UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Call control signalling for personal communications over interconnected metropolitan area networks Qian, Nanjian 1993

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1993_spring_qian_nanjian.pdf [ 5.39MB ]
Metadata
JSON: 831-1.0065103.json
JSON-LD: 831-1.0065103-ld.json
RDF/XML (Pretty): 831-1.0065103-rdf.xml
RDF/JSON: 831-1.0065103-rdf.json
Turtle: 831-1.0065103-turtle.txt
N-Triples: 831-1.0065103-rdf-ntriples.txt
Original Record: 831-1.0065103-source.json
Full Text
831-1.0065103-fulltext.txt
Citation
831-1.0065103.ris

Full Text

CALL CONTROL SIGNALLING FOR PERSONAL COMMUNICATIONSOVER INTERCONNECTED METROPOLITAN AREA NETWORKSbyNANJIAN QIANB.Sc., Xi-An University of Electronic Science and Technology, P.R.China, 1983A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF APPLIED SCIENCEinTHE FACULTY OF GRADUATE STUDIESDEPARTMENT OF ELECTRICAL ENGINEERINGWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAApril 1993© Nanjian Qian, 1993In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature) Department of "E( e c-t 1 cc, il/teRThe University of British ColumbiaVancouver, CanadaDate^ 1 9 ,^1 i'93DE-6 (2/88)AbstractPersonal communication services (PCS) are currently being developed to enable people to callother people regardless of their locations and movements. A distributed personal communication network(PCN) based on the IEEE 802.6 metropolitan area networks (MANs) has been recently proposed as ameans of supporting PCS. This network must be supported by efficient and reliable signalling, which hasnot been extensively examined and will be the main subject of this thesis.While pre-arbitrated (PA) slots are suitable for the transport of isochronous traffic in IEEE 802.6MANs, high priority, queue-arbitrated (QA) slots accessed by means of the distributed queue dual bus(DQDB) protocol are appropriate for the transport of signalling messages between the major networkelements. The use of bandwidth balancing and erasure nodes to improve the performance of the DQDBprotocol has been considered. As analytical models are not available, simulation models have beenimplemented to facilitate the evaluation of signalling performance in terms of signalling delays andsignalling traffic levels. Using these simulation models, the relationship between PA traffic capacityand QA signalling delay has been established, and the effects of alternative signalling architectures withdifferent bus configurations and locations of signalling network elements have been investigated. The bestsignalling performance is obtained in a closed bus configuration where the head of bus (HOB) stations forthe two buses are physically collocated and incorporate the functions of the bandwidth manager, while thefunction of the signalling termination is distributed among all the MAN nodes. Signalling performancecan be improved by employing erasure nodes to recover QA slots already read by their destinations. Theoptimal location for a single erasure node on each bus has been empirically determined by simulations.Logical procedures for call setup, clearing and handoff have been established for calls betweenmobile subscribers, through the access MANs and backbone MANs. The overlapping call setup procedure,which facilitates parallelism in call setup processing, has been proposed as a possible means of reducingcall setup delays.hi summary, call control signalling architecture, performance and procedures have been studiedin this thesis. The results further the viability of using the IEEE 802.6 MAN to support PCS.Table of ContentsAbstract^ iiList of Tables viiList of Figures^ ixAcknowledgments xiiChapter 1 Introduction^ 11.1 Background  11.1.1 Historical Perspective ^  11.1.2 A Distributed PCN Based on IEEE 802.6 MAN ^  41.2 Motivations, Objectives and Approach ^  61.3 Outline of The Thesis ^  7Chapter 2 Using MAN to Support PCS^ 92.1 Overview of IEEE 802.6 MAN  92.1.1 The IEEE 802.6 MAN Architecture ^  92.1.2 The DQDB Protocol ^  112.1.3 Capabilities to meet PCS Requirements ^  142.2 Modelling and Analysis of DQDB Protocol  162.2.1 DQDB Protocol Modelling ^  162.2.2 Performance of Priority QA Access ^  182.2.3 Using Bandwidth Balancing Mechanism in IEEE 802.6 MAN ^ 212.2.4 Using Erasure Nodes in IEEE 802.6 MAN ^  27iii2.3 Alternative Schemes for Supporting Voice Transport in IEEE 802.6 MAN ^ 292.3.1 Queue Arbitrated Voice Transport ^  302.3.2 Pre-Arbitrated Isochronous Voice Transport ^  32Chapter 3 PCN Architecture based on IEEE 802.6 MAN 363.1 Requirements for Management and Control Functions in PCN ^ 363.1.1 Mobility Management ^  363.1.2 Call Management  373.1.3 Handoff Management ^  373.2 PCN Architecture ^  383.2.1 Logical Architecture ^  383.2.2 Physical Architecture  393.3 Network Elements to Support PCS over MAN ^  403.3.1 Functional Descriptions ^  403.3.2 Physical Placement of ST and BWM^  423.3.2.1 Signalling Architecture 1 (SST, SBWM, CB)^ 433.3.2.2 Signalling Architecture 2 (SST, SBWM, CB)  443.3.2.3 Signalling Architecture 3 (MST, BBWM, OB) ^ 443.3.2.4 Signalling Architecture 4 (MST, SBWM, CB)  453.3.3 Performance Analysis ^  45Chapter 4 Signalling Procedures for Call Setup, Clearing and Handoff in PCN^514.1 Review of Signalling Procedures and Functions in Other Systems  524.1.1 Review of Call Setup/Clearing Procedures in ISDN ^  524.1.2 Call Handoff Functions in Cellular Networks  56iv4.2 Functional Elements of Call Setup, Clearing and Handoff in PCN ^ 584.2.1 Call Setup Functions ^  594.2.2 Call Clearing Functions  614.2.3 Call Handoff Functions ^  614.2.4 Definitions of Call Control Signalling Messages ^  624.3 Call Setup and Clearing Procedures ^  654.3.1 Stage 1: Call Setup and Clearing between Calling User and Access Network ^ 664.3.1.1 Sequential Call Setup Procedure ^  664.3.1.2 Overlapping Call Setup Procedure  694.3.1.3 Call Clearing Procedure ^  714.3.2 Stage 2: Call Setup and Clearing within Transit Network^ 724.3.2.1 Sequential Call Setup Procedure ^  734.3.2.2 Overlapping Call Setup Procedure  744.3.2.3 Call Clearing Procedure ^  754.3.3 Stage 3: Network to Callee Call Setup and Clearing^  764.3.3.1 Call Setup Procedure ^  764.3.3.2 Call Clearing Procedure  784.3.4 Effects of Different Signalling Architectures on Isochronous Channel Setup ^ 804.3.5 Performance of Call Setup Procedures ^  804.4 Call Handoff Procedures ^  83Chapter 5 Conclusions^ 925.1 Summary of the Results ^  925.2 Issues for Future Studies  93vBibliography^ 94Appendix List of Acronyms and Abbreviations^ 100Appendix Sample of Typical Data^ 104Appendix Sample of Source Code 107viList of TablesTable 1 Slot Format in IEEE 802.6 MAN ^  11Table 2 Definition of Call Management Functions ^  38Table 3 Relations Between Management and Control Functions and Network Elements ofPCN ^  42Table 4 Definitions of Call Setup Messages ^  63Table 5 Definitions of Call Clearing Messages ^  64Table 6 Definitions of Call Handoff Messages ^  64Table 7 Definitions of Facility Setup/Releasing Messages ^  65Table 8 Relationships between Type of Handoff, Change in Connectivity, and Facility SetupRequirements ^  86Table 9 Inter-MAN Call Handoff Delays vs. PA Throughput ^  91Table 10 Cell Transmission Delays at PA Throughput = 204Mbps ^ 104Table 11 Cell Transmission Delays at PA Throughput = 240.8Mbps ^ 104Table 12 Cell Transmission Delays at PA Throughput = 256.6Mbps ^ 105Table 13 Cell Transmission Delays at PA Throughput = 273.8Mbps ^ 105Table 14 Cell Transmission Delays at PA Throughput = 276.8Mbps ^ 106viiTable 15 Cell Transmission Delays at PA Throughput = 277.5Mbps ^ 106viiiList of FiguresFigure 1 Basic Architecture of MAN-based Personal Communication Network ^ 7Figure 2 IEEE 802.6 MAN Architecture ^  9Figure 3 IEEE 802.6 MAN Frame Format  10Figure 4 IEEE 802.6 MAN Protocol Layers ^  11Figure 5 Operation of The DQDB Protocol  13Figure 6 Mean Cell Transmission Delay vs. Normalized Offered Load (without erasurenodes) ^  19Figure 7 Normalized Throughput vs. Normalized Traffic Load (without erasure nodes) . . . 20Figure 8 Node Index vs. Mean Cell Waiting Delay (50 nodes, single priority level, offeredload=l.0) ^  23Figure 9 Node Index vs. Mean Cell Waiting Delay (10 nodes, single priority level, offeredload=1.0) ^  24Figure 10 Node Index vs. Mean Cell Waiting Delay (50 nodes, single priority level, offeredload=0.3) ^  25Figure 11 Node Index vs. Mean Cell Waiting Delay (10 nodes, 3 priority levels, offeredload=0.12 for each level) ^  25Figure 12 Mean Cell Waiting Delay vs. Node Index (10 nodes, 3 priority levels, offeredload=0.36 for each level) ^  26Figure 13 Comparison of Mean Cell Transmission Delays with and without Erasure Nodes 28Figure 14 Comparison of Normalized Throughput with and without Erasure Nodes ^ 29Figure 15 Node Index vs. Mean Cell Waiting Delay (10 nodes, single priority load, offeredload=l.0 with erasure nodes)^  30Figure 16 Mean Voice Cell Waiting Delay vs. Voice Throughput for QA Voice Transport ^ 32Figure 17 Mean Cell Transmission Delay vs. PA Throughput ^  35Figure 18 Mean Call Setup Delay vs. PA Throughput ^  35ixFigure 19 Logical Architecture of PCN ^  39Figure 20 Physical Architecture of PCN Sub-System ^  40Figure 21 Mean Call Setup Delay vs. PA Throughput for Different Signalling Architectureswithout erasure nodes, 90% confidence intervals indicated) ^ 46Figure 22 Call Setup Delay vs. Location of Erasure Nodes ^  48Figure 23 Effect of Erasure Nodes on Call Setup Delay (Signalling Scheme 3) ^ 48Figure 24 Effect of Erasure Nodes on Call Setup Delay (Signalling Scheme 4) ^ 49Figure 25 Mean Call Setup Delay vs. PA Throughput for Different Signalling Architectures(with one erasure node per bus, 90% confidence intervals indicated) ^ 50Figure 26 Comparison of Call Setup Delays, with and without Erasure Nodes ^ 50Figure 27 Call Setup/Clearing Procedures in ISDN^  53Figure 28 Process Interactions in the Call Setup Procedure ^  60Figure 29 Physical Partitioning of Call Setup Procedure  60Figure 30 Process Interactions in the Call Clearing Procedure ^  62Figure 31 Process Interactions in the Call Handoff Procedure  63Figure 32 Sequential Call Setup Procedure in Stage 1 ^  68Figure 33 Overlapping Call Setup Procedure in Stage 1  70Figure 34 Call Clearing Procedure in Stage 1 ^  72Figure 35 Sequential Call Setup Procedure in Stage 2 ^  74Figure 36 Overlapping Call Setup Procedure in Stage 2  75Figure 37 Call Clearing Procedure in Stage 2 ^  77Figure 38 Call Setup Procedure in Stage 3  78Figure 39 Call Clearing Procedure in Stage 3 ^  79Figure 40 Isochronous Channel Setup for Signalling Architecture 1 ^ 80xFigure 41 Isochronous Channel Setup for Signalling Architecture 2 ^ 81Figure 42 Isochronous Channel Setup for Signalling Architecture 3  81Figure 43 Isochronous Channel Setup for Signalling Architecture 4 ^ 82Figure 44 Mean Call Setup Delay vs. PA Throughput for Sequential Call Setup^ 83Figure 45 Mean Call Setup Delay vs. PA Throughput for Overlapping Call Setup ^ 83Figure 46 Comparison of Call Setup Delays between Sequential and Overlapping Call Setup 84Figure 47 Intra-MAN Handoff with Intra-MAN Connectivity^  86Figure 48 Inter-MAN Handoff with Different Changes in Connectivity ^ 87Figure 49 Intra-MAN Call Handoff Procedure ^  88Figure 50 Inter-MAN Call Handoff Procedure  90xiAcknowledgmentsI would like to express my sincere gratitude to my research supervisor, Dr. V.C.M. Leung,for his enthusiastic guidance, valuable suggestions and many helpful comments. I extend my sincereappreciation to Dr. A. Malyan for his help during the time he was in the department. I would also liketo give special thanks to my beloved wife,Yu Zhang, and my family for their consistent encouragementand support. Many thanks to the staff and friends in this department for their kindness and help inproviding me a pleasant place to stay during my studies at UBC. This research was supported by theCanadian Institute for Telecommunications Research through a Research Assistantship provided by Dr.R.W. Donaldson. His continued interest in this research is gratefully acknowledged.XII1Chapter 1. IntroductionRecently, advances in communication technologies have significantly reduced the distancesbetween people living in all parts of the world, by enabling them to keep in touch wherever telephones arelocated. Increasingly, people are demanding the ability to keep in touch regardless of the locations of fixedtelephone sets, and to exchange an increasing variety of information, such as data, images and video, inaddition to voice. Continued developments of broadband integrated services digital networks (B-ISDNs)and personal communication networks (PCNs) are intended to satisfy these demands. PCNs are intended"to provide ubiquitous communication coverage, enabling people to call other people irrespective of thegeographic location of either party" [1], and will employ wireless access technologies evolved frompresent cellular mobile telephone networks. While the first generation PCNs may support only narrow-band voice and data personal communication services (PCS), future generation PCNs are expected tosupport increasingly wide-band traffic, thus converging to the service capabilities of B-ISDN.In this introductory chapter, the evolution of wireless communication systems, the currentdevelopment of PCN and the new research related to distributed personal communication networkarchitecture based on the IEEE 802.6 MAN are reviewed in the first section. The objectives andmotivations of this thesis are explained in the second section. The approach that is employed to achievethe objectives is also outlined. Finally, the structure of this thesis is given in the last section.1.1 Background1.1.1 Historical PerspectiveThe first radio telephone services were introduced in the high-frequency radio band in 1921 [2].This kind of system was manually operated and used to support international telephone services, as wellas communications with ships and aircrafts. Immediately after the Second World War, a new productcombining a radio transmitter and receiver (the transceiver) operating over VHF (very high frequency)was developed [2]. First used in such services as the police, the fire brigade, and taxis, later developmentled to their use as full "radio telephones", connected to the public switched telephone network (PSTN)Chapter I. Introduction^ 2for the receipt and generation of ordinary telephone calls [3]. Later, a new technology emerged. It useda radio base station located on a hill and equipped with a powerful multi-channel radio transceiver tocommunicate with many mobile stations over a wide area. Several base stations were connected to amobile switching center (MSC) which provided an interface to the PSTN. The mobile stations were verybulky at that time. This kind of system grew rapidly in popularity in the mid-1970s [3]. However, theseradio telephone systems were limited by their small user capacities, because they were designed to providewide area coverage using a limited number of radio channels. Another drawback for these systems wasthat they required the caller to specify the radio telephone system providing coverage to the called mobileto setup the call. In-band signalling methods had been commonly used in these radio telephone systems.Continued development of radio communications results in the emergence of the cellulartelephones and residential cordless phones in the early 1980s [2]. Both products have experiencedrapid growth in public acceptance. These products, often described as first-generation systems, arecharacterized by analog frequency modulated voice transmissions, and increased system capacity overprevious radio systems.Cellular mobile telephone networks make efficient use of the radio spectrum through frequencyreuse. The coverage of each radio base station, called a cell, is very much reduced, so that widearea coverage is provided by many cells. Cells not immediately adjacent to each other may share thesame radio channels while keeping co-channel interference under control. Several cells, through theirrespective base stations, are managed by a MSC. Cellular networks and their MS Cs have more significantprocessing and signalling capabilities compared to previous radio telephone systems. An ongoing call isautomatically handed off from one cell to the next, through the allocation of a new radio channel, whenthe mobile terminal crosses a cell boundary. The ability of the mobile terminal to originate and receivecalls is maintained over networks spanning vast areas through the roaming capability. The networkcontrol signalling sub-system, however, imposes a severe limitation on these cellular systems. Duringthe processing of a call, all communications between a mobile terminal and the network infrastructureare confined to a single voice channel. Network control messages for handoff or power control, forChapter 1. Introduction^ 3example, are signalled over the voice channel on a "blank and burst" basis. Because control messagesinterrupt speech transmission and produce objectionable audible clicks, signalling must be reduced to abare minimum, so that the network control capacity of a cellular system is relatively limited.In cordless telephony, the main limitation is vulnerability to interference. Most cordless phoneshave access to only one radio channel. If two people in the same area have cordless telephones thatoperate over the same channel, their cordless telephone sets can not be used at the same time. Theoperating range is limited to tens of meters from a single base station [4].To overcome the limitations of first generation technology and to satisfy the exploding demandfor wireless access to communication networks, industry has been working on second generation cordlessand cellular technology. At present, several second-generation systems are being developed aroundthe world [5][6][7][8][9]. All of them employ digital voice transmission. Second-generation cellularsystems will have dedicated signalling channels for the exchange of network control information betweenmobile terminals and the network infrastructure. Cordless phones will have access to many channels,automatically selecting the best available channel at the beginning of a call and, in some cases, switchingto a new channel as conditions change during the call.Five specific second-generation systems have received extensive attention in recent years [6].There are three cellular standards including GSM, a Pan-European digital cellular mobile telephone system[10][7][11], IS-54, an emerging North American digital cellular mobile telephone standard [12], and theyet-to-be-named Japanese standard. In addition, there are two approaches to enhanced cordless telephony:CT2, a British standard [13], and DECT, digital European cordless telephone [14][15], under considerationby the association of European telephone operating companies. Furthermore, Bell CommunicationsResearch has proposed Universal Digital Portable Communications [16] as a means of delivering basicservices to telephone company subscribers. The main drawback of second generation systems is thatthey are not compatible with each other, because each system is designed to provide a different setof services over a specific environment. These systems still employ centralized network architecturesand centralized methods for call control procedures and call handoff procedures where call handoffs areChapter 1. Introduction^ 4supported [17][18][19]. Nevertheless, it is expected that by the end of the century, demand for wirelessservices will overtake the capacity of second-generation technology [6]In the upcoming third generation system [1][20][21][22], the distinction between cordless phonesand cellular mobile telecommunications may disappear. Lightweight, inexpensive and portable terminalswill give people access to worldwide integrated information networks. The challenges of third-generationwireless access are to create networking techniques that are capable of the following [6]:carrying many types of information;being integrated with current and future wireline communication networks;serving a mass market in urban areas;operating efficiently in sparsely populated areas;• operating indoors, outdoors, and within vehicles;• serving stationary terminals and terminals moving at high speed;Third generation wireless access technologies will be integrated into PCNs to provide accessto advanced information services. The PCNs will accommodate a diverse set of information sources(voice, video, fax, image, data etc.), enable users to initiate and receive calls regardless of their locationsand movements, operate in all transmission environments (wireless and wireline), and be capable ofserving very large user populations. A third generation system should embrace the integration of severalcommunication concepts, approaches, or systems into one interconnected and interworking network. Thisintegrated network should support several different tetherless communications devices optimized for theirspecific environments and should also include wireline communications. Today, researchers and plannersare devoting increasing attention to new architectures and technologies for third generation personalcommunication systems [23][24][20]. These include micro-cellular architecture, distributed processingfor call control and call handoff and advanced intelligent network capabilities. New ideas and newtechnologies are still emerging.1.1.2 A Distributed PCN Based on IEEE 802.6 MANThe main thrust of the Canadian Institute for Telecommunications Research project on "Inter-Chapter I. Introduction^ 5networking of wireless networks" being carried out at U.B.C. is the design and evaluation of a third-generation, microcellular interconnection architecture for personal communication networking based onthe IEEE 802.6 MAN [23]. The design criteria are as follows: (1) making use of current wireline pub-lic network facilities (i.e. MANs) to reduce the total costs of new system deployment; (2) employingdistributed architectures and procedures for call control to accommodate the increased processing load re-sulting from the growing demand for wireless communications; (3) compatibility with the future B-ISDNs.The system should have the capability of providing a wide range of services (voice, video, data etc.).The proposed design develops and takes advantage of the distributed circuit switching capabilityof the IEEE 802.6 MAN to provide the infrastructure for extending connection oriented PCS to users withfixed, portable or vehicular mobile terminals via micro-cells. The MAN architecture is being enhancedto provide call setup, clearing and handoff control capabilities, and speech activity-switched isochronoustransport capability, to support PCS. MANs will be an important element in the evolution towards B-ISDN, initially functioning as broadband gathering networks through the interconnection of local areanetworks (LANs). Due to similar cell structures of the 802.6 MAN and the Asynchronous Transfer Mode(ATM) employed in B-ISDN, the interworking of B-ISDNs and MAN-based PCNs will be simplified.In the new architecture as shown in Figure 1, MAN nodes have the interfaces to connect theappropriate base stations (BSs), LANs and private branch exchanges (PBXs). Within a MAN, there is abandwidth manager (BWM) which allocates isochronous channels for connection oriented services suchas voice and video, a signalling termination (ST) which processes the call control signalling messages,a trunk gateway (TGW) to interconnect the MAN with other public networks and a database (DB)containing the PCS profile of each subscriber within the MAN coverage area.It is sometimes necessary to use several MANs to cover an entire metropolitan area (MA).Initially, the relatively few MANs deployed in the MA could be interconnected via homogeneous networkbridges. As the number of MANs deployed rise however, network manageability will necessitate theuse of a centralized higher-speed multiport relay—a B-ISDN switch [23]. Since the existing connection-oriented telecommunication networks will continue to exist for many years, an interworking unit (IWU)Chapter 1. Introduction^ 6at this switch should have the functionality to support both ATM and synchronous transfer mode (STM)switching fabrics. Additionally, a backbone MAN could interconnect these access MANs to form acluster. The IWU switching nodes would then be connected to the backbone MAN(s) in a MA. Ifneighboring access MANs within a cluster are also bridged, call setup and interMAN handoff wouldbe simplified, and reliability against bridge failure would be enhanced. To effectively utilize this newPCN architecture, it is necessary to develop architectures, protocols and procedures for network controlsignalling to support PCS over the network.1.2 Motivations, Objectives and ApproachWhile the distributed MAN-based microcellular interconnection network architecture has beenshown to be effective in supporting future PCS in terms of its user traffic transport and signallingcapabilities [23][24] and its support of distributed database structures to facilitate callee location [25],detailed design and in-depth evaluation of call control signalling are lacking, and are the main subjectsof this thesis. Design and evaluation of call control signalling are essential elements in the investigationof the capabilities and characteristics of the proposed distributed internetworking architecture.The objectives of this thesis are:i^to design signalling network architectures which optimize signalling performance with respect to thelocations of signalling network elements;ii. to determine the network capacity by establishing the relationship between signalling performanceand the level of user traffic;iii. to design signalling procedures for call setup, clearing and handoffs over access MANs interconnectedby backbone MANs, and evaluate the design alternatives.The realization of these objectives depends on the capability to analyze the performance ofthe queue-arbitrated (QA) distributed queue dual bus (DQDB) protocol employed by the IEEE 802.6MAN. As analytical models are not available, we resort to discrete event simulation models, to gaina better insight about the general behaviors of the DQDB protocol, and more specifically, to evaluatesignalling performance relative to user traffic levels and signalling network element locations. SubstantialTo PBX To LAN, BSsAccess MAN`Backbone MANTo B-ISDN or PSTC ATM To other Backbone MANIWU STMAccess MAN`Backbone MANChapter 1. Introduction^ 7computation efforts are required to simulate individual MANs, such that simulations of interconnectedMANs are not achievable. The study of signalling procedures is based on qualitative evaluation ofsignalling diagrams, and using typical signalling delays from individual MANs to approximate end-to-end signalling performance.1.3 Outline of The ThesisIn chapter 2, the IEEE 802.6 MAN (DQDB MAN) is reviewed. The DQDB protocol and its twoFigure 1. Basic Architecture of MAN-based Personal Communication NetworkChapter 1. Introduction^ 8network access methods (PA and QA) are evaluated. Two alternative schemes (bandwidth balancing anderasure node) are considered for enhancing the performance of the DQDB protocol. Different methodsfor user traffic transport are considered, and the relationship between signalling delay and user trafficlevel is determined. In chapter 3, the network requirements and the logical/physical architecture at theMAN-based PCN are defined, and the functionality of each network elements is specified. The optimumpositions and structures for the major network elements in the MAN-based PCN are examined. Inchapter 4, several alternative call control procedures for call setup, clearing and handoff are proposedand examined. Finally, conclusions are drawn in chapter 5.9Chapter 2. Using MAN to Support PCSTo provide a framework for designing and evaluating control signalling architecture, performanceand procedures in the proposed distributed PCN based on the IEEE 802.6 MAN, the MAN architecture,protocol mechanisms and its capabilities to meet PCS requirements are reviewed in this chapter. Thegeneral performance of the DQDB protocol for QA access is examined, by means of simulation modelling,with respect to access priorities, bandwidth balancing and the use of erasure nodes, and alternative schemesfor voice transport over the IEEE 802.6 MAN are evaluated.2.1 Overview of IEEE 802.6 MAN2.1.1 The IEEE 802.6 MAN ArchitectureThe IEEE 802.6 MAN is a distributed communications network which provides integrated circuitswitching for isochronous traffic (such as voice and video) and packet switching for asynchronous traffic(such as data and signalling). Its capacity is dynamically allocated to meet the demand of the isochronoustraffic. Capacity that is not allocated to isochronous traffic can be shared by the asynchronous traffic.Its topology comprises two uni-directional buses for communications in opposite directions as shown inFigure 2. Every node of the network is attached to both buses allowing full duplex communicationsbetween any pair of nodes. The head of bus (HOB) stations on both buses can be collocated if the buseshave a closed configuration, i.e. they are looped to form a physical ring. Otherwise, the HOB stationsare separate on open buses.PA slot PADQA slot QA slotPA slot 000frameheader 000Chapter 2. Using MAN to Support PCS^ 10The transmission format is slotted. Each slot is a protocol data unit transferred between adjacentnodes along a unidirectional bus. Transmissions on each bus employs a timing structure shown inFigure 3 with 125,us frames which comprise a given number of slots shared between isochronous andnon-isochronous communications. Frames and slots are generated on each bus by the HOB station.Transmissions from each station must synchronize with this timing structure as it propagates down thebus through the point at which the station is attached to the bus.2^53^53^ octets^ 125x 1 0-6s^Figure 3. IEEE 802.6 MAN Frame FormatThe slot length has been defined as 53 octets which contains a one octet Access Control Field(ACF), 4 octets of segment header and 48 octets of segment payload, as shown in Table 1. The ACFcontains five fields which control slot access. The length of each field in bits is shown in Table 1. Reservedbits are set to 0. The BUSY bit indicates whether the slot contains information (BUSY=1) or does notcontain information (BUSY=0). The SL_TYPE bit indicates whether the slot is a QA slot (SL_TYPE=0)or PA slot (SL_TYPE=1). The Previous_Slot_Read (PSR) bit indicates whether the segment in theprevious slot has been read (PSR=1) or has not been read (PSR=0). The REQUEST field contains threeREQ bits, used for QA access at the three priority levels. The PA/QA segment header contains 4 fields.The twenty bit Virtual Channel Identifier (VCI) provides a means to identify the virtual channel to whichthe PA/QA segment belongs. There is a single VCI space which is shared by all services. The VCIvalue corresponding to all bits being set to one is the default value for the connectionless medium accesscontrol (MAC) service. The two bit Payload_Type field indicates the nature of the data to be transferred.The two bit Segment_Priority field is reserved for future use. The eight bit Header Check Sequence(HCS) field provides for detection of errors and correction of single-bit errors in the PA/QA SegmentHeader. This structure is similar with the cell structure defined in ATM network for future B-ISDN.Chapter 2. Using MAN to Support PCS^ 11Table 1. Slot Format in IEEE 802.6 MANAccess Control Field(1 octet)PA/QA Segment Header(4 octets)SegmentPayloadBusy SL_TYPEPSR RESER-VEDREQUEST VCI PAYLOADTYPESEGMENTPRIORITYHCS 48 octets1 bit 1 bit 1 bit 2 bits 3 bits 20 bits 2 bits 2 bits 8 bits2.1.2 The DQDB ProtocolAs shown in Figure 4, the IEEE 802.6 standard specifies the Physical and Distributed QueueDual Bus (DQDB) protocol layers (which is the data link layer) for a Metropolitan Area Network.In the physical layer, the standard supports CCITT G.703 [26] at 34.368 and 139.264 Mbps,ANSI DS3 [27] at 44.736 Mbps and CCITT G.707-9 [28] SDH at 155.520 Mbps. In this thesis, weassume the use of CCITT G.707-9.Figure 4. IEEE 802.6 MAN Protocol LayersChapter 2. Using MAN to Support PCS^ 12The DQDB layer uses the services of the physical layer to provide three major services: theconnectionless service, the connection-oriented service and the isochronous service. The connectionlessservice is consistent with the medium access control service defined in other IEEE 802 local area networks.It supports the transfer of variable length messages between peer entities without the need for theestablishment of a connection between them. The connection-oriented data service supports a virtualchannel between a pair of users at either ends of the connection. The details of this service are stillunder study by the 802.6 standards committee. The isochronous service supports a fixed bandwidthtransfer between two users over an already established isochronous connection. The DQDB layer can besubdivided into the adaptation sub-layer and the MAC sub-layer.The Adaptation Layer provides two major functions: the Segmentation/Reassembly (SAR)function and the Convergence function. The SAR function is used to support variable length messagesfor connectionless and connection-oriented data services. It fragments an initial MAC protocol dataunit (IMPDU) into one or more fixed segmentation units (44 octet for each segmentation unit) at thesource, and creates corresponding derived MAC protocol data unit (DMPDU) from each segmentationunits. These DMPDUs are transmitted along the network using QA slots and reassembled into an originalIMPDU at the destination. The Convergence function is used to support different IEEE 802.6 servicesmentioned above. Convergence functions in each IEEE 802.6 service are described in [29].The MAC layer supports a pre-arbitrated (PA) access method and a queue-arbitrated (QA) accessmethod to support isochronous and non-isochronous IEEE 802.6 services, respectively.The PA access method pre-allocates some of the slots within each frame for the transport ofisochronous traffic. Individual octets within a PA slot may be allocated to different isochronous trafficsources through the specification of a slot offset. The HOB node at each bus has the responsibility towrite a VCI into each PA slot and ensures that PA slots for such a VCI value are provided periodicallyat a rate that matches the required bandwidth for the isochronous service.The QA access method employs the DQDB protocol, a medium access protocol that employs adistributed queue to control the access to the payload of QA slots on each of the two buses. The protocolChapter 2. Using MAN to Support PCS^ 13Figure 5. Operation of The DQDB Protocolprovides statistical multiplexing with non-contention access control for bursty traffic. The operation ofthis protocol is fundamentally different from other MAC protocols in that information about the queueingstate of the sub-network is explicitly kept in each node to regulate network access.The operation of the basic DQDB protocol for QA access on the forward bus within each prioritylevel is illustrated in Figure 5, with Bus A being the forward bus and Bus B the reverse bus. An identical,but independent arrangement applies for access to Bus B as the forward bus. Each slot on the buses(whether a QA or PA slot) contains an ACF as shown in Table 1 which includes a Busy bit and a Requestfield of three REQ bits, one for each priority level. The Busy bit indicates whether or not the slot is usedon the forward bus. The REQ bits are used for signalling on the reverse bus to upstream nodes when aQA segment for the given priority has been queued for forward bus access at a downstream node. Twocounters for each priority level are employed in each node to track the status of the distributed queue forthe forward bus. The request counter (RQC) counts the number of outstanding requests from downstreamnodes. It is incremented each time a request passes the node on the reverse bus, and decremented for eachfree slot passing the node on the forward bus until it has reached a minimum count of zero. The countdown counter (CDC) remembers the number of segments queued ahead of a segment to be transmittedChapter 2. Using MAN to Support PCS^ 14at a node. If a node has a segment to transmit, it sets the REQ bit of the required priority level in aslot passing by on the reverse bus to register its request at all upstream nodes, transfers the current RQCvalue to the CDC, and resets the RQC counter to zero. The CDC is now decremented for each passingfree slot on the forward bus until it reaches zero, whereupon the node transmits the segment in the nextfree slot on the forward bus. While waiting for access, the RQC continues to be incremented for anynew request seen on the reverse bus. This counting operation establishes a single queue of segmentsaccessing each bus across the network. Three levels of access priority are specified in the standard. Theyare realized by operating separate queues for each priority level. At each node, The RQC and CDCoperating at a particular priority level record all queued segments at equal and higher priorities. Thisallows segments to gain access as soon as capacity becomes available and guarantees that priority isalways given to segments in higher level queues.A detail specification of the DQDB protocol is given in [30].2.1.3 Capabilities to meet PCS RequirementsPersonal communications services (PCS) have received widespread attention, and it is predictedthat the potential market for PCS is very large. The objective of PCS is to provide ubiquitouscommunication coverage, enabling people to call other people irrespective of their locations or movements.This objective necessitates the following major network functions:• Wireless Access — to give subscribers freedom of movement;• Mobility Management — to enable tracking of subscriber locations between calls and making thisinformation available for call processing;• Call Management — to enable call setup and call clearing regardless of the locations of caller andcallee;• Handoff Management — to enable maintenance of ongoing calls when the mobile callers or calleescross the boundaries of radio coverage areas (cells).To reliably and efficiently provide the above functions, the network infrastructure must incor-porate the following capabilities:Chapter 2. Using MAN to Support PCS^ 15• easy accessing and updating of network databases;• user traffic transport for connection-oriented and connectionless, isochronous and asynchronous PCS;• high speed transport of signalling traffic, preferably compatible with common channel signalling;• high speed processing of signalling and database information under very high traffic load;• interworking with existing and future networks, such as B-ISDN, to enhance connectivity.The above network capabilities are readily realized in a distributed network architecture basedon the IEEE 802.6 MAN, as discussed below.• The network architecture enables the distribution of call control process and other network functionsamong several nodes within the network, thus reducing the processing load of each node as comparedto centralized processing. Distributed processing is beneficial to both call management and handoffmanagement.• IEEE 802.6 MAN supports three priority levels for QA access. Common channel signalling can beprovided by reserving the highest priority QA access for signalling. This signalling channel can beshared by all the nodes in the MAN. If sufficient capacity is reserved for this signalling channel so thatit is not overloaded, other traffic in the MAN will have little impact on the signalling performance.• The topology of the IEEE 802.6 MAN enables broadcasting/groupcasting capability within thenetwork. This feature enables the implementation of a distributed hierarchical database managementsystem over the network [23] easing database accesses, searches, and updates.• The PA access method provided by the IEEE 802.6 MAN readily accommodates isochronous usertraffic such as voice and video. Traffic capacity is further increased by modifying the PA accessmethod to allow data rate variations, such as during and between talk spurts in voice calls.Data or other asynchronous traffic can be transported over connection-oriented or connectionless QAchannels at a lower priority than signalling.Since the 802.6 MAN slot structure is compatible with the ATM cell structure used in B-ISDN,interworking from MAN to B-ISDN is simplified.Chapter 2. Using MAN to Support PCS^ 16From the above discussions, it is apparent that the MAN-based distributed PCN can meet the basicneeds of PCS. However, it remains to explore the performance of the DQDB protocol with respect toits signalling and user traffic perfonnance.2.2 Modelling and Analysis of DQDB Protocol2.2.1 DQDB Protocol ModellingThere are two approaches to model the DQDB protocol: analytic modelling and discrete eventsimulation. The analytic modelling and performance analysis of the DQDB protocol is well known tobe a very difficult problem. The reason for this is the high degree of interactions among a plethoraof processes that makes an exact analysis of the protocol almost impossible. Hence, only approximateanalytical models are available [31][32]. As these models are not detailed enough to give a completeinsight of DQDB behavior, simulation modelling is employed in our investigations.The basic topology of the simulation model consists of two buses on the MAN. The numberof nodes (which corresponds to the network access points) and the distance between adjacent nodes areselectable within the MAN. The HOB nodes on the two buses, which may he separate or collocated,have the function of generating repeating frames of empty PA and QA slots. The number of PA slots ineach frame depends on the number of voice calls in progress, and the remaining slots are available forQA access. QA traffic, including signalling, arriving at each node from a traffic generator, is temporarilystored in three local queues, one for each priority level, until it is able to join the distributed queueto access the MAN. The distributed queue at each node is implemented using the counter mechanismdescribed in Section 2.1.2. General performance enhancement features including the bandwidth balancingmechanism and the erasure node scheme are included and can be enabled or disabled according to thepurpose of the simulation. The model also incorporates basic PCN call control functions for call setup,call handoff and call clearing to facilitate evaluation of signalling performance.The primary objective of the simulation model is to enable investigation of DQDB protocolperformance. It is used to evaluate a number of potential architectures consisting of various locationsof network elements, and various traffic scenarios with different mix of voice and data traffic sources.Chapter 2. Using MAN to Support PCS^ 17It is also used to evaluate different call setup, handoff and clearing schemes in the MAN-based PCN.The model produces statistical measures of network utilizations for different traffic sources, throughputaverages in individual nodes, means and standard deviations of cell waiting and transmission delays(in different priority levels) in individual nodes and over the entire network, and means and standarddeviations of call setup delays, call handoff delays and call clearing delays. These statistics can be usedto evaluate the effects of changes in network topology, call control procedures, and traffic variations inthe network. The terminology used in the model attempts to reflect the actual network elements.The simulation programs are written in SIMSCRIPT 11.5 [30] which is commonly adopted forlarge system simulations. All simulations are executed on Sun Workstations. The simulation model isoptimized for speed, with each simulation process designed to minimize computations. Samples of thesource code are given in appendix C.The following assumptions are common to all the simulation results:• The data rate on each bus is 155 Mbps• The propagation delay is 5.085µs/km• The slot length is 53 octets with 5 octets for header• The payload length is 44 octets for QA slots and 48 octets for PA slots• The slot transmission time is 2.73tts• The nodes are equally spaced along the bus• The number of buffers available for each priority level at each node is 20• The cell length is constant and equal to 1 slotThe cell inter—arrival time at each node is exponentially distributedAll signalling message lengths are equal to 1 slot except the SETUP message (which uses 5 slots)The inter-arrival time of voice calls at each node is exponentially distributedThe cell and call arrivals are independent, identically distributed among all the nodes in the networkThe mean call holding time for each voice call is 1 minuteThe data rate of each voice call is 64KbpsChapter 2. Using MAN to Support PCS^ 18The parameters used in the analysis of the simulation results are defined below:• The cell transmission delay is the elapsed time from the moment when a cell is introduced at thelocal queue in the source node to the moment when the cell is entirely transmitted and arrives atthe destination node.• The cell waiting delay is the elapsed time from the moment when a cell is introduced at the localqueue in the source node to the moment that transmission of the cell has begun at the source node.• The call setup delay is the elapsed time from the moment when a call setup request is generatedat the source node to the moment when the call related facilities have been established along theintermediate network and the call setup request message has reached the destination node over theMAN.• The network capacity is the maximum bit rate in the network.• The throughput (indicating PA or QA) is the total number of payload bits transmitted through thenetwork per second.• The normalized throughput (PA or QA) is the ratio of the current throughput to the maximumnetwork throughput (network capacity minus overhead).• The offered load (PA or QA) is the total number of bits (including payload bits and overhead bits)generated in the network per second• The normalized offered load (PA or QA) is the ratio of the current offered load to the networkcapacity.2.2.2 Performance of Priority QA AccessFigure 6 shows the mean cell transmission delay for QA access at each priority level as afunction of normalized offered load. Each set of vertically aligned data points represents the resultsof one simulation run, in which an equal offered load is presented to the MAN at each priority level.The MAN is configured with 50 nodes uniformly spaced along the dual bus. The distance of adjacentnodes is 3 slot transmission time. A QA cell generated at each node is equally likely to have any othernode as its destination. When the offered load is below 20% of network capacity, no difference in cellChapter 2. Using MAN to Support PCS^ 19Figure 6. Mean Cell Transmission Delay vs. Normalized Offered Load (without erasure nodes)transmission delay is observed among the three priority levels. The cell transmission delay is around0.1 ms. However, as the load increases to 30% of network capacity at each priority level (at this pointthe total traffic load is reaching the network capacity), there is a distinct difference among the lowestlevel and the two higher priority levels. The mean cell transmission delay for level 0 (lowest priority)increases rapidly as the offered load increases, while the mean transmission delay for level 1 increasesrelatively slowly comparing with level 0. For level 2 (highest priority), the cell transmission delayremains relatively constant. When the offered load reaches 50% of network capacity at each prioritylevel, it becomes difficult to collect the cell transmission data for level 0, because most of the cells atthis level have been blocked at the same nodes by higher priority traffic. The mean cell transmissiondelay for the middle priority level increases sharply whereas that of the highest priority level still remainsrelatively unchanged. When the offered load increases to 100% of network capacity at each prioritylevel, increasing number of level 1 cells are blocked so that data collection for cell transmission delayalso becomes difficult, and mean cell transmission delay for level 2 begins to increase. However, furtherincrease in traffic load does not result in further increase in mean cell transmission delay for level 2. Thisphenomena is caused by the limited number of buffers available in the local queue for each priority levelChapter 2. Using MAN to Support PCS^ 20Figure 7. Normalized Throughput vs. Normalized Traffic Load (without erasure nodes)at each node. Arriving cells which find the buffers full are blocked. From a cell transmission delay pointof view, while keeping delay at a minimum, the MAN can accommodate a very high offered load frompriority 2 QA traffic, up to the network capacity. However, to maintain the same minimum delay, andat the presence of higher priority traffic, the offered load for the lower priority levels must be reduced incorrespondence with the reduced priority level. This is further illustrated by examining the throughputfor each priority level, as shown below.Figure 7 shows the normalized throughput at each priority level versus the normalized offeredload, with the same set of assumptions as before. From this figure, we see that throughput keeps pacewith offered traffic when normalized traffic load is below 0.28 in each individual priority level. However,when traffic loads increase at the same rate beyond 0.28 for each priority level, priority 0 traffic is beingsqueezed out in favor of traffic at the two higher priority levels, so that priority 0 throughput graduallydecreases to zero. Throughput for priority levels 1 and 2 continues to keep up with increasing traffic loaduntil the normalized traffic at each level reaches 0.4. Further increase in offered traffic cause priority1 traffic to be squeezed out by priority 2 traffic, so that priority 1 throughput decreases to zero whileChapter 2. Using MAN to Support PCS^ 21priority 2 throughput continues to increase with offered traffic until the network capacity is reached at atraffic/throughput value of approximately 1.The above results assume uniform traffic for each priority level at each node. Different resultsmay be obtained depending on several parameters as follows:• The fraction of total traffic load for each priority level;• The distance between adjacent nodes;• The distribution of traffic load among the nodes.The IEEE 802.6 MAN is known to have a fairness problem, such that under certain conditionsequitable access is not available to all the nodes. For example, it has been reported that if the distancebetween adjacent nodes is large, lower priority levels may be able to maintain large shares of total networkcapacity, decreasing the effectiveness of the priority scheme [33]. Furthermore, if there are some nodesgenerating heavy traffic in the network, they are able to consume most of the capacity even if the trafficis at a lower priority level [34].2.2.3 Using Bandwidth Balancing Mechanism in IEEE 802.6 MANUnfairness in QA access using the DQDB protocol occurs when the network propagation delayis larger than a slot transmission time P41[33][35], and favors the nodes nearest the HOBs. Problemarises in most MAN implementations where the network span (over 50km), transmission rate (155Mbps),and slot size (53 octets) allow many slots to be in transit between the nodes. The long propagation delaycauses defects in the distributed queue in that a request generated by a node near the middle of the busis not known to nodes near the HOB until many slots have passed by these nodes, so that late arrivalsat these nodes could use up many empty slots before the request is registered in these nodes to causean empty slot to be set aside for the node at the middle of the bus. Extremely unfair service can resultduring large file transfers. To overcome this problem, a bandwidth balancing (BWB) mechanism hasbeen standardized by the IEEE 802.6 working group [29].The BWB ensures fair sharing of bandwidth among nodes. The mechanism uses a BWB counterin each node, and a system parameter called the BWB_MODE. Whenever a node transmits a QA segmentChapter 2. Using MAN to Support PCS^ 22on one of the buses, the BWB counter for that bus is incremented by one, provided that the counter doesnot have a value of (BWB_MODE — 1). If the counter does have a value of (BWB_MODE — 1) whenthe QA segment is transmitted, the BWB counter is reset to zero and at the same time, the RQC/CDCcounters of all priority levels on the node are incremented by 1. Effectively, this causes the node to skipthe use of one empty QA slot on that bus. If the value of BWB_MODE is zero, the BWB mechanism forthat bus is disabled. In this way, the apportionment of capacity for the different nodes can be controlledand smoothed at the cost of wasting some network capacity.The performance of the BWB was evaluated by simulation. As before, we assume there are 50nodes uniformly located among the DQDB buses. The distance of adjacent nodes is 3 slot transmissiontime.Figure 8 shows the variations of mean cell waiting delays with respect to the positions ofindividual nodes in the network with a single priority traffic load which is close to the network capacity.It is shown that without using the BWB (i.e. BWB_MODE=0), the mean cell waiting delays are unfairamong nodes in the network, in that the nodes near the HOBs have lower waiting delays than the nodesclose to the center of the buses. The difference between the maximum and minimum mean cell waitingdelay among the individual nodes on the buses is about 0.14,u,s. The use of non-zero BWB_MODEvalues result in dramatic changes in the distribution of cell waiting delay among the nodes. Two extremecases exist. When the BWB_MODE=1, the unfairness is reversed so that the mean cell waiting delays atnodes closer to the HOBs are now longer than those at nodes near the center of the buses. On the otherhand, BWB_MODE=16 gives a delay distribution which has the same characteristics as the case withoutbandwidth balancing (BWB_MODE=0). It is therefore expected that the proper BWB_MODE value isbetween these two extremes. For the given network configuration, experimentations with the simulationmodel reveal that the optimal BWB_MODE value is approximately 12. In this case, the difference ofmean cell waiting delays among all nodes in the network is about 0.05,us, and is minimized relative toother BWB_MODE values.Figure 9 shows the effects of different BWB_MODE values in a network with 10 nodes uniformlyspaced along the DQDB buses, the distance between adjacent nodes being 10 slot transmission time. InChapter 2. Using MAN to Support PCS^ 23Figure 8. Node Index vs. Mean Cell Waiting Delay(50 nodes, single priority level, offered load=1.0)this case, BWB_MODE=15 gives the best performance with regard to fairness for mean cell waitingdelays. The difference of mean cell waiting delays among all nodes in the network is about 0.01 bus. Itis apparent that with different network configurations, the optimal BWB_MODE value will be differentand result in different levels of improvements in the fairness behavior of the network. A higher level ofperformance improvement seems to be possible for networks with smaller number of nodes and shorterbuses.So far, we have shown that the BWB mechanism is effective for heavy traffic loads. However,the BWB mechanism is not useful under light traffic loads, as illustrated in Figure 10 which is based onthe same network configuration as Figure 8. When a non-zero BWB_MODE is used, not only does thefairness of cell waiting delay fail to be improved, but the actual mean cell waiting delay for each node isalso increased. The reason is that some empty slots are wasted due to the use of the BWB mechanism.However, in this case the mean cell waiting delays for all nodes may be small enough that fairness isnot an important concern. The above observation is true even when light traffic loads are offered atall three priority levels, as shown in Figure 11 for a network with the same configuration as the oneChapter 2. Using MAN to Support PCSconsidered in Figure 9. In this case, unfairness is evident of priority levels 1 and 2, whereas the meancell waiting delays for priority 0 traffic are relatively equal among all nodes. The reason for this is thatwhile the malfunction of the distributed queue counting mechanism caused by propagation delay giveshigh priority traffic generated at the nodes close to the HOBs an advantage in grabbing empty slots overthose generated at the nodes near the center of the buses, lowest priority cells do not have this advantagebecause they have to defer access in favor of higher priority cells.Figure 12 shows the mean cell waiting delays vs. node index in each priority level when thenetwork is fully loaded (i.e. each priority level provides up to 0.36 normalized offered traffic load, sothat the total load is beyond 100% network capacity). No matter what the BWB_MODE value is (i.e.BWB_MODE=0, 1 or 10), unfairness exists in each priority level, but with different behaviors. Forthe highest priority level, the cell waiting delays favor the nodes near the HOBs, while for the twolower priority levels, the cell waiting delays favor the nodes close to the center of the buses. Thisis an amplification of the phenomenon shown in Figure 11. As the higher priority traffic gains better24Chapter 2. Using MAN to Support PCS^ 25performance near the HOBs, it consumes a higher fraction of network capacity so that the throughput forChapter 2. Using MAN to Support PCS^ 26Figure 12. Mean Cell Waiting Delay vs. Node Index(10 nodes, 3 priority levels, offered load=0.36 for each level)the lower priority traffic is correspondingly reduced and delay increased.Therefore, the effectiveness of the BWB mechanism depends on the priority and load levelsChapter 2. Using MAN to Support PCS^ 27of the offered traffic. It is apparent that the BWB mechanism is not effective for light traffic loads orlow priority levels, but it is effective only for the highest priority traffic which also presents a heavyload to the network.2.2.4 Using Erasure Nodes in IEEE 802.6 MANIn the original DQDB protocol, a slot could only be used once to transmit information betweenany two nodes in the network. After the receiving node copies a cell addressed to it, the slot continues tocarry that cell until it reaches the end of the bus. Therefore, the protocol makes inefficient use of networkbandwidth. By introducing erasure nodes into the network, slots are allowed to be reused by other nodesonce the cells they carry have been copied by their destination nodes. The motivation behind the use oferasure nodes is to increase the network capacity and reduce the cell access or waiting delay.The basic mechanism of erasure nodes is as follows. All the nodes in the network observe slotspassing by on the two buses of the MAN, making local copies of the slots so as to introduce as littlelatency as possible. Each node identifies whether it is a recipient of the cell carried by each slot or not.The recipient node(s) flags the slot used by setting the previous segment read bit (PSR) in the subsequentslot. Special nodes called erasure nodes are selectively placed on the buses. Their purpose is to lookfor those slots that have been marked used and, if certain conditions are satisfied, erase the carried cellswhile maintaining the correctness of the network logic. Detail implementations are discussed in [36][37].Here, we use the erasure node scheme described in [36]. This scheme has the advantage of guaranteeingfair sharing of the extra capacity created through slot erasures, and erasing corresponding request bitspropagating up stream on the opposite bus as erased slots are made available to satisfy these requests.Figure 13 shows the relationship between mean cell transmission delay and equivalent normalizedoffered load in each priority level with and without erasure nodes. The network configuration is the sameas in Figure 6. One third of the nodes in the network are erasure nodes which are uniformly spacedalong the buses. The figure shows that equal cell transmission delays are maintained for both high andlow priority traffic up to a normalized offered load of 25% when erasure nodes are employed, comparedto 20% without erasure nodes. At an offered load of 30% network capacity, the mean cell transmissionChapter 2. Using MAN to Support PCS^ 28Figure 13. Comparison of Mean Cell Transmission Delays with and without Erasure Nodesdelays for priority 0 traffic are 0.26ms, and 0.6ms, with and without erasure nodes, respectively. When theoffered load increases to 50% network capacity, the mean cell transmission delays for priority 1 trafficare respectively 0.28ms and 0.49ms, with and without erasure nodes. When the offered load reaches100% of network capacity, the mean cell transmission delays for priority 2 traffic are 0.12m and 0.23ms,respectively, with and without erasure nodes. Thus, we see substantial reductions of mean cell waitingdelays at all priority levels when erasure nodes are incorporated in the network.Correspondingly, the use of erasure nodes also substantially increases the maximum throughputobtained at each priority level when the offered load is increased in equal parts for the three prioritylevels, as shown in Figure 14. The throughput of the highest priority traffic can exceed 100% of networkthroughput, reaching a maximum normalized throughput of 1.25. When offered load in all priority levelsare high (greater than 1), the two lower priorities can still maintain a normalized throughput of about0.1 when erasure nodes are used, compared to a normalized throughput of zero when no erasure nodesare used.Another advantage of using erasure nodes is the improvement in access fairness, as illustrated bythe relatively constant mean cell waiting delays relative to the node index in Figure 15. It is apparent thatChapter 2. Using MAN to Support PCS^ 29Figure 14. Comparison of Normalized Throughput with and without Erasure Nodesthe use of erasure nodes is more effective than the BWB mechanism in overcoming access unfairness in aMAN carrying multi-priority traffic loads. Erasure nodes have the effect of making empty slots availableat various locations along the buses. They favor QA access from nodes immediately down stream ofeach erasure node in the same manner that the nodes closest to the HOBS are favored for QA access. Ifa large number of erasure nodes are uniformly spaced along the buses, as in the case considered here,all nodes are favored and none of the nodes are at a disadvantage for QA access. Therefore the fairnessproblem in QA access is effectively eliminated.Using erasure nodes is not without disadvantages, however. Transmission latency and hardwarecomplexity are both increased in the nodes with erasure capability. Therefore, if the number of erasurenodes in the network is carefully selected to balance increased network throughput against increasedminimum mean cell waiting delay and increased equipment cost, the total performance of networkthroughput and the mean cell waiting delay can be improved.2.3 Alternative Schemes for Supporting Voice Transport in IEEE 802.6 MANPCS will include a wide variety of services such as voice, video, and data, etc. However, voiceChapter 2. Using MAN to Support PCS^ 30Figure 15. Node Index vs. Mean Cell Waiting Delay(10 nodes, single priority load, offered load=1.0 with erasure nodes)service is and will remain the most important among all the different PCS. It is therefore of criticalimportance to investigate the provision of voice service and the transport of voice traffic in future PCNs.In the following, several schemes for transporting voice in the IEEE 802.6 MAN are examined.2.3.1 Queue Arbitrated Voice TransportOne possible scheme to transport voice over the MAN is to transmit voice cells by means ofQA access. In this scheme, control signalling is sent by QA access at the highest priority, while voicetraffic is sent by QA access at second priority level, with each voice source accessing one full slot witha 44 octets payload at a time. The access employs the same frame length (5.5ms) as used in the airinterface. QA access is analogous to packet switching where statistical multiplexing recovers unusedcapacity in the silent periods between talk spurts. It also has the advantage of flexibility and simplicityfor the integration of traffic over the MAN.The main disadvantages are the non-uniform delay performance with respect to the locationof the access node, and the delay variations relative to the offered traffic load. Since voice quality isChapter 2. Using MAN to Support PCS^ 31sensitive to the variations of cell transport delays, which depend on factors unrelated to the volume ofvoice traffic, it is difficult to guarantee the voice user's quality of service using this scheme.The performance of this scheme has been evaluated using the simulation model describedpreviously. In this case, the MAN is configured with 50 nodes with the distance between adjacentnodes being 3 slot transmission time. Call control signalling is transmitted using priority 2 QA accessin the MAN, while voice traffic is transmitted using priority 1 QA access. It is assumed that each voicecall consume 64Kbps and the voice activity factor is 0.435. In order to evaluate the voice cell delayperformance for the QA scheme, an individual voice call tracking scheme is used. The description of thisscheme is as follows. For each voice call, a corresponding dynamic record is used to monitor the statusof this voice call. The arrival time and completion time for a particular voice call are pre-scheduled byusing exponential random number generators. The originating node and the terminating node are alsopre-selected by using random generators. If the voice call is active, the system generates appropriatevoice cells regularly, once every 1 lms from the originating node and from the terminating node, andtransmits the cell using priority 1 QA access over the MAN. When the call reaches the completingtime, the system stops generating voice cells for this call and activates the corresponding call clearingprocess. This individual call tracking scheme has the disadvantage of consuming a lot of computationtime. Since every individual call has to be monitored during the simulation, a large amount of computertime is used for changing status of individual calls. Figure 16 shows the mean voice cell waiting delayvs. voice throughput for this scheme. With increasing throughput, the voice cell waiting delay firstincreases gradually, until the total throughput reaches the maximum network throughput for this scheme(i.e. 310 x (44/53) = 257.36Mbps), at which point the delay increases sharply. By using this scheme,the maximum number of voice calls which can be accommodated by the network is around 4010 callswithout silence detection and 9220 with silence detection. The capacity consumed by total cell overheadis around 52.64Mbps. If only call setup and clearing signalling is considered, the maximum signallingtraffic in the network is around 0.88Mbps.Chapter 2. Using MAN to Support PCS^ 32Figure 16. Mean Voice Cell Waiting Delay vs. Voice Throughput for QA Voice Transport2.3.2 Pre-Arbitrated Isochronous Voice TransportThe transport of isochronous traffic (e.g. voice, video) over the IEEE 802.6 MAN is naturallysupported through the use of isochronous pre-arbitrated (PA) slots in the MAN. Control signalling anddata traffic, which is bursty in nature, is more suitable for QA access at appropriate priority levels.PA voice transport guarantees a constant throughput and constant cell delay for each voice call, and iscompatible with the existing digital PSTN based on time division multiplexing (TDM). However, themechanisms for allocating and de-allocation PA slots, necessary for supporting call setup and clearing invoice transport, have not been included in the existing IEEE 802.6 standards. Also, PA slots assigned toa voice call are wasted in the silent intervals between talk-spurts. Here, we review an extension of theIEEE 802.6 standard to enable the unused PA slots to be reclaimed for other traffic during silent intervals[23]. Network architecture and signalling protocols to support call setup and clearing are considered inchapters 3 and 4.For compatibility with TDM, 125ms frame length is usually used in PA schemes to provideisochronous services. Each frame consists of an integer number of slots with 48 byte payload, as before.As the cell payload is much greater than the 8—bit PCM voice sample transmitted every 125ms for eachChapter 2. Using MAN to Support PCS^ 33voice call, it is not economical to assign one slot in every 125,as frame for a single voice channel.Packing schemes have been proposed where up to 48 calls share each PA slot [38]. In the simpler, "first-fit" scheme, unoccupied octets or "holes" occur in voice slots as calls arrive and complete at differenttimes. Holes are eliminated in the "re-packing" scheme, where the octet assignments are reorganizedwhenever a call completion occurs. While overcoming the inefficiency of the "first-fit" scheme, thecomplexity of managing the "re-packing" scheme is considerable and may be unacceptable.Alternately, each PA voice slot can be used for only one voice source, necessitating a longerframe duration to fill each voice slot. This scheme is preferred for its simplicity and manageability,and is assumed for all subsequent discussions in this thesis. Here, we assume a 6ms frame duration forcompatibility with currently defined wireless TDMA air interfaces.As an improvement to this scheme, an extension of the IEEE 802.6 pre-arbitrated access protocolhas been proposed [23], which enables effective multiplexing of bi-state isochronous traffic sources, suchas voice traffic which is in the on-state during talk-spurts and the off-state during silent periods betweentalk-spurts. Silence detection allows PA slots unused during silent periods to be reclaimed for use byother traffic sources.The simulation model was employed to evaluate the effectiveness of the silence detection scheme.The network is configured for 50 uniformly spaced nodes as before. In addition, a nodal processing delayis assumed exponentially distributed with a mean value of 100,u,s. The standard call setup sequence forintraMAN calls (section 4.3.1) has been used in the simulation to account for the signalling delays duringcall setups. Call control signalling is transmitted using priority 2 QA access in the MAN, while voicetraffic is transmitted using PA access. It is assumed that each voice call consumes 64Kbps and the voiceactivity factor is 0.435 as before. In order to save computation time, a general voice call tracking schemeis used. This scheme is described as below. Instead of monitoring individual on-going call status, anactive call variable is used to keep track of the changing number of currently active calls. This variableis used to determine the percentage of used PA slots in each frame transmitted over the MAN. Theremaining slots of each frame are used for three priority QA access. Meanwhile, a separate call clearingChapter 2. Using MAN to Support PCS^ 34routine is used to activate the call clearing operation. In this way, the system needs not keep track ofindividual active calls. Substantial computation time can be saved. If simulation time is 10 sec, thecomputation time using the general voice call tracking scheme is about 4 hours instead of 12 hours byusing the individual voice call tracking scheme on a Sun Sparc II Station.Figures 17 and 18 show the comparison of the mean signalling cell transmission delays andmean call setup delays for PA voice transport with and without silence detection at different PA voicethroughput levels, respectively. In Figures 17 and 18, as the PA voice throughput increases due toan increasing number of voice calls being set up in the MAN, both the cell transmission delay andcorrespondingly the call setup delay increase due to the decreasing network capacity available for theQA signalling traffic. The maximum voice throughput supported by the MAN is therefore dictated bythe level at which insufficient signalling capacity is available so that the mean cell transmission andsetup delays become unbounded. The result shows that the use of silence detection more than doublesthe maximum voice throughput supported by the network, from 122Mbps without silence detection to279.7Mbps with silent detection. The difference between the 310Mbps network capacity and 279.7Mbpsmaximum throughput for PA voice transport with silence detection accounts for the cell overheads aswill as overheads for signalling and silence detection. Note that the total overhead is less than that forQA voice transport (Figure 16), so that the maximum voice throughput is about 23Mbps higher for PAvoice transport. The maximum number of voice calls accommodated by the network is around 4370calls without silence detection and 10050 calls with silence detection. The maximum signalling trafficfor call setup and clearing is around 0.97Mbps.Chapter 2. Using MAN to Support PCS^ 3536Chapter 3. PCN Architecture based on IEEE 802.6 MANThe main purpose of PCNs is to enable the users to call other users regardless of their locationsand movements. To serve this purpose, PCNs must have mobility management, call control managementand handoff management capabilities. This chapter examines the design of IEEE 802.6 MAN-based PCNarchitectures to provide these capabilities.In section 3.1, the basic requirements of management and control functions in a PCN arereviewed. The logical and physical architecture of the MAN-based PCN are described in section 3.2.Finally, the definitions of the basic network elements to support network management control are givenand four different network architectures employing different physical placements of the key networkelements, the Signalling Termination and the Bandwidth Manager, are evaluated by simulations in section3.3.3.1 Requirements for Management and Control Functions in PCN3.1.1 Mobility ManagementMobility management (MM) is used to support terminal and user mobility in a PCN betweencalls. To enable people to call people rather than places where terminals are located, it is expectedthat user identifications will be dissociated from user location in the PCN, i.e. a personal identificationnumber (PIN) with a flat address space will be assigned permanently to each user. The purpose ofmobility management is to dynamically bind the PIN to the user's current terminal and its location interms of appropriate network addresses. The following network functions are required.(i) Network Databases Databases are required to store each subscriber's profile, such as class of service, security codeetc., indexed against the PIN, and to provide a linkage between the PIN of a subscriber, the identificationof the terminal currently used by the subscriber, and the network address (i.e. location) through whichthe subscriber may be reached.Chapter 3. PCN Architecture based on IEEE 802.6 MAN^ 37Databases may be classified as home databases (HDBs) and visitor databases (VDBs), corre-sponding respectively to the home location registers (HLRs) and visitor location registers (VLRs) inGSM terminology [10]. A subscriber's HDB usually resides in the part of the network covering a region,e.g. a metropolitan area (MA), where the subscriber is normally located, and include all relevant entriesas discussed above. When the subscriber roams into a different region, a temporary VDB entry is setup for the subscriber in the network covering this region, into which all or a subset of the subscriber'sHDB content is copied.Designs and search strategies for network databases are considered in [39].(ii) Subscriber and Mobile Terminal (MT) Tracking CapabilitySignalling procedures must be implemented in the network, and over the air interface for MTs,to enable the PCN to ascertain what terminal each subscriber is using, where the terminal is located, andwhether the terminal is powered up in case the subscriber is using a MT. Changes are updated at theappropriate network databases. Tracking may require explicit subscriber actions when moving to a newfixed-terminal location, or may be automatically handled by the mobile access network for roaming MTs.3.1.2 Call ManagementCall management (CM) enables call origination, termination and disconnection in both the fixedpart and the mobile part of a PCN, regardless of the locations of the callers and callees. It requiresa collection of functions which provide call-related information access, resource assignment, routing,alerting, supervisory and other administrative support for the establishment, termination and disconnectionof calls. These functions are summarized in Table 2.3.1.3 Handoff ManagementHandoff management (HM) refers to the process of maintaining a call connection while themobile user using a MT is crossing the boundary between two radio cells. In a PCN, it is importantthat the system is capable of maintaining the connection when the MT is handed off between radio cellsconnected to different parts of the fixed network. The basic requirements for handoff processing in theChapter 3. PCN Architecture based on IEEE 802.6 MAN^38Table 2. Definition of Call Management FunctionsCall Management Function DefinitionCall ProcessingInteractions and procedures exercised by caller, callee andappropriate network elements to enable call setup and clearing.Call Control Signalling Communications of control messages to enable above interactionsSubscriber AuthenticationChecking the caller's and callee's subscriber profile to authenticateusers and ensure permission to access requested services.Callee LocationDatabase search to determine the network address currentlyassociated with the callee's PIN.Call RoutingDetermination of the best route to set up an end-to-end connectionbetween the caller and the callee's current network address.Resource Management Efficient allocation of communications capacity along the route.Interworking Enable routing of calls across heterogeneous networksSupplementary Functions Adjunct functions such as call details recording, special servicefeatures, etc.network are to realize fast and seamless handoff operation, which is transparent to the users, and tominimize the probability of handoff failures.3.2 PCN ArchitectureTo meet the service requirements of the PCN, a new PCN architecture based on distributednetwork intelligence and the IEEE 802.6 MAN as the transport media is being considered.3.2.1 Logical ArchitectureThe PCN architecture can be functionally described under three logical layers 1811241, as shownin Figure 19. The figure also indicates the main functions of each layer, and the network elementssupporting these functions. The three layers cooperate with each other to meet the requirements of thePCN.(i) Intelligent Layer This layer incorporates the network intelligence and processing capabilities used for realizinglocation independent services in the PCN. Network intelligence is embodied in databases. A distributedChapter 3. PCN Architecture based on IEEE 802.6 MAN^ 39Figure 19. Logical Architecture of PCNhierarchical database management system has been proposed to handle to expected demands from thevery large subscriber population in future PCNs [39].(ii) Transport Layer The purpose of the transport layer is to provide basic communication (i.e. transmission andswitching) services to enable the exchange of control signalling and user traffic. In a PCN based on theIEEE 802.6 MAN, it is proposed to use QA access for signalling transport, and PA access with silencedetection for voice traffic, as discussed in chapter 2.(iii) Access Layer This layer provides the means for fixed and mobile users to access the fixed network. Specifically,it incorporates the mobile access networks, possibly employing a variety of air interface standards.3.2.2 Physical ArchitectureIn a PCN supporting wireless access (Figure 20), the geographic coverage area is partitionedinto radio cells, each of which has a base station (BS). Several BSs can be coordinated by a base stationcontroller (BSC), which is a node on the access MAN. An access MAN may provide coverage over manycity blocks [23]. To extend the coverage to a metropolitan area and to provide world-wide connectivity,City streetoffice or mallTo B_ISDN, ISDNBackbone MANPSTN etc.Chapter 3. PCN Architecture based on IEEE 802.6 MAN^ 40Figure 20. Physical Architecture of PCN Sub-Systemone or several MAN nodes may provide interfaces to other access MANs, backbone MANs whichinterconnect several access MANs, or interworking units (IWUs) which interconnect to other networkssuch as the PSTN, N-ISDN or B-ISDN (Figure 1).3.3 Network Elements to Support PCS over MAN3.3.1 Functional Descriptions(i) Base Station Controller (BSC) It manages a set of BSs, each controlling a radio cell. It provides the interface for user trafficand control signalling between the wireless access network and the access MAN. It is assumed that eachBSC is physically associated with an access MAN node.(ii) Bridge (BG) Chapter 3. PCN Architecture based on IEEE 802.6 MAN^ 41It is used for the interconnection between two access MANs, or between an access MAN anda backbone MAN.(iii) Interworking Unit (IWU) It is used for interconnection of dissimilar networks, such as between a MAN and a LAN,N-ISDN or B-ISDN.(iv) Signalling Termination (ST) The function of the signalling termination is to manage call processing procedures by providingsignalling interactions with the BSCs, BG and IWU for call setup, clearing and handoff, and with thebandwidth manager and databases for specific call processing functions.(v) Bandwidth Manager (BWM) The BWM controls bandwidth assignment within the MAN. If the requested bandwidth for theisochronous or connection-oriented data traffic is available, the BWM assigns a virtual circuit identifier(VCI) and informs the HOBs to generate appropriate PA or QA slots. Otherwise, it postpone or rejectsthe request. Upon call completion, the BWM releases the bandwidth for assignment to other calls.(vi) Head of Bus (HOB) The function of the HOB in a DQDB MAN is to generate empty PA/QA slots on the respectivenetwork bus. It also cooperates with the BWM to write specified VCIs and control information intothese slots at regular intervals.(vii) Home Database (HDB) The function of the HDB is to store the information related to those subscribers normally locatedin the area covered by the associated network segment, such as access MAN, backbone MAN or MA. Thedatabase contains the subscribers' service profile, current terminal identifications and location informationfor the subscribers, indexed on the subscribers' PINs. The HDB entries are semi-permanent but contentsof each entry may change with the respective subscriber's current location.(viii) Visitor Database (VDB) Chapter 3. PCN Architecture based on IEEE 802.6 MAN^ 42Table 3. Relations Between Management and Control Functions and Network Elements of PCNManagement and Control Functions Network Elements InvolvedMMNetwork Databases HDB, VDBSubscriber and MT Tracking BSC, BG, IWU, HDB, VDB, STCMCall Processing ST, BWM, IWU, HOB, HDB, VDB, BSCCall Control Signalling ST, IWUSubscriber Authentication BSC, ST, HDB, VDBCallee Location ST, HDB, VDB, IWUCall Routing ST, HDB, VDB, IWU, BG, BSCResource Management ST, BWM, HOBInterworking IWU, BGSupplementary Functions ST, HDB, VDBHM Handoff Control Functions BSC, ST, BWM, HOB, HDB, VDB, IWUThe function of the VDB is to temporarily store the data related to those mobile subscriberscurrently roaming in the area covered by the associated network segment. The contents of this databaseis similar to that of the home database, but entries are deleted as soon as the corresponding subscribermove out of the area.The above network elements collectively provide the means to realize the network managementand control functions in the PCN described in section 3.1. Each element may be related to severalfunctions, and each function may require the cooperation of several networks elements as illustratedin Table 3. Each functional element may be implement with several alternative physical architectures,involving different means of mapping the functions to the MAN nodes at different locations.3.3.2 Physical Placement of ST and BWMWhile the physical placements of such network elements as the BSCs, BG, IWU and HOB arefixed by either physical constraints or locations of interconnected entities external to the MAN, the othernetwork elements including the ST, BWM, HDB and VDB may be implement at any MAN nodes. Thearchitectures and physical positions of these elements will directly influence the network performancefunctions such as signalling load, processing and transmission delay. The architecture and physicalChapter 3. PCN Architecture based on IEEE 802.6 MAN^ 43positions of HDB and VDB have been analyzed, and a distributed hierarchical database architecture hasbeen proposed [23][39]. In this architecture, each MAN node associated with a BSC has a databasepartition. Taken together they act as a logically centralized database on the MAN, but provide thecapability for parallel processing to improve response time in database searching and location updating.In the following, we examine the alternative architectures and physical positions for the ST and BWM,and investigate the performance associated with each alternative.Assume that every IEEE 802.6 MAN (access or backbone MAN) requires the functions of theST and BWM. The architectures defined below for mapping each of these network elements to physicalMAN nodes will be used in the evaluation of alternative ST and BWM positions in a MAN.• SST (single ST) — the ST is implemented in a single node.• MST (multiple STs) — the ST is distributed among several or all nodes.• SBWM (single BWM) — the BWM is implemented in a single node.• BBWM (bi-BWMs) — the BWM is distributed among two nodes.• OB (open buses) — separate HOB nodes for each bus.• CB (close buses) — HOB nodes for both buses are collocated.We consider and qualitatively evaluate four alternative signalling architectures based on differentcombinations of the above. The evaluation is based on the following assumptions.The BSCs, incorporating the database partitions, are uniformly spaced over the MAN.Each call setup requires several interactions between the originating/terminating BSC and the ST,and one interaction each between the ST and the database, the ST and the BWM, and the BWMand the HOBs.3.3.2.1. Signalling Architecture 1 (SST, SBWM, CB)With this architecture, it is best to collocate the ST and BWM at the center node of the dualbuses to minimize the average distance and hence signalling delay between the BSCs and the single ST.The collocation of the ST and BWM further reduces the need to exchange signalling between the ST andChapter 3. PCN Architecture based on IEEE 802.6 MAN^ 44BWM over the MAN. The ST/BWM functions like a centralized call processor, which has advantagesin hardware cost, and ease of operation, maintenance and upgrade. On the other hand, the drawbacksof this architecture are firstly, that under heavy call volume the ST/BWM could potentially become abottleneck for the whole network with long processing and queuing delays, and secondly, its vulnerabilityto breakdown of the ST/BWM. It also does not fully make use of the main feature of the network (i.e.distributed control capability) and is inconsistent with the distributed hierarchical database managementscheme proposed before.3.3.2.2. Signalling Architecture 2 (SST, SBWM, CB)If HOB nodes are collocated in the MAN so that the buses looks like a physical ring, it is bestto collocate the single ST and BWM with HOBs in the network. As in the previous case, the averagedistance between the BSC and the single ST is minimized, and signalling between the ST and BWM islocalized within the common node. In addition, the architecture also localizes signalling traffic betweenthe ST/BWM and the HOBs within the common node. Consequently, both the signalling traffic load andsignalling delays are reduced over the MAN. The drawbacks of this scheme are the same as those in theprevious scheme. In addition, the common node has the added complexity of the HOB functions.3.3.2.3. Signalling Architecture 3 (MST, BBWM, OB)In this architecture, the ST function is distributed in every nodes (BSCs, BG, IWU) over theMAN, in the same manner as the distribution of database partitions. BWMs are collocated with the HOBnodes at either ends of the dual buses. Call requests are locally processed in the ST collocated with theoriginating BSC. For bandwidth allocation, instead of having the ST signalling the BWM which in turnsignals the HOBs at either ends, the BSC/ST directly signals the BWM/HOB at either ends of the buses.Overall, this architecture further reduces signalling traffic load and signalling delays compared with theprevious architectures. Due to the distribution of call processing among the MAN nodes, processingtime is reduced and the processing bottleneck eliminated. If it is acceptable that a call could be assigneddifferent VCI's on each of the two buses, then the two BWMs could function independently with no needChapter 3. PCN Architecture based on IEEE 802.6 MAN^ 45for co-ordination between them. The drawbacks of this architecture are possible increase in hardwarecosts for each node, and complications in system maintenance and upgrades.3.3.2.4. Signalling Architecture 4 (MST, SBWM, CB)If the HOB nodes are collocated, the two BWMs in the previous architecture reduce to a singleBWM collocated with the HOBs in a single node. In addition to having all the advantages of signallingarchitecture 3, signalling traffic between the ST's and the BWM/HOBs is halved and delay substantiallyreduced, as the BSC/ST needs to signal the BWM/HOBs only on the bus where the propagation distanceis shortest. Furthermore, the same VCI can be assigned to each call for transmission over each bus. Thesame disadvantages as in alternative 3 are applicable. It appears that signalling architecture 4 gives thebest performance. This expectation remains to be verified by computer simulations.3.3.3 Performance AnalysisIn this section, we employ computer simulations to evaluate the impact of the signallingarchitectures considered above on the mean call setup delay and PA throughput over the network. Thesimulation model employs the PA transport mechanism with silence detection for voice transmission,whereas signalling traffic is transmitted using the highest priority QA access. It is assumed that thereare no other traffic sources in the network. As before, the network configuration consists of 50 stationsuniformly distributed in the network, the distance between adjacent nodes being 3 slot transmission timewith a mean nodal processing delay of 100ps for each signalling message. The same standardized callsetup sequence for intraMAN calls (section 4.3.1) is used for all four schemes in the simulation. Thelength of the STP message is assumed to be 5 slot time. The length of other signalling messages areassumed to be 1 slot time. The call setup time does not include the database search time for the callee.90% confidence intervals are presented for selected simulation results.Figure 21 shows the relationship between mean call setup delay and PA throughput for the fourdifferent signalling architectures in an IEEE 802.6 MAN without erasure node. In all cases, when the PAthroughput is below approximately 270Mbps, the mean call setup delays are relatively stable and increaseslowly with the PA throughput. Above this throughput level, call setup delays increase rapidly and becomeChapter 3. PCN Architecture based on IEEE 802.6 MAN^ 46Figure 21. Mean Call Setup Delay vs. PA Throughput for Different Signalling Architectureswithout erasure nodes, 90% confidence intervals indicated)unbounded as the PA throughput reaches its capacity. At the region of stable call setup delays, signallingarchitecture 4 gives the lowest delays, followed by architectures 3, 2 and 1 with increases in delays ofapproximately 0.4ms, 0.7ms and 0.5ms, respectively, between each adjacent pairs of curves. The PAcapacities are 275.7Mbps, 275.6Mbps, 275.5Mbps and 275.1Mbps for signalling architectures 4, 3, 2 and1, respectively. Thus, Figure 21 confirms our previous conjecture that among the signalling architectures,architecture 4 gives the best performance, followed by architectures 3, 2 and 1. By reducing the signallingtraffic and distributing the processing of signalling messages among the MAN nodes, architecture 4 givessubstantial reductions in call setup delays (greater than 50% reductions compared to architecture 1).To further improve the network performance and increase total network throughput, the use oferasure nodes in the MAN is considered. It is necessary to quantify the number of erasure nodes whichyields a good trade-off between performance improvements and increases in complexity and cost, and todetermine the optimal locations for these erasure nodes.We start with one erasure node on each of the DQDB buses and empirically determine its bestlocation by simulations. Figure 22 shows the relationship between mean call setup delay and erasureChapter 3. PCN Architecture based on IEEE 802.6 MAN^ 47node position index at two particular PA throughput levels on bus A. The position of the erasure nodeon bus B is symmetrical to that on bus A. The same 50-node network is employed as before. When thePA throughput is 273.6Mbps (which is close to the PA capacity), there exists an optimum erasure nodeindex, which minimizes the mean call setup delay m all the alternative signalling architectures. The besterasure node location on bus A is around node 14. The corresponding location on bus B is node 38. Thesame observations apply when the PA throughput is 204Mbps although the reduction in call setup delayis less pronounced at the optimum erasure node location. The reason for different erasure node locationsyielding different mean call setup delay is as follows. When the erasure node is close to the HOB, onlya small fraction of slots can be reclaimed because the fraction of slots that are occupied is small, andthe fraction of occupied slots that have reached destination is smaller still. On the other hand, whenthe erasure node is close to the end of the bus, although a large number of slots can be reclaimed, thenumber of downstream nodes is small, thus reducing the chance of re-using these slots. Note, however,that the best location of the erasure node depends on the traffic pattern on the network, and may vary ifthe traffic pattern changes. Here, the optimization of erasure node location is with respect to the trafficpattern for call control signalling.Figure 23 and Figure 24 show the effects of different number of erasure nodes on the meancall setup delay for signalling schemes 3 and 4, respectively. Three different cases are considered ineach figure: without erasure node (case 1), with one optimally located erasure node on each bus (case2) and with erasure capabilities at every node except the nodes at either ends of each bus (case 3). Asexpected, case 3 yields the best performance, followed by case 2 and case 1. Notice, however, that theimprovements between case 2 and case 1 are larger than the improvements between case 3 and case 2.That is, the use of a single erasure node on each bus brings about substantial performance improvements,whereas increasing the number of erasure nodes bring about progressively smaller improvements. Sincecomplexity and hardware costs increase with the number of erasure nodes in the network, it is reasonableto conclude that case 2 represents a good compromise between network performance and hardware costs.On the other hand, case 3 gives the performance upper bound. In subsequent simulations, we will adoptcase 2 as the default configuration.Chapter 3. PCN Architecture based on IEEE 802.6 MAN^ 48Figure 22.Figure 23. Effect of Erasure Nodes on Call Setup Delay (Signalling Scheme 3)The call setup delay performances of the four alternative signalling architectures are comparedin Figure 25 for the DQDB MAN with one erasure node per bus. The same relative performance asin Figure 21 is maintained between the four schemes. Furthermore, the use of erasure nodes results inCall Setup Delay vs. Location of Erasure NodesChapter 3. PCN Architecture based on IEEE 802.6 MAN^ 49Figure 24. Effect of Erasure Nodes on Call Setup Delay (Signalling Scheme 4)general improvements of call setup delays in each of the four signalling schemes. This is made explicitin Figure 26, which overlays Figure 21 and Figure 25. In the region of stable call setup delays, the use oferasure nodes result in a 10% reduction in mean call setup delay for signalling architecture alternative 4.Chapter 3. PCN Architecture based on IEEE 802.6 MAN^ 50Figure 25. Mean Call Setup Delay vs. PA Throughput for Different Signalling Architectures(with one erasure node per bus, 90% confidence intervals indicated)SIGNALLING ARCHITECTURES1. SST, SBWM, OB, EN=O^^2. SST, SBWM, CB, EN=0 03. MST, BBWM, OB, EN=O^e4. MST, SBWM, CB, EN=0 --1. SST, SBWM, OB, EN=14A,38B2. SST, SBWM, CB, EN=14A,38B3. MST, BBWM, OB, EN=14A,38B4. MST, SBWM, CB, EN=14A,38B5.54.50:37CDCD3.5=ct50,,E 2.5asa)21.51204 214^224^234^244^254^264^274PA Throughput (Mbps)Figure 26. Comparison of Call Setup Delays, with and without Erasure Nodes51Chapter 4. Signalling Procedures for Call Setup,Clearing and Handoff in PCNThe provision and maintenance of connection-oriented services across the PCN relies not onlyupon the availability of an appropriate network topology providing the required connectivity betweentwo end users, but also upon the correct functioning of a set of logical 'call control' procedures. Theseprocedures define the logical sequence of events for establishing, clearing and handing off calls. Theevents occur at distributed network elements which interact by means of signalling protocols. CCITThas standardized a series of signalling protocols for public switched telephone networks, such as SSNO.7 which has been adopted for ISDN inter-exchange signalling, and Q.920-921 data link layer andQ.930—Q.931 network layer for signalling over the ISDN user-network interface (UNI). These signallingprotocols and call control procedures are designed for wireline networks with centralized switching toexchange control signals between fixed points. However, future PCNs will allow wireless access frommobile users in the system and users will be identified by personal identification numbers (PIN) whichare independent of users' locations and terminal equipments in use. PCNs may adopt new networkarchitectures, such as the distributed MAN-based architecture being considered in this thesis. To takeadvantage of the broadcasting and distributed processing capabilities of this network architecture, itis necessary to add new features to the signalling protocols and design new call setup and clearingprocedures. Call handoff procedures also need to be investigated to support mobility of users duringcalls. This chapter considers the call setup, clearing and handoff procedures and signalling protocolsapplicable to a PCN architecture based on the IEEE 802.6 MAN. The call setup and clearing proceduresare derived from CCITT Q.931 and SS NO.7 signalling protocols for ISDNs.In section 4.1, a brief review of call setup/clearing procedures in ISDN and call handoff functionsin mobile communication systems are given. The functional definition of call setup/clearing proceduresare given in section 4.2. In section 4.3, call setup/clearing procedures in the PCN are proposed anddiscussed. Finally, call handoff procedures in the PCN are proposed and discussed in section 4.4.Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 524.1 Review of Signalling Procedures and Functions in Other Systems4.1.1 Review of Call Setup/Clearing Procedures in ISDNSince most of the signalling procedures in current mobile communication systems are derivedfrom the signalling protocols of 1.451 and the SS NO.7 ISDN User Part (ISUP), the call setup/clearingprocedures defined by the 1.451 network layer and the SS NO.7 ISUP protocols used in ISDN networks[40][41] are reviewed in this section. Only the procedures for the simple case involving a successful callsetup and a normal call clearing are considered.Figure 27 demonstrates the call control procedures across an ISDN network encompassing theoriginating exchange, two transit exchanges and the destination exchange. The messages used over theUNI are defined in the 1.451 signalling protocol. The inter-exchange messages are defined in the SSNO.7 signalling protocol. The local (i.e. originating and destination) exchanges provide interworkingcapabilities between 1.451 and SS NO. 7, and must handle both types of messages.The caller initiates call establishment by transferring across the UNI to the originating exchangeon the network side one or more messages containing the information elements (1E) required by thenetwork to process the call. This information may be transmitted by one of two methods, known asen-bloc and overlap signalling.In en-bloc signalling, all information is contained in a single Setup (STP) message. Aftertransferring this message, the user waits for further response from the network. The originating localexchange, on receiving the STP message, determines if the information received is complete andcompatible with its capabilities, and if so, returns the Call Proceeding (CPD) message to acknowledge theSTP message sent by the calling user. This message also includes the identity of the component channelallocated to the call, i.e. the channel in the immediate network segment between the caller and the callee.In overlap signalling, only partial call establishment information is transferred by the caller inthe initial STP message. The originating exchange returns the Setup Acknowledge (STP_ACK) messagecontaining the allocated component channel to the caller, with an indication that more data regarding thecall is required. On receiving the STP_ACK message, the caller then transmits additional InformationChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 53Figure 27. Call Setup/Clearing Procedures in ISDN(INF) messages to the network. Each INF message will be acknowledged by the network individuallyby sending a STP_ACK message back to the caller. This procedure may be invoked several times ifnecessary to convey all the required call setup details. Once the originating exchange determines that ithas obtained sufficient information to process the call, or it receives an indication from the caller thatsending is complete, it returns the CPD message to the caller.Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 54Several exception conditions may arise during the operation of this protocol. First, a specificcomponent channel or facility requested by the caller may not be available for the call. Second, the callestablishment information provided in en-bloc or overlap signalling may be insufficient or wrong. Third,the network timer that governs the maximum time allowed for a response from the caller may expire. Inall these cases, the originating exchange sends a Release Complete (RCP) message containing a causeIE that specifies the nature of the problem to clear the call.Following the transmission of the CPD message to the caller, the originating local exchangeinitiates appropriate inter-exchange signalling to transfer the call establishment information towards thedestination exchange using the SS NO.7 signalling network and protocols. The routing is determinedby analyzing the contents within the STP/INF messages and referring to a network database containingrouting information. This enables the originating exchange to select an appropriate initial transit exchangeand an interconnecting circuit specified by its circuit identification code. The originating local exchangethen generates an Initial Address Message (IAM), and one or more Substantial Address Messages (SAM's)if necessary (i.e. if overlap signalling is used over the UNI), and transmits these messages to theselected initial transit exchange over the SS NO.7 signalling network. The IAM and SAM's correspondrespectively to the STP and INF messages in 1.451, and contain the information supplied by the callerin the STP/INF messages, including the station numbers for the calling and called parties, as wellas the transmission medium requirement and all other information necessary to route the call to thedestination local exchange. The initial transit exchange analyzes the called number, the transmissionmedium requirement, and other parameters in the IAM/SAM's to determine the next transit exchangeand an appropriate interconnecting circuit, and then forwards the IAM/SAM's with a new routing labeland circuit identification code to the next transit exchange. The user transmission path is connectedthrough the transit exchange.This procedure is repeated at all subsequent transit exchanges until the IAM/SAMs arrive atthe destination local exchange. When all call control details have been received at the destination localexchange, they are analyzed to determine whether a connection can be established to the called party.Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 55Assuming that no incompatibilities exist, a STP message is then transferred over the remote UNI to thecallee. Again, the procedures of overlap signalling, as outlined earlier, may be employed to convey anyadditional call establishment details not contained in the STP message using INF messages.Assuming now that the callee is willing and able to accept the call which characteristics aredefined in the received STP and INF messages, including the specified component channel, and thatthe specification of the call is complete, it transmits the CPD message to the destination local exchangeindicating that the requested call establishment has been initiated and to verify the selected componentchannel. Upon receiving this message, the destination local exchange transmits an Address Completemessage (ACM, corresponding with CPD message in 1.451) which is forwarded by the successive transitexchanges to the originating local exchange to indicate that complete addressing information is nowavailable at the destination local exchange. Once the callee's terminal equipment is ready, it transmits anAlerting (ALT) message to the destination local exchange, which causes the relaying of a Call Progressing(CPG) message via the transit exchanges to the originating local exchange. The latter then sends an ALTmessage to the caller. Next, the callee responds to the call by sending out the Connect (CNT) message tothe destination local exchange, resulting in the return of a Connect Acknowledge (CNT_ACK) messageto the called terminal, and the relaying of an Answer (ANM) message through the network to theoriginating local exchange, whereupon a CNT message is conveyed to the caller as the final step in thecall setup procedure. Optionally, the caller sends a CNT_ACK message to the originating local exchangeto indicate the completion of the call setup procedure.A number of anomalies must be considered. The callee could be busy, could be incompatiblewith the characteristics of the requested call, or could refuse the call for other reasons. In all suchcases, the callee transmits a RCP message to the destination local exchange including appropriate causeIEs indicating the reasons for the refusal of the call. In response, the network initiates appropriate callclearing procedures.Figure 27 also shows the call clearing procedure when the caller initiates call clearing. Anessentially symmetrical procedure applies to call clearing by the callee. Release of the connection isChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 56initiated by sending the Disconnect (DISC) message across the UNI to the local exchange, which causesthe release of the circuit to the transit exchange, the transfer of a Release (REL) message to the clearinguser, and the relaying of the REL message toward the remote local exchange. The clearing user normallyacknowledges the REL message by a RCP message. At each transit exchange the reception of RELmessage causes the release of the through-connection, the return of a RCP message to the precedingexchange, and the relaying of the REL message to the next exchange. At the remote local exchangethe reception of the REL message also causes the transfer of the DISC message to the remote userterminal, which responds with the REL message. The transfer of the RCP message to the cleared userthen completes the procedure.4.1.2 Call Handoff Functions in Cellular NetworksOne of the major requirements of cellular mobile communication systems is to maintain anacceptable quality of communications to and from the mobile subscribers during calls. In these systems,the MTs frequently move across cell boundaries during calls. As a MT moves away from a BS, theradio link between the MT and the BS currently serving it will gradually degrade. When the MT reachesa location where a better quality radio link could be established with a different BS, i.e. the MT hascrossed a cell boundary, it becomes necessary for the system to automatically and quickly switch the MTto a new radio channel serviced by the new BS in the new cell so that the quality of communications canbe maintained and the on-going conversation can be continued without dropping the call. This processis called a call handoff.In the first generation cellular mobile communication systems [2][3], the handoff functionsare centralized at the mobile switching center (MSC). The system controller in the MSC assemblesinformation from several BSs about the strengths of the signal paths from each active MT. On the basisof these measurements, the controller decides when the MT should switch over to a new channel at adifferent BS. With the number of subscribers and the offered traffic increasing rapidly, cell sizes arebeing decreased and the number of cells increased to handle this demand. Consequently, the increasingvolume of handoffs may exceed the processing capacity of the MSC.Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 57With the development of second generation cellular systems [6], the trend is to distribute networkcontrol functions more evenly between the MSCs, BSs and MTs. An important feature of both the NorthAmerican and Pan-European digital cellular systems is mobile assisted handoff. In this approach, theMT measures during a call the signal strengths from BSs in neighboring cells. These measurements arereported to the serving BS, which determines the need for a handoff and informs the system controller atthe MSC the preferred new cell. In this way, a large portion of the information processing and controloperations associated with handoffs are moved from the system controller in the MSC to the BSs andMTs, thus avoiding overload at the controller and enabling faster handoffs.Since the trend towards smaller cells will continue in future third generation cellular systems,management of handoffs will become more important. Call handoffs will occur more frequently and willhave to be carried out more rapidly. Due to the large number of base stations required in micro-cellulararchitectures, the MSC control function may not be suitable to handle the resultant increase in processingload. In this thesis, we consider the use of a MAN-based distributed network architecture to replace thecentralized MSC in the third generation cellular system which will form an integral part of the PCN. Asfuture PCNs may be required to service isochronous traffic that is loss sensitive, the call handoff procedurebased on this network architecture is required to be 'seamless' in the sense that user information flow isnot subjected to cut off during handoffs, and 'fast' to minimize the probability of handoff failures.In keeping with the trend of relieving the processing loads of network elements, a MT-directedcall handoff scheme is proposed [23]. The MT, by measuring the signal qualities of neighboring BSs,initiates a handoff request to a new BS following an indicated improvement in link quality. During therequired time to setup a radio channel (or perhaps an isochronous channel across a MAN) to the new BS,the MT maintains the existing circuit to the old BS, to guarantee a seamless handoff and maintain thecommunications quality. This handoff request also includes the MT's present BS address and attachedMAN node address, previously stored in the MT's random access memory. Using this information, thenew BS can acquire addition information related to the call from the old BS, and determine the type ofhandoff by comparing its hierarchical address with that of the old BS. The new BS sets up isochronousChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 58channels using high priority QA signalling in the PCN if needed, and sets up a new radio channel tothe MT. Finally, when the new channel has been set up over the air interface and the MAN, the MTdisconnects the old channels and switches to the new channel.4.2 Functional Elements of Call Setup, Clearing and Handoff in PCNDue to the unique characteristics of PCNs, specifically the one based on IEEE 802.6 MANas discussed in previous chapters, the call setup/clearing procedures require additional functionalitiessupported by new signalling messages.First, the PCN is required to support mobile users, enabling them to originate and terminate callswhile roaming in different coverage areas in the network. In contrast with fixed users, which networkidentity is synonymous with the address of a fixed location, mobile users' current location and networkaddress are not directly related to the users' personal identification number (PIN). The PCN must havethe capability to track roaming users and bind their PIN's to their current locations and network address.This information is kept in network databases, the organization of which is outside of the scope of thisthesis [39]. However, PCN call management requires the capability to access these databases during callsetup to authenticate the caller, to locate the callee, and to check the profiles of the caller and the calleeagainst the specifications of the call.Second, the use of MAN as a distributed switching and interconnection network requires differentcontrol procedures than centralized networks employing local and transit exchanges. In fact, call controlsignalling procedures and protocols have not been well defined for MAN-based PCN's. While theDQDB protocol standardized for IEEE 802.6 MAN is suitable for supporting call control signalling at thephysical layer and data link layer, no applicable standards exist for network layer functions, procedures,and protocols. These network layer capabilities facilitate the use of network resources, specific to theMAN-based PCN architecture detailed in previous chapters to support connection-oriented PCS, usingthe services provided by the DQDB protocol. The call setup/clearing procedures should be functionallycompatible with currently standardized ISDN call setup/clearing procedures to eliminate future problemsin interworking. However, the specific network configuration under consideration requires new proceduresChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 59for setting up the facilities (such as isochronous channels) and routing calls. These will be described inmore detail. In addition to the signalling messages used in ISDN, as defined by 1.451 and SS NO. 7,extra signalling messages need to be introduced to support these new procedures.In the following, we will discuss the functions of the call setup, clearing and handoff procedures,and then we will define the messages used in these procedures.4.2.1 Call Setup FunctionsCall setup procedures (whether for fixed calls or mobile calls) can be functionally divided intothree interacting processes (i.e. call information process, locate callee process and routing/facility setupprocess), as shown in Figure 28, with distinct characteristics. During the call setup operation, theseprocesses exchange information among each other to realize the call setup procedures.(i) Call Information ProcessOn receiving call related messages, the call information process analyzes the informationreceived, makes decisions regarding the type of call and the procedure for proceeding, forms new messagesand forwards them to the process responsible for the next step of the call setup.(ii) Locating Callee Process This process is actually a sequence of database searches. It uses the information provided byprevious processes and selects an appropriate algorithm to access the network databases and search forthe location of the callee. The search results, which may include the current network address of thecallee and suitable routing information for this call, are forwarded to the appropriate network entities forthe subsequent call setup process.(iii) Routing and Facility Setup ProcessThe purpose of this process is to route the call to the called user and set up the facilities alongthe routing path from source to destination. Particularly, the process sets up isochronous channels inevery MANs along the routing path.Call setup procedures can also he physically divided into three independent stages (i.e. callinguser-to-network; network-to-network; network-to-called user) as shown in Figure 29. The generalSEARCHING RESULT INQUIRING CALLEESETUP SETUPCALL PROCEEDING^CallCalling • ALERTING^InformationUser^CONNECT ProcessALERTINGCONNECTCalledUser7-Routing& Facility SetupProcessCONNECT ACKNOWLEDGE IREQUEST^RESPONSESTAGE1 STAGE2 STAGE3•^SETUPCALL PkOC•ALERTINGCONNECT •000ALERTINGCONNECT •SETUPALERTING•CONNECT-41SETUPCalledUserAccessMANAccessMANNetworkCallingUserChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 60Figure 28. Process Interactions in the Call Setup ProcedureFigure 29. Physical Partitioning of Call Setup Procedureprocedures for each stage can be realized using several alternative time sequences for the requiredmessage flows. The overall procedures are achieved using the general message flows between the threestages indicated in Figure 29.The main functions of stage 1 are to realize the handshake between the caller and the localChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 61access network (at the calling user side), inquire the network database for callee information, and set upthe facilities in the local access network.The main functions of stage 2 are conveying messages within the network between the callinguser side and the called user side for the purpose of routing and setting up facilities. Several networksmay be involved in this stage if the call is over long distance. However, if the call is local, this stagemay be bypassed.The main functions of stage 3 are to realize the handshake between the local access network (atthe called user side) and the callee, and set up the facilities in the local access network.4.2.2 Call Clearing FunctionsThe call clearing procedure can be divided into two interacting processes: the call releaseinformation process and facility release process, as shown in Figure 30. When call clearing is requiredduring call setup or while the call is in progress, these processes exchange information between eachother to realize the call clearing function.(1) Call Release Information ProcessOn receiving the call clearing request message, the call information process decides which relatedfacilities need to be released and informs the appropriate network entities to release the facilities.(ii) Facility Release Process The purpose of this process is to physically release the facilities as required, and upon completionreturn a response message to the call information process. The release request can be sent from the callinguser side, the called user side or the transit network side.4.2.3 Call Handoff FunctionsThe call handoff procedure can be divided into three interacting processes as shown in Figure31: the handoff information process, facility setup process and facility release process.(i) Handoff Information ProcessChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 62Figure 30. Process Interactions in the Call Clearing ProcedureOn receiving a call handoff request, the handoff information process retrieves related informationfrom the previous BS, decides the type of handoff, and informs the facility setup and release processesto setup and release call related facilities to the next and the previous BSs, respectively.(ii) Facility Setup Process The purpose of this process is to set up the new facilities associated with the next BSs andreporting to the handoff information process when done. To reduce the probability of call handofffailures, facility setup for handoffs may be given a higher priority than that for call setups.(iii) Facility Release Process This process is used to release facilities associated with the previous BS. The handoff informationprocess is informed upon completion of this process. For seamless handoffs, facility setup for the nextBS should precede facility release for the previous BS.4.2.4 Definitions of Call Control Signalling MessagesTables 4-7 define the major signalling messages used for call control in the MAN-based PCN.The abbreviations of the message names are used in subsequent descriptions of call setup, clearing andhandoff procedures. Most of the messages have adopted the names used in the 1.451 signalling protocolbut the contents of the messages may be different. Some new messages have been introduced to realizethe new functions related to callee location and setting up/releasing isochronous channels over the MAN.Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 63Figure 31. Process Interactions in the Call Handoff ProcedureMessages have local significance if they are applicable over a particular interface (e.g. the UNI) only;otherwise they have global significance.Table 4. Definitions of Call Setup MessagesMessage Abbreviation Significance DefinitionALERTING ALT global Indicate that the callee is generally ableto accept the call, and is being alertedCALLPROCEEDING CPD localIndicate that information provided bycaller is complete and call establishmenthas been initiatedCONNECT CNT global Indicate call acceptance by called partyCONNECTACKNOWLEDGE CNT ACK localAcknowledge the Connect messageSETUP STP global Initiate call establishmentPROGRESS PRG global Report progress of call setupINFORMATION INF global Additional information provided by callerSETUPACKNOWLEDGE STP ACKlocal Indicate that call establishment has beeninitiated but more information is requiredChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 64Table 4. (Continued) Definitions of Call Setup MessagesMessage Abbreviation Significance DefinitionINQUIRINGCALLEE INQ_CLE global Search the location of called partySEARCH RESULT S_RST global Search result consisting of routing andlocation information about called partyTable 5. Definitions of Call Clearing MessagesMessage Abbreviation Significance DefinitionDISCONNECT DISC local Sent by user/network to request connectionclearingRELEASE REL global Request to release the facilities assigned tothe callRELEASECOMPLETE RCP local Indicate completion of call clearingTable 6. Definitions of Call Handoff MessagesMessage Abbreviation Significance DefinitionHANDOFF REQUEST HO_REQ local Request a handoff operationHANDOFFPROCEEDING HO PD localIndicate that call handoff has beeninitiated and information provided iscompleteCONFIRM HANDOFF CFM_HO local Indicate that handoff has beenacceptedREQUEST CALLINFORMATION REQ_INF global Request call related informationSUPPLY CALLINFORMATION SPY_INF global Supply call related informationEXCHANGE CALLINFORMATION EXG _ INF localExchange call related informationbetween two network entitiesChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^65Table 6. (Continued) Definitions of Call Handoff MessagesMessage Abbreviation Significance DefinitionRESERVE RADIOCHANNEL RES —RAD localNew BS informs MT assignment of anew radio channelDISCONNECT RADIOCHANNEL DISC RAD localPrevious BS disconnects the old radiochannel to the MTTable 7. Definitions of Facility Setup/Releasing MessagesMessage Abbreviation Significance DefinitionREQUEST ISOCHRONOUSCHANNEL REQ_ISOC local Request an isochronous channelRESERVE VCI RES_VCI local BWM informs HOB to reserveVCICONFIRM VCI CFM_VCI local HOB confirms that VCI can beusedNOTIFY VCI NTY_VCI local Inform the node about VCIassignmentCONFIRM ISOCHRONOUSCHANNEL CFM ISOC localBWM informs ST that isochronouschannel has been reservedRELEASE ISOCHRONOUSCHANNEL REL ISOC localRequest to release an isochronouschannelRELEASE COMPLETEISOCHRONOUSCHANNELRCP_ISOC local ndicate the release ofI ^ anisochronous channelRELEASE VCI REL VCI local Request to release a VCIRELEASE COMPLETE VCI RCP VCI local Indicate the release of a VCI4.3 Call Setup and Clearing ProceduresThe following discussions of call setup and clearing procedures are based on the PCN architectureemploying interconnected IEEE 802.6 MANs as described in Section 3.2. The call control signallingprocedures are described in terms of the exchange of signalling messages between functional networkelements, with no regard for the physical locations of the network elements. In addition, the effectsChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 66of alternate signalling architectures, with different placements of the ST and BWM, on the proceduresfor setting up isochronous channels over each MAN are discussed. To simplify the discussions, a fairlysimple example is used in the description of the call setup procedures, which involves the successfulsetup of an end-to-end connection from mobile caller A, using MT_A, to mobile callee using MT_B,over four network segments including the access MANs at either ends and two transit MANs in between,interconnected by transit bridges (BGs). It is assumed that the transit bridges have complete routinginformation stored in tables which enable them to route signalling messages and setup virtual circuitsbetween any specified pairs of source and destination nodes specified by hierarchical network addresses.4.3.1 Stage 1: Call Setup and Clearing betweenCalling User and Access NetworkTwo alternative sets of call setup procedures as well as call clearing procedures are presented.4.3.1.1. Sequential Call Setup ProcedureThis set of procedures represents the straight forward performance of call setup functions in astep-by-step manner, as shown in Figure 32, and elaborated below.1. A STP message from MT_A, which contains the requested service class and bandwidth capacity, thecaller's personal identification number (PIN), The MT identification number used by the caller, thecallee's PIN and other information related to the call, arrives at the BSC at node A in the originatingaccess MAN. This message is forwarded to the ST over the access MAN.2. The ST analyzes the message received, and sends a STP_ACK message back to MT_A via node Aif further information is needed from MT_A. MT_A responds to the STP_ACK message by sendingadditional INF messages to the ST.3. When the required setup information is complete, The ST returns a CPD message to MT_A throughnode A, indicating that call setup is proceeding in the network.4. Meanwhile, the ST sends an INQ_CLE message to the corresponding network database to searchfor the callee's location and routing information. The detailed method of locating callee is outsidethe scope of our discussion.Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 675. The network database sends a S_RST message back to the ST indicating MT_B's current locationand the associated routing information.6. The ST begins the process of setting up an isochronous channel over the access MAN by sendinga REQ_ISOC message to the BWM.7. If the BWM accepts this request, it assigns a VCI and s/s VCl/offset for this call and sends thisinformation in a RES_VCI message to the HOBs.8. If the HOBs can reserve the requested VCI for the call, it transmits a CFM_VCI message back tothe BWM. The HOB does not start generating PA slots at this time. Only the s/s slots are generatedand attached at the end of each frame. PA slots are generated for this call when the incoming s/sslots indicate that speech activity has started.9. The BWM then transmits NTY_VCI message(s) to node A and the node which has been assignedas the next transit destination of the call (transit bridge BG1 in this case) indicating the VCI ands/s VCl/offset assigned for this call. Notice that if node A and the transit node are all on the sameside, i.e. both downstream or upstream of the BWM on the MAN, a single NTY_VCI message canbe used to inform both nodes.10. Node A and the transit BG1 then return a CFM_VCI message to the BWM to confirm the acceptanceof this notification.11. The BWM then sends a CFM_ISOC message to the ST indicating that an isochronous channel hasbeen reserved.12. The ST then reorganizes the STP message to include MT_B's location and routing information,and sends it to the transit network through BG1. This STP message is acknowledged by the PRGmessage immediately after the transit BG1 receives the STP message.13. The ST waits for the arrival of an ALT message from the transit network. Upon receiving thismessage, the ST forwards it to MT_A indicating that the callee is being alerted about the call.14. The transit network finally relays a CNT message to the ST indicating that the callee has accepted thecall. The ST then passes this message to MT_A via node A. MT_A returns a CNT_ACK messageto the ST to conclude the stage 1 call setup procedure. When speech activity starts, node A requestsChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 68the generation of PA slots by setting the proper bit in the s/s slot in each frame transmitted on thebuses, according the VCl/offset assignment.Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 69From the above description of the stage 1 sequential call setup procedure, we can identifyseveral phases in the procedure.• Information Phase: call setup related information is transmitted and analyzed (steps 1-3, step 12).• Locating Callee Phase: searching the network database for the callee's location and routing infor-mation (steps 4, 5).• Isochronous Channel Setup Phase: setting up the PA slot VCI and s/s slot VCl/offset for theisochronous channel over the MAN (steps 6-11 give a generic description of the procedure ofthis phase; the effects of different signalling architectures on the actual procedure are consideredin section 4.3.4).• Awaiting Callee Response Phase: the callee is alerted and finally accepts the call (steps 12-14).We define T stagel as the time from the transmission of the initial STP message by the callinguser to the time that the STP message has reached BG1 at the interface to the transit network; T,,,fas the time used in the information phase; T search as the time used in the locating callee phase andT zsoc as the time used in the isochronous channel setup phase. In this procedure, the call setup phasesoccur sequentially. Therefore,T stagel = T z.nfl+T search+ T z.socl+ T z.71f2 (1)where T Znfl corresponds to the first information phase involving the caller and the access network, andT 27if2 corresponds to the second information phase involving the access and transit networks.Note that in the case of an intra-MAN call, where nodes A and B are on the same MAN, the stage1 and stage 3 call setup procedures would be combined, and T stagel would represent the overall call setupdelay. This is the case for the simulation results involving call setup delays presented in chapters 2 and 3.4.3.1.2. Overlapping Call Setup ProcedureTo reduce call setup time, some call setup phases, specifically the isochronous channel setupphase and the second information phase, can overlap so that the corresponding operations are performedChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 70Figure 33. Overlapping Call Setup Procedure in Stage 1in parallel, as illustrated in Figure 33. In comparison with the sequential procedure described in the lastsection, the steps are the same for both procedures, but the sequence timing is different. In particular,steps 6 and 12 are performed together, so that steps 6-11 and steps 12-14 overlap in their execution.The stage 1 call setup time for the overlapping call setup is therefore reduced toT^ = T^+ T^+ T .^ (2)stag el^in f 1^search^11 f2Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 714.3.1.3. Call Clearing ProcedureIn addition to normal call clearing initiated by either users at the end of a call, the call clearingprocedure must also handle exception conditions which may arise during call setup, such as invalid callinformation, expiration of response timers in the signalling protocol, unavailability of radio channel,isochronous channel or other facilities needed to set up the connection, or unavailability of the calleeto accept the call.In all these cases, the stage 1 call clearing procedure must respond to call clearing initiated byone of following entities—the calling user, the network (i.e. local access network, transit network, orthe remote access network), or the called user. The latter case is indicated by a REL message arrivingat the ST from BG1. These cases result in three different scenarios of call clearing procedures in stage1, as shown in Figure 34.Supposing the release command comes from the network via BG1 for whatever reasons, thecall clearing steps are as follows:1. A REL message which includes the reason of the release is transmitted from transit BG1 to the STin the local access MAN.2. The ST forwards this message to node A. The ST also informs the BWM to release the isochronouschannel used for this call. Several internal messages (REL_ISOC, REL_VCI, RCP_ISOC, RCP_VCI)are employed in this process. Upon completion, a RCP message is sent back to BG1.3. When node A receives the REL message, it forwards this message to MT_A and returns a RCPmessage to the ST.4. When MT_A receives the REL message, it returns a RCP message back to node A. Both the BSCat node A and MT_A release the radio channel assigned to the call.The other two scenarios shown in Figure 34 involve similar steps as described above.The above descriptions of the call clearing procedure assume all required facilities have alreadybeen set up for the call and therefore need to be cleared, i.e. the procedure involves the maximumnumber of steps. The number of steps is reduced if call clearing occurs when some facilities have not yetChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 72Figure 34. Call Clearing Procedure in Stage 1been set up. Since the overlapping call setup procedure could potentially set up more facilities sooner, itcould require extra signalling steps for call clearing resulting from abnormal terminations. For example,if a REL message arrives from MT_A during the isochronous channel setup phase, the local accessMAN needs only to terminate this process if sequential setup is employed, whereas an additional RELmessage must be forwarded to the transit network in correspondence with the STP message sent earlier,if overlapping setup is employed.4.3.2 Stage 2: Call Setup and Clearing within Transit NetworkStage 2 defines the signalling procedure used within the transit network. As in stage 1, twoChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 73alternatives are considered for the call setup procedure, as described below. The call clearing procedureused in stage 2 is also described.4.3.2.1. Sequential Call Setup ProcedureThe procedure, as shown in Figure 35 and described below, involves sequentially setting upisochronous channels in each intermediate transit MAN.1. Upon receiving the STP message from the caller's access MAN, BG1 forwards this message to theST1 in transit MAN!. ST1 then informs the BWM in transit MANI to reserve the correspondingisochronous channel. The procedure used for reserving an isochronous channel is the same as before.When ST1 receives from BWM1 confirmation of the isochronous channel reservation, it forwardsthe STP message to BG2, which interconnects transit MAN1 to transit MAN2. BG1 maintains themapping between VCIs in the access and transit MANs assigned to the isochronous channels set upfor the call in the respective MANs.2. The above operation is repeated in transit MAN2. Upon successful reservation of an isochronouschannel in transit MAN2, ST2 forwards the STP message to BG3 which links transit MAN2 to thecallee's access MAN.3. BG3 waits for an ALT message from the callee. When this message arrives, it is forwarded to ST2,BG2, ST1, and then BG1. BG1 then forwards this message to the ST in the caller's access MAN.4. BG3 waits for a CNT message from the callee and forwards the message to ST2, BG2, ST1, BG1,and finally to the ST in the caller's access MAN. Each of these network elements responds to theCNT message by returning a CNT_ACK message to the previous network element.We define the T stage2 as the time from when the start of transmission of the STP messageby BG1 to the time when the STP message has reached BG3. The time used in stage 2 of call setupprocedure is thereforeT stage2 = T.n f3 + T t.soc2+ T t.nf4+T.soc3 + T.77.0 (3)where the elements are as defined previously and indicated in Figure 35.iChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 74Figure 35. Sequential Call Setup Procedure in Stage 24.3.2.2. Overlapping Call Setup ProcedureAs an alternative, when a STP message is received at a transit BG, it can be immediatelyforwarded to the next transit BG without waiting for the completion of the isochronous channel setupat the transit MAN. This enables the isochronous channel setup processes in adjacent transit MANS toVCPmALTALTCNTCNTCATT ciceNT AckSTPPRGsCPM vetALTALTCNTCNTCIVT ACKCAIT ACKTransitBGIHOB1 ST1 BWM1 TransitBG2HOB2 ST2 BWM2 TransitBG3Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 75Figure 36. Overlapping Call Setup Procedure in Stage 2overlap and thus reduces the call setup time. The signalling sequence for this alternative procedure isillustrated in Figure 36.The call setup time for stage 2 is reduced to:T =T^ + T^+ T istage2^inf3^inf4^nf5 (4)4.3.2.3. Call Clearing ProcedureAs before, the call clearing procedure in stage 2 can also be summarized under three differentscenarios, i.e., release initiated from the caller's side, release initiated from the callee's side, and releaseChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 76initiated from within the transit network. The procedural steps are shown in Figure 37, the descriptionof each step being similar to that described under stage 1. Again, the actual number of steps in anabnormal call clearing depends on the extent that facilities have been set up for the call. Therefore, theoverlapping call setup scheme may need more signalling steps than the sequential call setup scheme incall clearing before call setup completion.4.3.3 Stage 3: Network to Callee Call Setup and ClearingStage 3 involves signalling between the callee and the callee's access network. Only one set ofcall setup procedure exists in stage 3, as described below. The call clearing procedure is also discussed.4.3.3.1. Call Setup ProcedureThe major steps in the stage 3 call setup procedure as shown in Figure 38 are described below.1. A STP message sent by BG3 reaches the ST in the callee's access MAN.2. The ST begins to reserve the isochronous channel for this call using the procedures describedpreviously and elaborated in section 4.3.4.3. After the isochronous channel has been reserved, the ST forwards the STP message to node B,which then pages MT_B over its control area.4. When MT_B receives this message, it transmits an ALT message to the ST via node B to indicateits general ability to accept the call. The ALT message is forwarded over the network to the callerat MT_A.5. When the callee answers the call, MT_B transmits a CNT message to the ST via node B. Thismessage is forwarded over the network all the way to the caller indicating acceptance of the call.6. The ST returns a CNT_ACK message to MT_B as a response.Call setup time T sta ,3 is defined as the time from the start of STP message transmission atBG3 to the time of STP message reception by the callee, and is given by:T = T .^+ T .^+ T .^ (5)stag e3^171 f 6^Isoc4^inf 7Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 77InsitG3Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 784.3Figure 39 shows the call clearing procedures in stage 3 under three different scenarios, i.e.,release initiated from callee, release initiated from remote network, and release initiated from callee'saccess MAN. Previous descriptions of the call clearing procedure are also applicable here.Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 79Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 80Figure 40. Isochronous Channel Setup for Signalling Architecture 14.3.4 Effects of Different Signalling Architectureson Isochronous Channel SetupFigures 40 to 43 show the variations of the procedures for setting up isochronous channelsin each stage of the call setup procedure, with respect to the four signalling architecture alternativesconsidered in chapter 3. Different signalling architectures have different signalling sequences for settingup an isochronous channel, which in turn affects the overall performance of the call setup operationin the network, as illustrated in the next section. While the signalling sequences show that signallingarchitecture 2 requires the minimal number of signalling message transmissions, it does not reflect theheavy processing load at the single ST, so that its delay performance is inferior to signalling architectures3 and 4, which distributes the ST processing load among all the MAN nodes.4.3.5 Performance of Call Setup ProceduresIn chapter 3, call setup performance over a single MAN, in terms of mean call setup delays vs.PA throughput, was evaluated with respect to different signalling architectures by simulations. Here weextend this investigation to the multiple interconnected-MAN network configuration being considered inthis chapter, and evaluate call setup performance with respect to the two alternative, i.e. sequential andoverlapping, call setup procedures. Due to the limitation of computer resources, a single computer modelencompassing the interconnected MANS for simulating end-to-end call setup is not attainable. Instead,Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 81Figure 42. Isochronous Channel Setup for Signalling Architecture 3we obtain approximations to the end-to-end call setup delays by means of delay decomposition. Thefour interconnected MANs are assumed to be independent in signalling transport and equally loaded.The simulations are performed using a single MAN (with the same configuration as the one consideredin chapter 3) which is considered representative of each of the four interconnected MANs. Detaileddelay statistics for signalling transfer between appropriate pairs of network elements are collected (seeAppendix B for some typical results) for each signalling architecture. The end-to-end call setup delay isthen calculated by summing up the appropriate delay elements according to the call setup procedure.Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 82Figure 43. Isochronous Channel Setup for Signalling Architecture 4Figure 44 and 45 compare the mean call setup delays of the four different signalling architecturesusing the sequential and overlapping call setup procedures, respectively. As expected, for both sets ofprocedures, signalling architecture 4 gives the best call setup performance. Also, it is apparent that the CBarchitectures (2 and 4) result in a much more substantial performance improvement over the correspondingOB architectures (1 and 3) when the sequential call setup procedure is employed, compared to the caseemploying the overlapping call setup procedure. This is because the CB architecture mainly brings aboutimprovements in the time to setup isochronous channels, which effect is very much reduced with the useof the overlapping call setup procedure.Figure 46 shows the comparison of the results from Figure 44 and 45. It shows that for eachsignalling architecture, the call setup delay for the overlapping procedure is significantly shorter than thatfor the sequential procedure. Considering the minimum call setup delays shown in the graph, the savingsin call setup delays for the overlapping procedure over the sequential procedure are approximately 2.2ms,1.6ms, 2.4ms and 1.8ms for signalling architectures 1, 2, 3 and 4, respectively, which correspond topercentage savings of 23%, 19%, 32% and 27%. Overall, the combination of signalling architecture 4and overlapping call setup procedure yields the best call setup performance.141413 SIGNALLING ARCHITECTURES1. SST, SBWM, OB2. SST, SBWM, CB-e3. MST, BBWM, OB ^04. MST, SBWM, CB 121311109• .............. ..... • .........87 76200 210 2206230 240 250 260 270 280 290Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 83PA Throughput (Mbps)Figure 44. Mean Call Setup Delay vs. PA Throughput for Sequential Call SetupFigure 45. Mean Call Setup Delay vs. PA Throughput for Overlapping Call Setup4.4 Call Handoff ProceduresThe following discussions of call handoff procedures are focussed on the seamless transfer ofthe call connection within the MAN-based PCN, from the BS (BS.p) with which the MT undergoing a'^I^'^I^' I^'^1^'^1^'^IC/)C0TD• 812141354200 210 220 230 240 250 260 270 280 2906975486971013121114MST,SBWM,CB Overlap Scheme _e— MST,BBWM,OB Overlap Scheme ^0SST,SBWM,CB Overlap Scheme --e,--SST,SBWM,OB Overlap Scheme ^^ MST,SBWM,CB Sequential Scheme- MST,BBWM,OB Sequential Scheme- SST,SBWM,CB Sequential SchemeSST,SBWM,OB Sequential Scheme• ................................ .......... ---dChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 84PA Throughput (Mbps)Figure 46. Comparison of Call Setup Delays between Sequential and Overlapping Call Setuphandoff is associated previous to the handoff, to the next BS (BS.n) with which the MT will be associatedafter the handoff. The general call handoff procedure consists of the following steps.1. The MT signals BS.n to request a call handoff. The request includes the identities of the MT aswell as BS.p. If BS.n is able to accept the handoff request, it acknowledges the request by assigninga radio channel to the MT.2. BS.n sends a REQ_INF message to BS.p over the PCN to request for call information. BS.p respondsby providing the requested information in a SPY_INF message.3. BS.n compares the hierarchical address of BS.p with its own address to ascertain the type of handoffand the change in call connectivity resulting from the handoff, and takes appropriate actions to setup the facilities required to connect itself to the terminating BS (BS.t) at the other end of the call.The possible actions are:a. retaining the existing isochronous connection to the common BSC shared by BS.p and BS.n,and re-routing the BSC connection from BS.p to BS.n;b. re-routing one end of an existing isochronous channel from BS.p to BS.n;Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 85c. re-routing one end of an exiting isochronous channel from BS.p to a transit BG;d. re-routing one end of an exiting isochronous channel from a transit BG to BS.n;e. re-routing one end of an exiting isochronous channel from a transit BG to another transit BG;f. establishing a new isochronous channel to connect BS.n to a transit BG;g. establishing a new isochronous channel to connect BS.n to BS.t;h. establishing a new isochronous channel to connect two transit BGs over a transit MAN;i. establishing a new isochronous channel to connect BS.t to a transit BG;j. establishing a new connection to BS.t through the common BSC shared by BS.n and BS.t.4. BS.n sends CFM_HO messages to BS.p and the MT to confirm the completion of the call handoffover the network.5. The MT continues the call via BS.n.6. BS.p releases the radio channel previously assigned to the call, and signals the MAN to releaseisochronous channels in the previous connection to BS.p that are not used in the new connectionto BS.n.The actual action taken by BS.n in step 3 to setup facilities for the new connection to BS.n in theabove general call handoff procedure depends on the type of handoff and the change in call connectivityresulting from the handoff, and is tabulated in Table 8. Three types of call handoff are identified:• Intra-MAN-node handoff — BS.n and BS.p are attached to a common BSC (at the same MAN node);• Intra-MAN handoff BS.n and BS.p are attached to different MAN nodes over the same MAN;• Inter-MAN handoff — BS.n and BS.p are attached to MAN nodes over different MANs.Likewise, there are three types of call connectivity, depending on the relative positions of the two BSs(BS.n/BS.p and BS.t) terminating either ends of the call connection:• Intra-MAN-node call — the two BSs are attached to a common BSC (at the same MAN node);• Intra-MAN call — the two BSs are attached to different MAN nodes over the same MAN;• Inter-MAN call — the two BSs are attached to MAN nodes over different MANs.Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 86Table 8. Relationships between Type of Handoff,Change in Connectivity, and Facility Setup RequirementsType of HandoffConnectivity beforeHanfoffConnectivity afterHandoffBS.n Actions inProcedure Step 3Intra-MAN-nodeIntra-MAN-node Intra-MAN-node a/jIntra-MAN Intra-MAN aInter-MAN Inter-MAN aIntra-MANIntra-MAN-node Intra-MAN gIntra-MANIntra-MAN-node jIntra-MAN bInter-MAN Inter-MAN bInter-MANIntra-MAN-node Inter-MAN i, [h,^...^, h,]^fIntra-MAN Inter-MAN c, [h,^. . .^, h,] fInter-MANIntra-MAN-node jIntra-MAN dInter-MAN e, [h,^. . .^, h,] fBS.pBS.n^• S.tIntraMAN to IntraMANFigure 47. Intra-MAN Handoff with Infra-MAN ConnectivityFigures 47 and 48 show examples of different call connectivities in an intra-MAN and an inter-MANcall handoff, respectively.To re-route an existing isochronous channel within a MAN to a new node, the existing VCIis retained. EXG_INF messages are exchanged between the new node and the node at the other endof the channel to determine each node's direction of transmissions on the buses. Figure 49 illustratesthis procedure in an infra-MAN handoff situation where the connectivity remains intra-MAN before andafter the handoff. The signalling procedures for establishing a new isochronous channel and releasingan unused isochronous channel are as described in the previous section. Specifically, detail proceduresc. interiviAiN to interiviAINFigure 48. Inter-MAN Handoff with Different Changes in Connectivityare given in Section 4.3.4 for isochronous channel establishments using different signalling architectureswithin the MAN. To reduce the probability of handoff failures due to unavailability of facilities, somePA slots and VCIs could be set aside in each MAN for allocation specifically to handoff connections.As Table 8 indicates, except for three of the five inter-MAN call handoff cases, call handoffprocedures over the MAN-based PCN architecture are relatively simple. This is particularly advantageousin reducing the call handoff delay and minimizing the probability of call handoff failures.Figure 50 shows an example of the inter-MAN handoff procedure. In this scenario, the BS.n,BS.p and the BS.t are in different MANS as shown in Figure 48 case c. The handoff procedure requiresthe following activities:Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 88• The MT sends a HO_REQ message to BS.n indicating a handoff request.• When BS.n receives the HO_REQ message, it returns a HO_PD message and a RES_RAD messageto the MT indicating the handoff request is being processed. Meanwhile, BS.n sends a REQ_INFmessage to BS.p to retrieve the original call related information which includes the routing infor-mation for the old call.• Upon receiving the REQ_INF message, BS.p returns a SPY_INF message to BS.n.• BS.n analyzes the information received from BS.p. By comparing the hierarchical addresses of eachBS involved in this call, it determines which kind of handoff is being requested and the resultingchange in call connectivity. In this case, it is an inter-MAN handoff within an inter-MAN to inter-MAN connectivity change. A REQ_ISOC message is sent to the ST in BS.n's MAN, and a HO_REQmessage is sent to BG2 interconnecting the MAN of BS.n and the MAN of BS.t. The purpose ofChapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 89this message is to request the MAN of BS.t to re-use the existing isochronous channel for the oldcall and update the routing table in BG2 to interconnect the two MANs.• The new isochronous channel on the MAN of BS.n is set up using the procedure described before.The ST sends a CFM_ISOC message to BS.n when the isochronous channel is ready. Meanwhile,BS.t and BG2 exchange information between each other using EXG_INF messages and sends aHO_PD message to BS.n indicating that the isochronous channel is ready.• After BS.n receives both the CFM_ISOC message from the ST and the HO_PD message from BG2,it confirms that the new handoff channel is ready and sends a CFM_HO message to the MT, BG2,BS.t and BS.p to indicate the continuation of the call over the new channel and the release of theold channel.• When BS.p receives the CFM_HO message, it releases the isochronous channel between BS.p andBG1, and the radio channel between BS.p and the MT.The procedure described above insures minimum cut off time for the call during handoff, andcan extend to other handoff procedures.Table 9 shows the simulation results for the inter-MAN handoff procedure in case c of Figure 48.This is representative of the most complicated handoff process. The same simulation model is used asin the evaluation of call setup procedures. Signalling architecture 4 is employed in this simulation. Theresults show that the call handoff delays are between 4 and 5.2ms at different levels of PA throughput inthe network. Other handoff cases are simpler and have shorter delays. The intra-MAN handoff delays arearound 1 to 2ms using the procedure described before. The results confirm that the distributed MAN-basedarchitecture facilitates simplification of the handoff procedure and minimization of the handoff delay.Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN 90Chapter 4. Signalling Procedures for Call Setup, Clearing and Handoff in PCN^ 91Table 9. Inter-MAN Call Handoff Delays vs. PA ThroughputPA Throughput(Mbps)Percentage ofMaximumThroughputInter-MAN HandoffDelay (ms)204 73% 4.18222 79% 4.34240 86% 4.40257 91% 4.58274 97% 4.79278 99% 5.2192Chapter 5. ConclusionsSince worldwide needs for PCS are growing so rapidly, the current centralized wireless commu-nication network architectures may not he suitable to handle the increased processing load resulting fromthe anticipated demand for PCS in this environment. A distributed microcellular network architecturebased upon the IEEE 802.6 MAN has been proposed as a means of distributing the processing load andproviding satisfied PCS services. Efficient and reliable signalling is required to enable such a networkarchitecture, and is the subject studied in this thesis.5.1 Summary of the ResultsIn this thesis, the architectural and procedural design and performance of call control signallingin the distributed MAN-based PCN have been thoroughly examined and extensively evaluated. Asanalytical models are not available, simulation models have been employed and implemented to evaluatethe general performance of the IEEE 802.6 MAN, which forms the basis of this distributed architecture.The bandwidth balancing and erasure-node schemes to improve the general performance and fairness ofthe DQDB protocol have been evaluated and considered to facilitate call control signalling. With trafficof only one priority level in the network, the bandwidth balancing scheme has been shown to be effectivein improving fairness in QA network access. However, with multi-priority level traffic in the network, thebandwidth balancing scheme is not able to give consistent performance improvements in the network. Onthe other hand, the erasure-node scheme improves network performance at all priority levels in terms ofaccess fairness, network throughput, and cell transmission delay. Different methods of transporting voiceand signalling traffic over the IEEE 802.6 MAN have been investigated. The scheme using PA accesswith silence detection to transport voice traffic and priority 2 QA access to transport signalling traffichas been adopted. The performance of using this scheme has been examined and compared with otheralternatives. The relationships between PA traffic capacity and QA signalling delay have been establishedusing the simulation models, and the effects of alternative signalling architectures with different busconfigurations and location of signalling network elements have been investigated. The best signallingChapter 5. Conclusions^ 93performance is obtained with the signalling termination function distributed among all the MAN nodes,and the bandwidth manager function incorporated in the HOB stations. The closed bus configuration,where the HOB stations for the two buses are physically collocated, gives better performance than theopen bus configuration, where the HOB stations for the two buses are physically separate. Signallingperformance can be improved by employing erasure nodes to recover QA slots already read by theirdestinations. The optimal location for a single erasure node on each bus has been empirically determinedby simulations. Logical procedures for call setup, clearing and handoff have been established for callsbetween mobile subscribers, through the access MANs hosting the respective subscribers, and backboneMANs interconnecting the access MAN's by means of bridges. In particular, call setup procedureshave been analyzed by separating the process into several disjoint stages and considering alternativeprocedures for each stage. The sequential and overlapping call setup procedures have been proposed,the latter facilitating parallelism in call setup processing resulting in reduced call setup delays. Callhandoff procedures have been classified in terms of the type of handoff and the change in connectivityresulting from the handoff.5.2 Issues for Future StudiesAreas for further research could include the following:• call control signalling across heterogeneous networks;• call control signalling in multi-service environments;• co-operative handoff procedures based upon the MT's measurement of radio link quality as wellas the network's evaluation of other performance criteria, such as the traffic and processing loadof the network.Chapter 5. Conclusions^ 94Bibliography[1] D.C. Cox, "Personal communications- a viewpoint," IEEE Commun. Mag., pp. 8-92, Nov. 1990.[2] G. Calhoun, Digital Cellular Radio. Norwood, MA, Artech House: Englewood Cliffs, 1988.[3] R.J. Holbeche, Land Mobile Radio System. Peter Peregrinus Ltd., London, UK: Peter PeregrinusLtd., 1985.[4] D.J. Goodman, "Second generation wireless information networks," IEEE Trans. on Vehicular Tech.,vol. 40, pp. 366-374, May 1991.[5] A.E. Detecon, "Radio access protocol of the new GSM land mobile communication standard," Proc.IEEE VTC, pp. 30-37, 1988.[6] D.J. Goodman, "Cellular packet communications," IEEE Trans. Commun., vol. 38, pp. 1272-1280,Aug. 1990.[7] J.A. Audestad, "GSM: General overview of network functions," Proc. Int. Conf. on Digital LandMobile Radio Commun., pp. 59-79, 1987. Venice, Italy.[8] A. Nakajima, "Intelligent digital mobile communications network architecture for universal personaltelecommunications (UPT) services," Proc. IEEE VTC, pp. 83-87, 1991.[9] D.J. Goodman, "Trends in cellular and cordless communications," IEEE Commun. Mag., pp. 31-40,Jun. 1991.[10]B.J. Mallinder, "An overview of the GSM system," the Third Nordic Semi. Digi. Land Mobile RadioCommun., pp. 3.1.1-3.1.13, Sep. 1988. Copenhagen, Denmark.[11]M. Pautet, "GSM protocol architecture: radio sub-system signalling," Proc. IEEE VTC, pp. 326-332,1991.[12]EIA, "Dual-mode subscriber equipment network equipment compatibility specification," InterimStandard 54, Dec. 1989.Chapter 5. Conclusions^ 95[13] R.S. Swain and A.J. Motley, "Requirements of an advanced cordless telecommunication product,"Proc. IEEE ICC, pp. 81-85, June 1987. Seattle, WA.[14] H. Ochsner, "DECT-digital European cordless telecommunications," Proc. IEEE VTC, pp. 718-721,May 1989. San Francisco, CA.[15] D. Akerberg, "Properties of a TDMA pico cellular office communication system," Proc. IEEE GlobalCommun. Conf, pp. 1343-1349, Dec. 1988. Hollywood, FLA.[16] D.C. Cox, H.W. Arnold and P.T. Porter, "Universal digital portable communications: a systemsperspective," IEEE Journ. on Select. Areas of Commun., vol. SAC-5, pp. 764-773, June 1987.[17] Y.N. Doganata, T.X. Brown and E.C. Posner, "Call setup strategy tradeoffs for universal digitalportable communications," Comput. Networks and ISDN Syst., vol. 20, pp. 455-464, 1990.[18] B. Chartrs and C. Vernhes, "Mobile application part design principles," Proc. ISS, 1990.[19] A. Nakajima, K. Yamamoto and H. Sawada, "Automatic pursuit routing for mobile communicationsnetwork ," Inst. of Elec. and Commun. Eng., pp. 81-91, May 1989.[20]T. Hattori and H. Sekiguchi, "Personal communication -concept and architecture," Proc. IEEE ICC,pp. 1351-1357, 1990.[21] R.S. Wolff, S. Parlamas and D. Hakim, "A functional model and analysis of personal communicationsservices," Proc. IEEE ICC, pp. 356.4.1-356.4.7, 1992.[22] P.A. Ramsdale, "Personal communications - towards the third genegation," The 3rd IEE Conf. onTelecomm., pp. 214-218, 1990.[23] A.D. Malyan, L.J. Ng, R.W. Donaldson and V.C. M. Leung, "A microcellular interconnectionarchitecture for personal communication networks," Proc. IEEE VTC, May 1992.[24] A.D. Malyan and V.C.M. Leung, "Network architecture and signalling for urban personal commu-nications," Proc. IEEE Int. Conf. on Select. Topics in Wireless Commun., pp. 305-309, Jun. 1992.[25] L.J. Ng, R.W. Donaldson and A.D. Malyan, "Distributed architectures and databases for intellignetpersonal communication networks," Proc. IEEE Int. Conf. on Select. Topics in Wireless Commun.,Jun. 1992. Vancouver, B.C.Chapter 5. Conclusions^ 96[26] CCITT Blue Book, "Physical/electrical characteristics of hierarchical digital interfaces," Recom.G.703, 1988.[27] ANSI, "American national standard for telecommunications digital hierarchy-electrical interfacesand formats specifications," TI.102, T1.107, 1987.[28] CCITT Blue Book, "Network node interface and synchronous multiplexing structure," Recom. G.707 -709, 1988.[29] IEEE 802.6 Committee, " Draft of proposed standard 802.6 distributed queue dual bus (DQDB),"IEEE Standard, Aug. 1990.[30] E.C. Russell, Building Simulation Models with Simscript 11.5. CACI, 1990.[31] C. Bisdikian, "Waiting time analysis in a single buffer DQDB (802.6) network," Proc. IEEEINFOCOM, pp. 610-616, June 1990.[32]P. Potter and M. Zukerman, "A discrete shared processor model for DQDB," Proc. 7th ITC Sem.,Spet. 1989.[33] E.L. Hahne, "DQDB networks with and without bandwidth balancing," IEEE Trans. Commun.,vol. 40, no. 7, pp. 1192-1204, 1992.[34] V. Catania and L. Mazzola, "Throughput analysis of DQDB in overload condition," Proc. IEEEICC, pp. 741-747, 1990.[35] J. Filipiak, "Access protection for fairness in a distributed queue dual bus metropolitan area network,"Proc. IEEE ICC, pp. 639-639, June 1989.[36] M. Zukerman and P.G. Potter, "A protocol for eraser node implementation within the DQDBframework," Proc. IEEE ICC, pp. 1400-1404, 1990.[37] M. Rodrigues, "Erasure node: performance improvements for the IEEE 802.6 MAN," Proc. IEEEINFOCOM, pp. 636-643, 1990.[381 M. Zukerman, "Circuit allocation and overload control in a hybrid switching system," Comput.Networks and ISDN Syst., vol. 16, pp. 281-298, 1989.Chapter 5. Conclusions^ 97[39]L.J. Ng, "A distributed architectures and databases for personal communication network services,"M.A.Sc. Thesis in E.E. Dept. UBC, Mar. 1993.[40]Helgert and J. Hermann, Integrated Services Digital Networks. California, U.S.A.: Addision-WesleyPublishing Company, Inc., 1991.[41] W. Stallings, ISDN and Broadband ISDN. New York: Macmillan Publishing Company, 1992.[42]R. Steele and M. Nofal, "Teletraffic performance of city street microcells catering for pedestrianmobile users," 1EE colloq., pp. 9/1-9/6, Nov. 1990.[43]E.D. Sykas, "Numbering and addressing in IBCN for mobile communications," Proc. IEEE ICC,vol. 79, pp. 230-240, Feb. 1991.[44]J. Homa and S. Harris, "Intelligent network requirements for personal communications services,"IEEE Commun. Mag., pp. 70-76, Feb. 1992.[45]S.M. Boyles, "Common channel signalling: the nexus of an advanced communications network,"IEEE Commun. Mag., pp. 57-63, July 1990.[46]B. Jabbari, "Intelligent network concepts in mobile communications," IEEE Commun. Mag., vol. 30,pp. 64-69, Feb. 1992.[47]R.B. Robrock, "The intelligent network-changing the face of telecommunications," Proc. IEEE ICC,vol. 79, pp. 7-20, Jan. 1991.[48] W.R. Byrne, G. Clapp and H.J. Kafka, "Evolution of metropolitan area networks to broadband ISDN,"IEEE Commun. Mag., vol. 29, pp. 69-82, Jan. 1991.[49]J. Filipiak, "Shaping interworking MAN's into an evolutionary B-ISDN," Comput. Networks andISDN Syst., vol. 20, pp. 343-349, 1990.[50]M. Wizgall, W. Weiss and W. Stier, "The ISDN approach for mobile radio," Proc. ISS, 1990.[51]A. Nakajima and K. Yamamoto, "Advanced mobile communication network based on signalingsystem NO.7," Proc. ISS, pp. 747-751, 1987.[52]K. Kohiyama, T. Hattori and H. Sekiguchi, "Advance personal communication system," Proc. IEEEVTC, pp. 161-166, 1990.Chapter 5. Conclusions^ 98[53] A.R. Modarressi and R.A. Skoog, "Signaling systems no.7: a tutorial," IEEE Commun. Mag., pp. 19-35, July 1990.[54] S. Rappaport, "Modeling the handoff problem in personal communications networks," Proc. IEEEVTC, pp. 517-523, 1991.[55] A.D. Kamal, "Efficient multi-segment message transmission with slot reuse on DQDB," Proc. IEEEICC, pp. 869-878, 1991.[56] K.J. Bye, "Handover criteria and control in cellular and microcellular systems," Proc. IEEEINFOCOM, pp. 91-98, 1990.[57] M.W. Garrett and S.Q. Li, "A study of slot reuse in dual bus multiple access networks," Proc. IEEEINFOCOM., pp. 617-629, 1990.[58] S.T. Chia, "The control of handoff initiation in microcells," Proc. IEEE INFOCOM, pp. 531-536,1991.[59] M. Conti and E. Gregori, "A methodological approach to an extensive analysis of DQDB performanceand fairness," IEEE on Sele.in Commun., vol. 9, pp. 76-86, Jan. 1991.[60] S.Y. Kim amd K.J. Han, "An enhanced bandwidth balacing mechanism of the DQDB MAN," Proc.IEEE ICC, pp. 423-427, 1991.[61]E.L. Hahne and A.K. Choudhury, "Improving the fairness of distributed queue," Proc. IEEEINFOCOM, pp. 175-184, 1990.[62] E.L. Hahne and N.F. Maxemchuk, "Fair access of multi-priority traffic to distributed queue dual busnetworks," Proc. IEEE ICC, pp. 889-900, 1991.[63] C. Bisdikian, "Waiting time analysis in a single buffer DQDB (802.6) network," Proc. IEEEINFOCOM, pp. 610-616, 1990.[64] S.J. Jeon and G.T. Kim, "Effectiveness of priority schemes of the DQDB and the FDDI," Proc.IEEE ICC, pp. 736-740, 1991.[65] M.P. Spratt, "Bandwidth allocation in IEEE 802.6 using non unity ratio bandwidth balancing," Proc.IEEE ICC, pp. 729-735, 1991.Chapter 5. Conclusions^ 99[66] M. Zukerman, "Bandwidth allocation for bursty isochronous traffic in a hybrid switching system,"IEEE Trans. Commun., vol. 37, pp. 1367-1371, Dec. 1989.[67] B.G. Kim, "Packet delays in the IEEE 802.6 DQDB protocol," Proc. IEEE ICC, pp. 1692-1696, 1990.Appendix A. Acronyms and Abbreviations^ 100Appendix A List of Acronyms and AbbreviationsACF^Access control fieldACK Acknowledge messageACM^Address complete messageALT Alerting messageANM^Answer messageATM Asynchronous transfer modeBB WM^Bi-bandwidth managerBG BridgeB-ISDN^Broadband integrated services digital networkBS Base stationBSC^Base station controllerBWB Bandwidth balancingBWM^Bandwidth managerCB Closed bus — head of bus stations are collocatedCCITT^International Telegraph and Telephone Consultative CommitteeCDC Count-down counterCFM_HO^Confirm handoff messageCFM_ISOC^Confirm isochronous channel messageCFM_VCI^Confirm VCI messageCM Call managementCNT^Connect messageCNT_ACK^Connect acknowledge messageCPD Call proceeding messageCPG^Call progressing messageAppendix A. Acronyms and Abbreviations^ 101CT2^A cordless phone system in BritainDB DatabaseDECT^A cordless phone system in EuropeDISC Disconnect messageDISC_RAD^Disconnect radio channel messageDMPDU Derived message protocol data unitDQDB^Distributed queue dual busEN Erasure nodeEXG_INF^Exchange information messageGSM Group special module — European digital cellular systemHCS^Header check sequenceHDB Home databaseHLR^Home location registerHM Handoff managementHOB^Head of BusHO_PD Handoff proceeding messageHO_REQ^Handoff request messageIAM Initial address messageIE^Information elementIMPDU Initial message protocol data unitINF^InformationINQ Inquire messageINQ_CLE^Inquire callee informationIS-54 An American cellular telephone standardISDN^Integrated Services Digital NetworkISOC Isochronous channelIS UP^ISDN user partAppendix A. Acronyms and Abbreviations^ 102IWU^Interworking unitLAN Local area networkMA^Metropolitan areaMAC Medium access controlMAN^Metropolitan area networkMM Mobility managementMSC^Mobile switching centerMST Multiple signal terminationMT^Mobile terminalN-ISDN Narrow-band integrated services digital networkNTY_VCI^Notify VCI messageOB Open bus — head of bus stations are separatedPA^Pre-ArbitratedPBX Private branch exchangePCD^ProceedingPCM Pulse code modulationPCN^Personal Communication NetworkPCS Personal Communication ServicesPIN^Personal identification numberPSR Previous segment read bitPSTN^Public switch telephone networkQA Queue-ArbitratedQOS^Quality of ServiceREL ReleaseRCP^Release completeRCP_ISOC^Release complete of ISOC messageRCP_VCI Release complete of VCI messageAppendix A. Acronyms and Abbreviations^ 103REQ^Request messageREQ_INF^Request information messageREQ_ISOC^Request isochronous channel messageRES_RAD^Reserve radio channel messageRES_VCI Reserve VCI messageRQC^Request counterSAM Substantial address messageSAR^Segmentation/ReassemblySBWM Single bandwidth managerSDH^Synchronous data hierarchicalSPY_INF Supply information messageS_RST^Search result messages/s Silence/speechSS7^signalling system No.7SST Single signalling terminationST^Signalling TerminationSTM Synchronous transfer modeSTP^Setup messageSTP_ACK^Setup acknowledge messageTDMA Time division multiple accessTGW^Trunk gatewayUNI User-Network InterfaceVCI^Virtual circuit identifierVDB Visiter databaseVHF^Very high frequencyVLR Visitor location registerAppendix B. Typical Data^ 104Appendix B Sample of Typical DataScheme Data Names Mean Cell Trans. Dly. STD of Dly.Scheme 1OB, SST, SBWMNode_i to ST 0.17438 0.11854ST to HOB 0.21313 0.02781HOB to ST 0.23744 0.06862ST to Node_i 0.18036 0.12794Scheme2CB,SST, SBWMNode_i to HOB 0.12413 0.06689HOB to Node_i 0.20316 0.127140Scheme3OB,MST,BBWMNode_i to HOB 0.24557 0.15685HOB to Node_i 0.25311 0.13285Nodei to Node j 0.20087 0.13109Scheme4CB,MST,SBWMNode_i to HOB 0.12389 0.06668HOB to Node_i 0.18317 0.110330Node_i to Node j 0.19617 0.11891Table 10. Cell Transmission Delays at PA Throughput = 204MbpsScheme Data Names Mean Cell Trans. Dly. STD of Dly.Scheme 1OB, SST, SBWMNode_i to ST 0.19453 0.13927ST to HOB 0.21545 0.02546HOB to ST 0.25670 0.09193ST to Node_i 0.20086 0.14821Scheme2CB,SST, SBWMNode_i to HOB 0.13202 0.07009HOB to Node_i 0.214520 0.12668Scheme3OB,MST,BBWMNode_i to HOB 0.26153 0.17470HOB to Node i- 0.28426 0.13899Node_i to Node _j 0.22204 0.13656Scheme4CB,MST,SBWMNode_i to HOB 0.13201 0.06999HOB to Node i_ 0.213060 0.12644Node_i to Node _j 0.22128 0.13526Table 11. Cell Transmission Delays at PA Throughput = 240.8MbpsAppendix B. Typical Data^ 105Scheme Data Names Mean Cell Trans. Dly. STD of Dly.Scheme 1()B, SST, SBWMNode_i to ST 0.21056 0.14908ST to HOB 0.21763 0.02893HOB to ST 0.25847 0.09160ST to Nodei 0.22166 0.15800Scheme2CB,SST, SBWMNode_i to HOB 0.13270 0.06898HOB to Node_i 0.23401 0.12981Scheme3OB,MST,BBWMNode_i to HOB 0.26805 0.18241HOB to Node_i 0.29798 0.13729Node_i to Node j 0.23146 0.15514Scheme4CB,MST,SBWMNodei to HOB 0.123264 0.06890HOB to Node_i 0.23285 0.12964Node_i to Node _j 0.22835 0.13836Table 12. Cell Transmission Delays at PA Throughput = 256.6MbpsScheme Data Names Mean Cell Trans. Dly. STD of Dly.Scheme 1OB, SST, SBWMNode_i to ST 0.22499 0.16151ST to HOB 0.22493 0.03399HOB to ST 0.33859 0.15117ST to Node_i 0.24592 0.18805Scheme2CB,SST, SBWMNode_i to HOB 0.13535 0.07033HOB to Node_i 0.25598 0.13806Scheme3OB,MST,BBWMNode_i to HOB 0.27636 0.19032HOB to Node_i 0.32753 0.14403Node_i to Node_ 0.24380 0.14959Scheme4CB,MST,SBWMNode_i to HOB 0.13503 0.06962HOB to Node_i 0.25535 0.13755Node_i to Nodej 0.24077 0.14523Table 13. Cell Transmission Delays at PA Throughput = 273.8MbpsAppendix B. Typical Data^ 106Scheme Data Names Mean Cell Trans. Dly. STD of Dly.Scheme 1()B, SST, SBWMNode_i to ST 0.23927 0.18540ST to HOB 0.23068 0.03809HOB to ST 0.41657 0.18607ST to Node_i 0.26721 0.20718Scheme2CB,SST, SBWMNode_i to HOB 0.13687 0.07164HOB to Node_i 0.28895 0.15843Scheme3OB,MST,BBWMNode_i to HOB 0.28635 0.20995HOB to Node_i 0.37137 0.16268Node_i to Node_j 0.25707 0.17375Scheme4CB,MST,SBWMNode_i to HOB 0.13644 0.07156HOB to Node_i 0.28460 0.15716Node_i to Node_j 0.24759 0.16047Table 14. Cell Transmission Delays at PA Throughput = 276.8MbpsScheme Data Names Mean Cell Trans. Dly. STD of Dly.Scheme 1OB, SST, SBWMNode_i to ST 0.27014 0.26728ST to HOB 0.23285 0.04028HOB to ST 0.51952 0.43670ST to Node_i 0.30297 0.28995Scheme2CB,SST, SBWMNode_i to HOB 0.14000 0.07756HOB to Node_i 0.32372 0.23683Scheme3OB,MST,BBWMNode_i to HOB 0.29784 0.24720HOB to Node i 0.43884 0.32207Node_i to Node_j 0.29293 0.25871Scheme4CB,MST,SBWMNode_i to HOB 0.14050 0.09865HOB to Node i- 0.32058 0.23509Node_i to Node_j 0.26816 0.21279Table 15. Cell Transmission Delays at PA Throughput = 277.5MbpsAppendix C. Source Code^ 107Appendix C Sample of Source CodeThe simulation program is optimized for speed. Each simulation process is designed to minimizethe number of computations. A sample of the source code is given in the following pages.-April 15 1993 Source Code Name: q13d3.sim"The PREMAMBLE routine gives a static description of each modelling"element. Such as temporary element, permanent element, process and"general variables, etc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l, Z2,SD,Z3,Y,Y1,Y2,Y3,Y4,Y5,Y6,Y7,Y8,Y9,Y10,DIVISION, SWITCH,NODEA,NODEB AS INTEGER VARIABLESDEFINE SLOT, CELLO.INTERARRIVAL.TIME, CELL1.INTERARRIVAL.TIME,CELL2.INTERARRIVAL.TIME, STOP.TIME, DO.DELAY, Dl.DELAY, D2.DELAY,El.DLY,E2.DLY,E3.DLY,E4.DLY,E5.DLY,E6.DLY,E7.DLY,E8.DLY,E9.DLY,E10.DLY,SEARCH, S.DELAY, TOT.ERL, ERL.PER.CELL,CALL.RATE.PER.CELL, LOADO, LOAD1, LOAD2, AREA, P.LOAD,CALL.PER.HOUR, CALL.HOLD.TIME,CALL.ARV.RATE,T.DELAY, CALL.DPT.RATE, X2, X3, X4, X5, X6 AND X7 AS REALVARIABLESTALLY MINDDO AS THE MINIMUM, MAXDDO AS THE MAXIMUM,MEANDDO AS THE MEAN AND SDDO AS THE STD OF DO.DELAYTALLY MINDD1 AS THE MINIMUM, MAXDD1 AS THE MAXIMUM,MEANDD1 AS THE MEAN AND SDD1 AS THE STD OF Dl.DELAYTALLY MINDD2 AS THE MINIMUM, MAXDD2 AS THE MAXIMUM,MEANDD2 AS THE MEAN AND SDD2 AS THE STD OF D2.DELAYTALLY MINSD AS THE MINIMUM, MAXSD AS THE MAXIMUM,MEANSD AS THE MEAN AND SSD AS THE STD OF S.DELAYTALLY MINTMD AS THE MINIMUM, MAXTMD AS THE MAXIMUM,MEANTMD AS THE MEAN AND STMD AS THE STD OF T. DELAYTALLY MINE' AS THE MINIMUM, MAXE1 AS THE MAXIMUM,MEANE1 AS THE MEAN AND SDE1 AS THE STD OF El.ELYTALLY MIHE2 AS THE MINIMUM, MAXE2 AS THE MAXIMUM,MEAME2 AS THE MEAN AND SDE2 AS THE STD OF E2.ELYTALLY MINES AS THE MINIMUM, MAXE3 AS THE MAXIMUM,MEANE3 AS THE MEAN AND SDE3 AS THE STD OF E3.DLYTALLY MINE? AS THE MINIMUM, MAXE4 AS THE MAXIMUM,MEANE4 AS THE MEAN AND SDE4 AS THE STD OF E4.DLYTALLY MINES AS THE MINIMUM, MAXES AS THE MAXIMUM,MEANE5 AS THE MEAN AND SDE5 AS THE STD OF E5.DLYTALLY MINES AS THE MINIMUM, MAXE6 AS THE MAXIMUM,•EANE6 AS THE MEAN AND SDE6 AS THE STD OF E6. ELYTALLY MINE% AS THE MINIMUM, MAXE7 AS THE MAXIMUM,MEANE7 AS THE MEAN AND SDE7 AS THE STD OF E7.DLYTALLY MIME8 AS THE MINIMUM, MAXE8 AS THE MAXIMUM,MEANE8 AS THE MEAN AND SDE8 AS THE STD OF E8.DLYTALLY MINES AS THE MINIMUM, MAKES AS THE MAXIMUM,MEANE9 AS THE MEAN AND SDE9 AS THE STD OF ES.DLYTALLY MINE10 AS THE MINIMUM, MAXE10 AS THE MAXIMUM, O00MEANE10 AS THE MEAN AND SDE10 AS THE STD OF ElO.DLYDEFINE .SECONDS TO MEAN DAYSDEFINE .MILLISECONDS TO MEAN HOURSDEFINE .MICROSECONDS TO MEAN MINUTESEND"The major function of MAIN routine is to input variable parameters"into the system,^setup the basic configuration for the system,"schedule the beginning time for SETUP process; Generator 0,^1,^2" process; Update process and Stop process.MAINDEFINE I AS INTEGER VARIABLESDEFINE CON.START AND CON.START1 AS REAL VARIABLESPRINT 1 LINES AS FOLLOWSPLEASE INPUT STATIONARY.PER.CELL AND THEN RETURN:READ STATIONARY.PER.CELLPRINT 1 LINES AS FOLLOWSPLEASE INPUT MOBILE.PER.CELL AND THEN RETURN:READ MOBILE.PER.CELLPRINT 1 LINES AS FOLLOWSPLEASE INPUT NUMBER OF NODES IN MAN AND THEN RETURN:READ N.STATIONPRINT 1 LINES AS FOLLOWSPLEASE INPUT NODE DISTANCE IN SLOT AND THEN RETURN:READ SDPRINT 1 LINES AS FOLLOWSPLEASE INPUT HOW MANY CALLS/PER HOUR IN EACH SUBSCRIBER ANDTHEN RETURN:READ CALL. PER. HOURPRINT 1 LINES AS FOLLOWSPLEASE INPUT MEAN CALL HOLDING TIME IN MINUTES AND THEN RETURN:READ CALL.HOLD.TIMEPRINT 1 LINES AS FOLLOWSPRINT 1 LINES AS FOLLOWSPLEASE INPUT BANDWIDTH MODE AND THEN RETURN:READ BWB.MODPRINT 1 LINES AS FOLLOWSPLEASE INPUT BUFFER CAPACITY AND THEN RETURN:READ BUFFER.CAPACITYPRINT 1 LINES AS FOLLOWSPLEASE INPUT STOP TIME AND THEN RETURN:READ STOP.TIMEPRINT 1 LINES AS FOLLOWSPLEASE INPUT DIVISION OF TIME INTERVAL AND THEN RETURN:READ DIVISIONPRINT 1 LINES AS FOLLOWSPROGRAMM CONTINUE, PLEASE WAIT!OPEN unit 2^for output,^file name is^"q13a"use 2 for outputCREATE EVERY STATIONFOR EACH STATIONDOLET CDAO(STATION)=-2LET CDA1(STATION)=-2LET CDA2(STATION)=-2LET CDBO(STATION)=-2LET CDB1(STATION)=-2LET CDB2(STATION)=-2LOOPLET LOADO=LOADO*1000000.LET CELLO.PER.SEC=LOADO/(44.*8.)IF CELLO.PER.SEC>0LET CELLO.INTERARRIVAL.TIME=1.:CELLO.PER.SECALWAYSLET LOAD1=LOAD1*1000000.LET CELL1.PER.SEC=LOAD1/(44.*8.)PLEASE INPUT LEVEL2 TRAFFIC LOAD IN MBPS AND THEN RETURN: IF CELL1.PER.SEC>0READ LOAN LET CELL1.INTERARRIVAL.TIME=1.,'CELL1.PER.SECPRINT 1 LINES AS FOLLOWS ALWAYSPLEASE INPUT LEVEL]. TRAFFIC LOAD IN MBPS AND THEN RETURN: LET LOAD2=LOAC2*1000000.READ LOAD' LET CELL2.PER.SEC=LOAD2:(44.*8.)PRINT 1 LINES AS FOLLOWS IF CELL2.PER.SEC>0PLEASE INPUT LEVELC TRAFFIC LOAD IN MBPS AND THEN RETURN: LET CELL2.INTERARRI•AL.TIME=1.:CELL2.PER.SECREAD LOADO ALWAYSPRINT 1 LINES AS FOLLOWS LET P.LOAD=310000000PLEASE INPUT ERASURE NODE SWITCH (0 1) AND THEN RETURN: LET ISOC.CHAN1=INT.F(4360./0.43)READ SWITCH LET RATIO=STATIONARY.PER.CELL/(STATIONARY.PER.CELL+PRINT 1 LINES AS FOLLOWS 2.*MOBILE.PER.CELL)PLEASE INPUT E_HOPE NUMBER IN BUS A H0' AS ALL NODES) AND THEN LET ERL.PER.CELL=MOBILE.PER.CELL*CALL.HOLD.TIME*CALL.PER.HOJ2R60.RETURN: LET ERL1.PER.CELL=STATIONARY.RER.CELL*CALL.HOLD.TIME*READ NOTEA CALL.PER.HOUR/60.PRINT 1 LINES AS FOLLOWS LET TOT.ERL=64.*(ERL1.PER.CELL +^2.*ERL.PER.OELL)PLEASE INPUT E_NOTE NUMBER IN BUS B (O AS ALL NOTES)^AND THEN LET CALL.RATE.PER.CELL=ERL.PER.CELL:24.RETURN: LET MOBILE.TOTAL.TALKED=OREAD NODEB LET TOTAL.SS=64.*^(STATIONARY.PER.CELL+2.*MOBILE.PER.CELL) OLET Ho.MAHREs=0LET IsC ,C.CHAN=ISOc.cHAN1-HOMANRESLET HO.CELLREs=6LET NO.TALKINc=0LET CON.START=0LET cON.START1=0LET V.LosT=SLET cALL.LosT,LET cSP.LosT= •LET AREA=0.^33LET sEAROH=_ , 01LET HouP.S.7=1L.. .LET MINriTES.•I=100LET SLOT=2.7354839LET N.SLOTA=SD*II.STATION-SD+1LET N.SLOTB=sD*N.sTATION-SD+1LET M.STATIoN=INT.F;FFSTATION/2. +1LET Y=0LET 11=5LET 12=1LET 13=1LET 14=1LET 14=1LET 14=1LET 17=1LET 18=1LET 1*=1LET 110=1LET .71=1LET E2=1LET E3=1CREATE EVERY SLOTACREATE EVERY SLOTSLET CALL.AR7.RATE=TOTAL.SS*CALL.PER.HOUR:3600.FOR I=1 TO TOTAL. SSDOLET R=EXPONENTIAL.E3600.:CALL.RER.HOUR,1)IF R-(=STOR.TIMEADD 1 TO No.TO.TALKCREATE A CON -,:LET COFFsTART1=EXPONENTIAL.F(l.;CALL.AR.RATE,1)LET Cc.N_sTART=CON.sTART+CON.START1LET SOURcECCNV)=RANDI.FWN.STATICN,1)WHILE soURCECoNV)=M.STATIONDC,LET saJRCE=M=RANDI.E1,N.STATION,1)LOOPLET DESTINATION co ,NY=SoURCE(CONV)WHILE DESTINATION(CONV)=SOuRCE(CONV) ORDESTINATION; CONY)=M.STATIONDC ,LET DESTINATION(CONV)=RANDI.F1_,N.STATION,1)LOOPLET BEGIN(CONV)=COtiSTARTFILE CONY IN SU1ALWAYSLET R=RANDOM.E(1)IF R<=(CALL.PER_HouR.*CALL.HOLD.TIME/60.)LET R1=RANDOM.F(2)IF R1<RATIOLET MOBILE=0OTHERWISELET MOBILE=1ALWAYSLET SOUROE1=RANDI.F(1,N.STATION,1)WHILE SOURCE1=M.STATIONDOLET SOURcE1=RANDI.F(1,11.STATION,1)LOOPIF /40BILE=1. AND CHANNEL.UTSOURCE1).)—(300-Ho.CELLRES»OR (No.TALKINC+1>IS0C.CHAN AND MoBILE=0.) OR(NO.TALKING+1>ISCC.CHAN1 AND MOBILE=1.)IF MOBILE=1ADD 1 TO V.LOSTALWAYSADD 1 TO CALL.LOSTOTHERWISEIF MOBILE=1ADD 1 TO cHANNEL.UTsOURCE1)ADD 1 TO NOBILE.TOTAL.TALKEDALWAYSADD 1 TO TOTAL.TALKEDADD 1 TO NO.TALKINcALWAYSALWAySLOOPLET TRUE.TALK=INT.F(No.TALKINC*0.43LET BOUNDARY=INT.Fl(TRUE.TALK*64.*N.SLOTA;310000.48.:53.LET CALL.DPT.RATE=(NO.TALKINC);(CALL.HOLD.TI14E*60.FOR I=1 TO NO.TALKINCDoLET R3=EXpONENTIAL.F(CALL.HOLD.TIME*60.,2IF R3<=STOP.TIMEADD 1 TO NO.TO.DERARTCREATE A CONYLET cON.DEPART1=EXPONENTIAL.F1./CALL.DPT.RATE,2)LET cCN.DEPART=COI'I.DEPART+CON.DEPAP_TlLET SOURCE(OONV)=RANDI.F;1,N.STATION,2WHILE SOURCECONV)=M.STATIONDOLET ScuROE(O0IFI)=RA)IDI.E1,N.STATICN,2)LOOPLET DESTINATION(CONv)=sCNRCEOONV)OWHILE DEsTINATION(CONV)=SOURCECONV) ORDEsTINATIcMCOUV)=M.STATIONDOLET DESTINATION(CONV)=RANDI.FC1,N.STATIoN,2)LOOPLET TERMINATE(CONV)=CON.DEPARTFILE CON1 IN TERM.BUFFER1ALWAYSLOOPIF N.SU1>0REMOVE FIRST CON''.T FROM SU1LET SCHED=BECIN(CoNV)FILE CONY IN SU1ACTIVATE A SETUP1 IN SCHED .SECONDSALWAYSIF N.TERM.BUFFER1>0REMOTE FIRST CONV FROM TERM.BUFFER1LET SCHED1=TERMINATECCnV)FILE 00117 IN TERM.BUFFER1ACTIVATE A TERM.pRO1 IN SCHED]. .SECONDSALWAYSIF CELL9.PER.sEC>0ACTIVATE A GENERATOR° NOWALWAYSIF CELL1.PER.SEC:0ACTIVATE A GENERATOR1 NOWALWAYSIF CELL2.PER.SEC>0ACTIVATE A GENERATOR2 NOWALWAYSACTIVATE All UPDATE NOWFOP 3=1 TO DIVISIONCOACTIVATE A STOP.SIM Ill (STOP.TIME/DIVISION)*3 .SECONDS"ACTIVATE A VARRESET Ill 0 . SECONDSACTIVATE A VARRESET IN STOP.TIME DIVISION)*3 .SECONDSLOOP" ACTIVATE A STOP.SIM IN STOP.TIME .SECONDSSTART SIMULATIONclose unit 2END"The varrest process is used to reset the general parameters"to the original value. The purpose of this process is to"t,:pass the statistic values of unstable period in the system"and let the statistic result reflect the true value when"system go into stable state.PROCESS VARRESETRESET TOTALS OF DI.DELAYRESET TOTALS OF 02.DELAYRESET TOTALS OF DO.DELAYRESET TOTALS OF S.DELAYRESET TOTALS OF T.DELAYLET SENTO=0LET SENT1=0LET SENT2=0LET P.SENT=OLET CSP.CREATED=0LET P.CREATED=0LET CSP.COUNT=0LET CEND.COUNT=0LET CELLO.LOSS=0LET CELL1.LOSS=0LET CELL2.LOSS=OLET CELLO.CREATED=OLET CELL1.CREATED=0LET CELL2.CREATED=0LET CSP.LOST=OEND"Setup 1, 2 and 3 processes are used to implement the individual"Call setup, procedures within the system. Call setup procedures"are usually divided into several messages transmitted between"caller's node, heads of the bus, bandwidth manager, signalling"termination. The sequences are depend on the different call"setup schemes, the configuration of bandwidth manager, ST and"DB's position. In the mean time, it automatically collect the"statistic data such as )mean, max, min) call setup time, the"number of successful call setup procedures and the number of"failed call setup procedures, etc.PROCESS SETUP1DEFINE STATION AS AN INTEGER VARIABLEWHILE TIME.V<STOP.TIME AND INSUl>0DC,REMOVE THE FIRST CONY FROM SU1LET STATION=SOURCE(COITI)IF CHANNEL.UT(STATION)>=300-HO.CELLRESADD 1 TO 7.LOSTADD 1 TO CALL.LOSTOTHERWISELET R=RANDON. F 1IF R<RATIOADD 1 TO CHANNEL.UT(STATICI)ADD 1 TO MOBILE.TOTAL.TALKEDALWAYSLET X=Y1FOR I=1 TO XDOCREATE A CELLLET STARTC(CELL)=TIME.VLET BEGINC(CELL)=TIME.VLET SOURCECcCELLi=SCTRCECONV)LET DESTINATIONC(CELL)=DESTINATION(CONV)LET NEXT.DEsTINATIONC(CELL)=M.STATIONLET TyPEc10ELL)=1IF I=XLET MIDC;CELL =99OTHERWISELET MIDC(CELL;=IALWAYSIF soURCEC(cELL)<M.sTATIoNFILE CELL IN BUFFERA2(STATION)ADD 1 TO CSP.CREATEDADD 1 TO CELL2.CREATEDIF cDA2(STATIoN)=-2LET CDA2(STATION)=RCA2;STATION)LET RCA2(STATION)=0ADD 1 TO RQB2)STATION)ALWAYSOTHERWISEFILE CELL III BUFFERB2(STATION)ADD 1 TO CSP.CREATEDADD , 1 TO CELL2.CREATEDIF cDB2(STATION)=-2LET CDB2(STATION)=RCB2(STATION)LET RCB2(STATION)=0ADD 1 TO ROA2(STATION)ALWAYSALWAYSLOOPALWAISDESTROY THIS =NVIF N.SU1>0REMCE FIRST CON1 FROM SU1LET sCHED1=BEGIN(CCIP:)-TIME.VFILE Colf: IN sU1WAIT SCHED1 .SECONDSALWAysLOOPElIDPROCESS SETUP2DEFINE STATION AS INTEGER VARIABLESWHILE TIME.V<sT0P.TIME AND N.SU2>9DOREMOVE FIRST CELL FROM sU2LET STATION=SOURCEC(CELL)IF N0.TALKIN0+1›isoC.CHAN1LET^ 1 )IF R<RATIOSUBTRACT 1 FROM CHANNEL.UT)STATION)SUBTRACT 1 FROM •oBILE.TOTAL.TALKEDADD 1 TO ''.LOSTALWAYSADD 1 To CALL.LOSTDESTROY THIS CELLOTHERWISEIF BEGINC)CELL)<TIME.VLET STARTc(cELL)=STARTC;CELLTIME.V-BEGINC;CELL»ALWAYSADD 1 TO NO. TALKINGLET TRUE.TALK=INT.F(NO.TALKING*0.43)LET BouNDARy=INT.FHTRUE.TALK*64.*N.sLOTA/210090.):(48.153.)1ADD 1 TO TOTAL.TALKEDLET sT=STARTC(CELL)LET sG=STAGEC(CELL)LET SR=SOURCEC(CELL)LET D=DESTINATIONC(CELL)LET TYPEC(CELL)=1LET TC=TYpEC(CELL)LET N.D=N.STATIONLET STATION=M.STATIONDESTROY THIS CELLFOR I=1 TO Y2DOCREATE A CELLLET BEGINC)CELL)=TIME.VLET STARTC(CELL)=STLET DEsTINATIONC(CELL)=DLET TYPEC(cELL)=TCLET NEXT.DESTINATIONC(CELL)=N.DLET sOURCEc(cELL)=SRLET STAGEC(CELL)=SGIF I=Y2LET MIDC(CELL)=99OTHERWISELET MIDC;CELL)=IALWAYSFILE CELL IN BUFFERA2(STATION)ADD 1 TO CSP.CREATEDADD 1 TO CELL2.CREATEDIF CDA2(STATIoN)=-2LET CDA2(STATIOM)=RCA2;STATION;LET RCA2(STATION)=0ADD 1 TO ROB2(STATION)ALWAYSLOOPLET M.D=1FOP. 1=1 TO Y2DOCREATE A CELLLET DESTINATIONC;CELL;=DLET SOURCEC)CELL)=SRLET TYPEc(CELL)=TCLET BEGINC(CELL)=TIME.VLET STARTC(CELL)=STLET STAGECICELL)=SGLET NEXT.DESTINATIONC(CELL)=N.DIF I=Y2LET MIDC(CELL)=99OTHERWISELET MIDC(CELL)=IALWAYSFILE CELL IN BUFFERB2 ( STATION)ADD 1 TO CSP.CREATEDADD, 1 TO CELL2.CREATEDIF CDB2STATION)=-2LET CDB2STATION)=RCB2(STATION)LET RCB2;STATION)=0ADD 1 TO P.QA2 ( STATION)ALWAYSLOOPALWAYSLET X2=0.UNTIL X2=1.DOIF 11.502>CREMOVE FIRST CELL FROM SU2LET SCHED2=BEGINC(CELL)FILE CELL IN SU2IF SCHED2<TIME.VLET X2=1.OTHERWISEWAIT SLOT/5000000. .SECONDSALWAYSOTHERWISELET X2=1.ALWAYSLOOPLOOPENDPROCESS SETUP3WHILE TI•E.V<STOP.TIME AND N.503>0REI.17E FIRST CELL FROM 503IF BEGINC•CELLH<TIME.VLET STARTCCELL=STARTC(CELL)+TIME.V-BEGINO(CELL))ALWAYSLET STATION=NEXT.DESTINATIONC(CELL)LET ST=STAR=CELL))LET SR=SOURCECCELLLET D=DESTINATIONCL:ELL)LET B=BEGINCELLLET SG=STAGECCELL)IF STAGE" CELL )LET TC=2LET N.E,M.STATIONLET Y=Y3ALWAYSIF STAGEC(CELL)=4LET TC=2LET N.D=SOURCEC(CELL)LET Y=Y4ALWAYSIF STAGEC(CELL)=SLET TC=5LET N.D=M.STATIONLET Y=Y5ALWAYSIF STAGEC(CELL)=6LET TC=6LET N.D=SOURCEC(CELL)LET Y=Y6ALWAYSIF STACEC(CELL)=7LET TC=7LET N.D=M.STATIONLET Y=Y7ALWAYSIF STAGEC(CELL)=8LET TC=7LET N.D=DESTIN.ATIONC)CELL )LET Y=Y8ALWAYSIF STAGEC(CELL)=9LET TC=8LET N.D=M.STATIONLET Y=Y9ALWAYSIF STAGEC(CELL1=10LET TC=9LET N.D=DESTINATIONC\CELL)LET Y=Y10ALWAYSDESTROY THIS CELLFOR I=1 TO YDOCREATE A CELLLET TYPEC(CELL)=TCLET BEGINC(CELL)=TIME.VLET STARTC(CELL)=STLET SOURCECCCELL)=SRLET DESTINATIONC(CELL)=DLET NEXT.DESTINATIONC(CELL)=N.DLET STAGEC(CELL)=SCIF I=YLET MIDC(CELL)=99OTHERWISELET MIDCCELL)=IALWAYSIF STATICN<HEXT.RESTINATIONC(CELL)FILE CELL IN BUFFERA2(STATION)ADD 1 TO CSP.CREATEDADD 1 TO CELL2.CREATEDIF CDA2(STATIOH(=-2LET CDA2(STATION)=RCA2(STATIOH)LET RCA2(STATION)=0ADE , 1 TO RC ,B2(STATION)ALWAYSOTHERWISEFILE CELL Ill BUFFERB2(STATION)ADD 1 TO CSP.CREATEDADD, 1 TO CELL2.CREATEDIF CDB2(STATION)=-2LET CDB2 ( STATION) =RCB2 ( STATION)LET RCB2(STATIOH)=0ADD 1 TO RQA2 ( STATION)ALWAYSALWAYSLOOPLET X2=0.UNTIL X3=1.IF H.SU2>0REMOVE FIRST CELL FROM SU3LET SCHED3=BEGINC(CELL)FILE CELL IN SU3IF SCHED3<TIME.VLET X3=1.OTHERWISEWAIT SLOT:5000000. .SECONDSALWAYSOTHERWISELET X3=1.ALWAYSLOOPWHILE TINE.V<STOP.TIME AMR INTERM.BUFFER1>0DOREMOVE THE FIRST COLIV FROM TERM.BUFFER1LET STATION=DESTINATICN(CCNV)FOR I=1 TO Z1DOCREATE A CELLLET STARTC(CELL)=TIME.VLET BEGIHC(CELL)=TIME.VLET SOURCEC(SELL)=SOURCECONV)LET DESTINATIONC(CELL)=DESTINATION(CONV)LET NEXT.DESTINATIONC(CELL)=SOURCE(CONV)LET TERMC(CELL)=TERMINATE(COMV)LET TYPEC(CELL)=21IF I=Z1LET MIDC(CELL)=99OTHERWISELET MIDC(CELL)=IALWAYSIF STATION<NEXT.DESTINATIONC(CELL)FILE CELL IN BUFFERA2 (STATION)ADD 1 TO CTM.CREATEDIF CDA2(STATION)=-2LET CDA2(STATION)=RCA2(STATION)LET RCA2(STATION)=0ADD 1 TO RQB2 (STATION)ALWAYSOTHERWISEFILE CELL IN BUFFERB2 STATION(ADD 1 TO CTM. CREATEDIF CDB2(STATION)=-2LET CDB2(STATION)=RCB2(STATION(LET RCE2(STATICAN=0ADD 1 TO RQA2(STATIOH)ALWAYSALWAYSLOOPDESTROY THIS CONVIF IRTERM.BUFFER1>0REMOVE FIRST CONV FROM TERM.BUFFER1LET SCHEE1=TERMIHATE(CONV)-TIME.VFILE CONY IN TERM.BUFFER1WAIT SCHED1 .SECONDSALWAYSLOOPEIID"TERM.PRO1 and TERM.PRO2 processes are used to implement the"(iallterrninating procedure. When the call has been hanged off,"these processes will be activated. As call setup , procedure ,"it collects the useful data such as call terminating delay,"etc.PROCESS TERM. PRO1DEFINE STATION AS AN INTEGER VARIABLEENDPROCESS TERM.PRO2WHILE TIME.V<STOP.TIME AND H.TERM.BCFEER2>0DC,REMOVE FIRST CELL FROM TERM.BUFFER2LET STATION=HEXT.DESTINATIOHC(CELL)LET ST=STARTC(CELL)LET SR=SOURCEC(CELLLET N.D=SOURCEC(CELL)LET D=DESTINATIONC(CELL)LET SG=STAGEC(CELL)LET^ CELL )IF STAGEC.!.CELL=22LET TC=22LET N.D=DESTINATIONC(CELL)LET -,_]=Z2ALWAYSIF STAGEC(CELL)=23LET TC=23LET N.D=SCNRCEC;CELL!LET Y-1'33ALWAYSDESTROY THIS CELLFOP. I=1 TODC,CREATE A CELLLET TYPECLOELL=TCLET BECINCCELL)=TIME.VLET STARTC(CELL)=STLET SOURCECCCELL=SRLET DESTINATIONCCELL)=DLET NEXT.DESTINATIONC(CELL)= ,MDLET TYPECCELL)=TCLET STAGEO(CELL=S0LET TERMCCELL)=TMIF I=2LET MIDCHOELL =99OTHERWISELET MIDC(CELL)=IALWAYSIF STATION<NEXT.DESTINATIONC(CELL)FILE CELL III BUFFERA2(STATION)ADD 1 TO CTM.CREATEDIF CDA2(STATICN)=-2LET CDA2STATION)=RCA2(STATION)LET RCA2o5'TATION)=0ADD 1 TO RQB2STATION)ALWAYSOTHERWISEFILE CELL III BUFFERB2STATION!ADD 1 TO CTM.CREATEDIF CDB2(STATION)=-2LET CDB2STATION) =RCB2(STATION)LET RCB2STATICIM=0ADD 1 TO RQ02(STATION)ALWAYSALWAYSLOOPLET K3=0.MTIL X3=1.DGIF N.TERM.BUFFER2>0REMOVE FIRST CELL FROM TERM.BUFFER2LET SOHED3=BEGINC(CELL)FILE CELL IN TERM.BUFFER2IF SCHED3<TIME.VLET X3=1.OTHERWISEWAIT SLOT, 5000000. .SECONDSALWAYSOTHERWISELET X3=1.ALWAYSLOOPLOOPEND"The GENERATORO, 1 and 2 processes control the generation of"user cells in each priority level of MAN. It takes the"parameter of traffic load of each level. For each message,"it contains the different number of cells determined by a"random generator. The source and destination addresses foreach message are also generated by a random generator. So,"the cell.gens can know where to put this message and"determine the bus that the message will be transmitted. The"message inter-arrival time are determined using exponential"distribute generator. As we know, each node hasthree buffers"in each bus direction corresponding to the priority level."The buffer size is pre-determined at the beginning of the"simulation program execution. When a message arrived, it"first to determine number of cells it need to transmit, If"the cells already in that buffer plus the number of cells"just arrived beyond the buffer size. This message is said to"be lost. Otherwise, it puts into this buffer. When this"operation finished. The cell.gen will wait for the next"message arrive.PROCESS GENERATORODEFINE S AND D AND K AND COUNT AS INTEGER. VARIABLESWHILE TIME.V<STOP.TIMEDC,LET I=RANDI.F(1,3,3)LET S=RANDI .E^STATION, IWHILE S=M.STATIONDOLET S=RANDI.E(1,N.STATION,I)LOOPLET D=SWHILE S=DDOLET D=RANDI.F(1,N.STATION,I)LOOPLET X=RANDI.F,1,2,I )FOR COUNT=1 TO XDOCREATE A CELLADD 1 TO CELLO.CREATEDLET SOURCEC(CELL)=SLET DESTINATIONC(CELL)=DLET NEXT.DESTINATIONCCELL)=DLET STARTC(CELL)=TIME.VIF S<DIF N.BUFFERAOS)<BUFFER.CAPACITYFILE CELL IN BUFFERAO(S)IF CDAO,S)=-2LET CDAO;S1=RCAO(S)LET RCAO(S)=0ADD 1 TO RQBO(S)ALWAYSOTHERWISEADE , 1 TO CELLO.LOSSDESTROY THIS CELLALWAYSOTHERWISEIF H.BUFFERBO(S)<BUFFER.CAPACITYFILE CELL III BUFFERBO(S)IF CDBO(S)=-2LET CDBO(S)=RCBO(S)LET RCBOS=0ADD 1 TO RQAO!S)ALWAYSOTHERWISEADD 1 TO CELLO.LOSSDESTROY THIS CELLALWAYSALWAYSLOOPWAIT EXPONENTIAL.F;CELLO.INTERARRIVAL.TIME*X,1) .SECONDSLOOPEl IDPROCESS GENERATOR1DEFINE S AND D AND X AND COUNT AS INTEGER. VARIABLESWHILE TIIIE.V<STOP.TII1EDOLET I=RANDI.F1,3,2LET S=RANDI.F(1,11.STATION,I)WHILE S=N.STATIONDOLET S=RANDI.F1,N.STATION,I)LOOPLET D=SWHILE S=DLET D=RANDI.F(1,H.STATION,DLOOPLET X=RANDI .F ( 1 , 2 , I )FOR COUNT=1 TO XDOCREATE A CELLADD 1 TO CELL1.CREATEDLET SOURCEC(CELL)=SLET DESTINATIONC(CELL)=DLET NEXT.DESTINATIONCCELL=DLET STARTC(CELL)=TINE-VIF S<DIF N.BUFFERA1(S)<BUFFER.CAPACITYFILE CELL IN BUFFERA1;SJIF CDA1(S)=-2LET CDA1S)=RCA1S)LET RCA1(S)=0ADD 1 TO RQB1(S)ALWAYSOTHERWISEADD 1 TO CELL1.LOSSDESTROY THIS CELLALWAYSOTHERWISEIF N.BUFFERB1;S)<BUFFER.CAPACITYFILE CELL IN BUFFERB1 ) S)IF CDB1(S)=-2LET CDB1S1=RCB1S)LET RCB1(S)=0ADD 1 TO RQA1(S)ALWAYSOTHERWISEADD 1 TO CELL1.LOSSDESTROY THIS CELLALWAYSALWAYSLOOPWAIT EXPONENTIAL.F,CELL1.INTERARRIVAL.TIME*X,U SECONDSLOOPENDPROCESS GENERATOR2DEFINE S AND D AND F AND COUNT AS INTEGER VARIABLESWHILE TIME.V<STOP.TIMEDOLET I =RAI 'DI F ( 1 , 3 , 1 1LET S=RANDL.F(1,I 'I.STATION,I )WHILE S =14 .^211DOLET S=RANDI.F(1,P.STATION,I)LOOPLET D=SWHILE S=DDOLET D=RANDI.F)1,N.STATION,I)LOOPLET X=RANDI.F(1,2,I)FOR COUNT=1 TO XDOCREATE A CELLADD 1 TO CELL2.CREATEDLET SOURCECCCELL)=SLET DESTINATIONC)CELL)=DLET NEXT.DESTINATIONC(CELL,=DLET STARTC)CELL)=TIME.VLET BEGINC(CELL)=TIME.VIF S<DIF H.BUFFERA2)S)<BUFFER.CAPACITYFILE CELL IN BUFFERA2(S)IF ODA2(S)=-2LET CDA2)S)=RCA2(S)LET RCA2S)=0ADD 1 TO RQB2(S)ALWAYSOTHERWISEADD 1 TO CELL2.LOSSDESTROY THIS CELLALWAYSOTHERWISEIF N.BUFFERB2)S)<BNFFER.CAPACITYFILE CELL IN BUFFERB2)S)IF CDB2(S)=-2LET CDB2(S)=RCB2(S)LET RCB2)S)=0ADD 1 TO RQA2(S)ALWAYSOTHERWISEADD 1 TO CELL2.LOSSDESTROY THIS CELLALWAYSALWAYSLOOPWAIT EXPOIJENTIAL.E)CELL2.INTERARRIVAL.TIME*X,1) .SECONDSLOOPEND"This process is used to change the status of each node in the"network in a slot time period. It realizes the function of"D.:TB protocol. When a slot passing by each node, it will do"fcilowing things: check this slot is empty or not; change the"cell request counter, count-down counter and request queuecounter; decide to implement the BWB mechanism; decide to do"erasure node function; collect general statistic data; change"the status of call setup, call handoff and call terminate"procedures.PROCESS UPDATEDEFINE SLOT; AND SLOTB AND STATION AS INTEGER RIABLESWHILE TIME.V<STOF.TIMEDOLET BOUNDARY.A=0LET BOUNDARY.B=0FOR COUNT=1 TO N.SLOTADOFOR SLOTA=1 TO (N.SLOTA-1)DOLET REQBITA2(N.SLOTA-SLOTA+1)=REQBITA2(N.SLOTA-SLOTA)LET REQBITA1(N.SLOTA-SLOTA+1)=REQBITA1)N.SLOTA-SLOTA)LET REQBITAO(N.SLOTA-SLOTA+1)=REQBITAO(N.SLOTA-SLOTA)LET BUSYA(N.SLOTA-SLOTA+1)=BUSYA(N.SLOTA-SLOTA)LET PSRA(N.SLOTA-SLOTA+1)=PSRA(N.SLOTA-SLOTA)LET DSTA(N.SLOTA-SLOTA+1)=DSTAW.SLOTA-SLOTA)LET SLOT.TYPEA(11.SLOTA-SLOTA+1)=SLOT.TYPEA(H.SLOTA-SLOTA)LOOPLET REQBITA2(1)=0LET REQBITA1(1)=0LET REQBITAC)(1)=0LET BUsYA(1)=0LET DSTA)1)=0LET PSRA(1)=0LET SLOT.TYPEA(1)=3IF BOUNDARY.A<BOUNDARYLET SLOT.TYPEA(1)=1LET BUSYA(1)=1LET TEMP1=RANDI.F(1,N.STATION,3)LET TEMP2=RANDI.F(1,N.STATION,1)IF TEMPI > TEMP2LET DSTA(1)=TEMP1OTHERWISELET DSTA(1)=TEMP2ALWAYSADD 1 TO BOUNDARY.AADD 1 TO P.SENTALWAYSFOR sLoTB=2 TO N.SLOTBDOLET REQBITB2(SLOTB-1)=REOBITB2sL•TB)LET REQBITB1SLoTB-1)=REQBITB1(SLOTB)LET REQBITBO(SLOTB-1)=REQBITBO(sLoTB)LET BUSYBsLOTB-1)=BUSYBSLOTB)LET PSRB(SLOTB-1)=PSRB(SLOTB)LET DSTB(SLOTB-1)=DSTB)SLOTB)LET SLOT.TYPEB)SLOTB-1)=SLOT.TYPEB)SLOTB)LOOPLET REQBITB2(N.SLOTB)=0LET REQBITB1(N.SLOTB)=0LET REQBITBO(N.SLOTB)=0LET BUSYB(N.SLOTB)=0LET PsRB(N.SLOTB)=0LET DSTB(H.sLOTB)=0LET SLoT.TYpER(11.SLOTB)=0IF BOUNDARy.B<BOUNDARYLET SL0T.TYPEB(N.SLOTB)=1LET BUSYB(N.SLOTB)=1LET TEMP1=RANDI.F(1,N.STATION,1)LET TEMP2=RANDI.F(1,M.sTATION,2)IF TEMPI TEMP2LET DSTB(N.SLOTB)=TEMplOTHERWISELET ESTB(M.sLOTB)=TEMP2ALWAYSADE , 1 TO BOUNDARY.BADD 1 TO P. SEATALWAYSFOR STATION=1 TO I1. STATIONDOLET SLoTA=SD*STATION-sD+1LET SL0TB=SD*sTATION-SD+1IF REQBITA2(SLOTA)=1ADD 1 TO RCB2(STATION)IF CDB1(STATION)=-2ADD 1 TO RCB1(STATION)OTHERWISEADE , 1 TO CDB1(STATION)ALWAYSIF CDBO(STATION(=-2ADD 1 TO RCBO(STATION)OTHERWISEADD 1 TO CDBO(sTATION)ALWAYSIF REQBITA2(SLOTA)=1 AND ERCTB2(STATION)>0LET REQEITA2SLCTA)=0SUBTRACT 1 FROM ERCTB2(STATION)ALWAYS.ALWAYSIF REQBITAl(SLOTA)=1ADE , 1 TO RCB1(STATICN)IF CDK(STATION)=-2ADD 1 TO RCBO(STATION)OTHERWISEADD 1 TO CDB0)STATION)ALWAYSIF REQBITAl(SLOTA)=1 AND ERCTB1(STATION)>0LET REQBITAI(SLOTA)=0SUBTRACT 1 FROM ERCTB1(STATION)ALWAYSALWAYSIF REQBITAO(SLOTA)=1ADD 1 TO RCBO(STATION)IF REQBITAG(sLOTA)=1 AND ERCTBO(STATION)>0LET REQBITAOSEOTA)=0SUBTRACT 1 FROM ERCTBO(STATION)ALWAYSALWAYSIF REQBITA2(SLOTA)=0 AND R2A2(STATION)>0 ANDFRCTB2)sTATION)> , 0LET REOBITA2(SLOTA)=0SUBTRACT 1 FROM RQA2(STATION)SUBTRACT 1 FROm ERCTB2(STATION)OTHERWISEIF REQBITA2(SLOTA)=0 AND RQA2(STATION)>0 ANDERCTE2STATION)=0LET REQBITA2(SLOTA)=1SUBTRACT 1 FROM RQA2(STATIOIEIF CDB1(STATION)=-2ADD 1 TO RCB1)STATION)OTHERWISEADD 1 TO CDB1)STATION)ALWAYSIF CDBO)STATION)=-2ADD 1 TO RCBO)STATIC1 (1)OTHERWISEADD 1 TO CDBO(STATION)ALWAYSALWAYSALWAYSIF REQBITAl(SLOTA)=0 AND F(QA1STATION)>0 ANDERcTB1(STATION))(0LET REOBITAl(SLOTA)=0SUBTRACT 1 FROM RQA1(STATION)SUBTRACT 1 FROM ERCTB1(sTATion)OTHERWISEIF REQBITA1SLOTA)=0 AND P.QA1)STATION)>0 ANDER.CTB1)STATION(=0LET REQBITA1(sLoA)=1SUBTRACT 1 FROM R.:A1)STATION(IF CDBO(STATION)=-2ADC , 1 TO RCBO)STATION)OTHERWISEADD 1 TO cDBO(STATION)ALWAYSALWAYSALWAYSIF REQBITAO(SLOTA)=0 AND RC,A0(STATION)>0 ANDERCTBO(STATION)>3LET REQBITAC(sLOTA)=0SUBTRACT 1 FROM RQA0(sTATION)SUBTRACT 1 FROM ERCTBO(STATION)OTHERWISEIF REOITAO(SLOTA)=0 AND R2A0(STATION)>0 ANDERCTBO(STATION)=000LET REC , BITAO(SLOTA) , 1SUBTRACT 1 FROM RQAO(STATION)ALWAYSALWAYSIF REQBITB2)SLOTB)=1ADD 1 TO RCA2(sTATION)IF CCA1(sTATION)=-2ADD 1 TO RcAl(STATION)OTHERWISEADD 1 TO CDA1(STATION)ALWAYSIF CDAO(STATION=-2ADD 1 TO RCAO(STATION)OTHERWISEADD 1 TO CDAO(STATION)ALWAYSIF REQBITB2)SLOTB)=1 AND ERCTA2(STATION)>0LET REQBITB2)sLOTB)=0SUBTRACT 1 FROM ERCTA2(STATION)ALWAYSALWAYSIF REcoiTB1 sLOTB)=1ADD 1 TO RCA1)STATION)IF cLAO(STATION)=-2ADD 1 TO RCAC)STATION)OTHERWISEADD 1 TO CDAC(STATION)ALWAYSIF REQBITB1)SLOTB)=1 AND ERCTAl(STATION>0LET REQBITB1(SLCTB)=0SUBTRACT 1 FROM ERCTAl(STATION)ALWAYSALWAYSIF REOBITBI sLOTB)=1ADD 1 TO BCAO(STATIONIF REC.BITBO)SLOTB)=1 AND ERCTAO(STATION>0LET REC,BITBO(SLCTB)=0SUBTRACT 1 FROM ERCTRO)STATION)ALWAYSALWAYSSUBTRACT 1 FROM RQB2STATION)IF CDA1(STATICN)=-2ADD 1 TO RCAUSTATION)OTHERWISEADD 1 TO CCAl(STATICM)ALWAYSIF cDA0(sTATIoN)=-2ADD 1 TO RcAO(STATIoN)OTHERWISEADD 1 TO CDAOsTATION)ALWAYSALWAYSALWAYSIF REQBITB1(SLCFB)=0 AND RQB1)STATION)>0 ANDERCTAUSTATION)>0LET REQBITB1(SLOTB)=0SUBTRACT 1 FROM RQB1(STATION)SUBTRACT 1 FROM ERCTA1(STATION)OTHERWISEIF REQBITB1(SLOTB)=0 AND RQB1)STATION)>0 ANDERCTA1(STATION)=0LET REQBITB1)SLOTB)=1SUBTRACT 1 FROM RQB1(STATICN)IF CDAO(STATION)=-2ADD 1 TO RCAO(STATION )OTHERWISEADD 1 TO CDAO(STATION)ALWAYSALWAySALWAYSIF REQBITBO(SLOTB)=0 AND RUO)STATICN»0 ANDERCTAO(STATION)>0LET REQBITBO(SLOTB)=0SUBTRACT 1 FROM RUG(STATION)SUBTRACT 1 FROM ERCTA0(sTATION)OTHERWISEIF REQBITBO(SLCTB)=O AND P. ,132(STATICII)>O ANDERCTAO(STATION)=0LET REQBITBO)SLOTB)=1SUBTRACT 1 FROM ROBO(STATION)ALWAYSALWAYSIF REQBITB2)SLOTB) , 0 AND RQB2(STATION)>0 ANDERCTA2(sTATION)>0LET REQBITB2(SLOTB)=0SUBTRACT 1 FROM RQB2(STATION)SUBTRACT 1 FROM ERCTA2(STATION)OTHERWISEIF REQBITB2(sLOTB)=0 AND RO ,B2(STATION)>0 ANDERCTA2)STATIOH)=0LET REQBITB2 • SLOTB)=1IF BUSYA)SLOTA)=1 AND DsTA SLOTA)=STATIOM AND SWITCH=1LET PSRA(SLOTA)=1ALWAYSIF BUSYB)SLOTB)=1 AND DSTB)SLOTB)=STATION AHD SWITCH=1LET PSRB)SLOTB)=1ALWAYSIF (SWITCH=1 AND NODEA=G AND STATION>1 ANDSTATICM<II.STATION)OR SWITCH=1 AND STATION=NODEA)IF RCA2(sTATIoN)>0 OR CDA2(STATION)>=0 ORRQB2(STATIoN)>0LET P=2OTHERWISEIF PcA1(sTATION)>0 OR ODA1(STATION)>=0 OR.RQB1(STATION)>0LET P=1OTHERWISEIF RCAO(STATION)>0 OP. CDAOSTATION)>=0 ORRQB0(sTATION)>0LET P=0OTHERWISELET P=-1ALWAYSALWAYSALWAYSIF BUsYA(sLoTA)=1 AND PSRA(SLOTA)=1LET BUSYA(SLOTA)=0LET PSRA(SLOTA)=0IF p=2ADD 1 TO ERCTA2(STATION)OTHERWISEIF P=1ADD 1 TO EROTAl(STATION)OTHERWISEIF P=0ADD 1 TO ERCTAO(STATION)ALWAYSALWAYSALWAYSALWAYSALWAYSIF (SWITCH=1 AND NODEB=0 AND STATION>1 ANDSTATION<N.STATIOM)GP. ,.SWITCH=1 AND STATICN=NoDEBIF RCB2!.STATION(>0 OR CDB2(STATION)>=0 ORR ,QA2(sTATION)>0LET Q=2OTHERWISEIF RCB1(STATICN)>0 CR CDB1 STATION)>=0 OP.RQA1(STATICN)>0LET ', )-(=1OTHERWISEIF ROBO(STATION)^OR CDEO(STATION)>=0 CRRQAO(STATIOM)>0LET ,Q,0OTHERWISELET Q=-1ALWAYSALWAYSALWAYSIF BUsyB(SLoTB)=1 AND PSRB(SLOTB)=1LET BUSYB(SLOTB)=0LET PSRB(SLOTB)=0IF Q=2ADD 1 TO ERCTB2(STATION)OTHERWISEIF Q=1ADD 1 TO ERCTB1(STATIoN)OTHERWISEIF Q=0ADD 1 TO ERCTBO(STATION)ALWAYSALWAYSALWAYSALWAYSALWAYSIF BUsYA(SLOTA)=0IF BUsYA(SLOTA)=0 AND RoA2(STATION)>0AND CDA2(STATION)=-2SUBTRACT 1 FROM RcA2(STATION)ALWAYSIF BUsIA(SLCIA)=0 AND RcA1(STATION)>0AND CDA1(STATION)=-2SUBTRACT 1 FROM ROAUSTATION)ALWAYSIF BUSYA(SLOTA)=0 AND RCA0(STATION)>0AND CDAO(STATION)=-2SUBTRACT 1 FROM RCAO(STATION)ALWAYSLET K=0.IF BUsYA(SLOTA)=0 AND CDA2(STATION)>-2SUBTRACT 1 FROM CDA2 (STATION)IF CDA2(STATICN)=-1LET BUSYA(SLOTA)=1REMOVE FIRST CELL FROM BUFFERA2(STATION)LET DETA(SLOTA(=NEHT.DESTINATICNO(CELL)LET D2.DELAY=ABs.F(STATION-NEXT.DESTINATIONC(OELL))*SD*sLCT/1000000.+TIME.V-BECINC(CELL)ADD 1 TO SENT2IF TYPEC(CELL; HE 0IF NEXT.DESTINATIONC(CELL)=M.STATION ANDTYpEC(CELL)=1AND MIDO(CELL)=99LET El.DLY=Y1*ABS.F(M.STATION-STATION)*SD*SL0T/100000 0 +TIME.V-BEGINC(CELL)LET D=Y1*ABS.F(M.STATI0H-STATI011)*sEJ(sLOT:1000000.+0.0000013LET BEOINC(CELL)=D , +TIME.VOIF N. SU2=0FILE CELL IN SU2ACTIVATE A SETUP2 IN D .SECONDSOTHERWISEFILE CELL IN SU2ALWAYSALWAYSIF NEXT.DESTINATIONCCELL)=N.STATION ANDTYPECCELL)=1AND NIDC)CELL)=99LET E2.ELY=Y2 *ABS.F)N.STATION-STATION)*SD*SLOT/1000000.+TIME.V-BECINC(CELL)LET D=12*j1.STATION-STATION)*SD*SLCT:1000000.LET BECINC(2ELL)=D+TIME.VIF N.SU3=0ACTIVATE A SETUP3 IN D .SECONDSALWAYSLET STAGEC(CELL)=3FILE CELL IN SU3ALWAYSIF NEXT.DESTINATIONC)CELL)=M.STATION ANDTYPEC)CELL)=2AND MIDC(CELL)=99LET E3.DLY=Y3*ABS.E(STATION-M.STATION)*SD*SLOT/1000000.+TINE.1-BEGINC)CELL)LET F=1.ALWAYSIF NEXT.DESTINATICNC)CELL)=SOURCECCELL) ANDTYPEC)CELL)=2AND NIDCCELL)=99LET E4.DLY=Y4*ABS.F(NEXT.DESTINATIONC(CELL)-STATIOW*SD* SLOT, 1003000.+TIME.V-BEOINC)CELL)LET D=Y4*ABS.FTIEXT.DESTINATICVCCCELL STATION)*SD*SLOT, 1000000.LET BEGINCH2ELL)=D+TIME.VIF M.SU3=0ACTIVATE A SETUP3 IN D .SECONDSALWAYSLET STAGEC(CELL)=5FILE CELL IN SU3ALWAYSIF NEXT.DESTINATIONCCELL)=M.STATION ANDTYPEC)CELL)=5AND MIDC(UELL)=99LET E5.DLY=Y5*ABS.F)NEXT.DESTINATICNC;CELL -STATICNYSD*SLOT:100000C.+TINE.V-BEGIN):.)CELL)LET E,Y5*ABS.F(NEXT.DESTINATIONC(CELL)-STATION)*SD*SLOT/1000000.LET BEGINC)CELL)=D+TIME.VIF N.SU3=0ACTIVATE A SETUP3 IN D .SECONDSALWAYSLET STAGEC(CELL)=6FILE CELL IN SU2ALWAYSIF NEXT.DESTINATIONC(CELL)=SCURCEC)CELL) ANDTYPEC(CELL)=6AND MIDC(CELL)=99LET E6.DLY=Y6*ABS.F(NEXT.DESTINATIONC(CELL)-STATION)*SD*SLOT/1000000.+TIME.V-BECIMC)CELL)LET D=Y6*ARS.F(NEXT.DESTINATIONC(CELL)-STATION)*SD*SLOT/1000000.+0.0000003+UNIFORM_F(0.,AREA,1)LET BEGINC(CELL)=D+TIME.VIF N.SU3=0ACTIVATE A SETUP3 IN D .SECONDSALWAYSLET STAGEC(CELL)=7FILE CELL IN SU3ALWAYSIF NEXT.DESTINATIONC(CELL)=M.STATION ANDTYPEC)CELL)=7AND MIDC)CELL)=99LET E7.DLY=Y7*ABS.F(NEXT.DESTINATIONC(CELL-STATION)*SD*SLOT/1000000.+TIME.V-BEGINC)CELL)LET D=Yi*ABS.FNEXT.DESTINATIONC)CELL)-STATION)*SD*SLOT/1000000.+0.0000003LET BEGINC(CELL)=D+TIME.VIF N.SU3=0ACTIVATE A SETUP3 IN D .SECONDSALWAYSLET STACECCELL)=8FILE CELL IN SU3ALWAYSIF NEXT.DESTINATIONC(2ELL)=DESTINATIONC(CELL) AHDTYPEC(CELL)=7AND MIDCCELL)=99LET E8.DLY=Y8*ABS.F(NEXT.DESTINATIONC)CELL)-STATION)*SD*SLOT;1000000.+TIME.V-BEGINC(CELL)LET D=Y8 *ABS.F(NEXT.DESTINATIONC)CELL)-STATION)*SD*SLOT:1000000.LET BEGINC)CELL)=D+TIME.VIF N.SU3=0ACTIVATE A SETUP3 IN D .SECONDSALWAYSLET STAGECCELL)=9FILE CELL IN SU3ALWAYSIF NEXT.DESTINATIONC(CELL)=M.STATION ANDTYPECCELL)=8AND MIEOCELL)=99LET ES.DLY=Ye*ABS.FMEXT.DESTINATIONC(CELL)—STATION)*SD*SLOT:100000(3.+TIME.V—BEINCOO,ELL)LET D=i9*ABS.F(NEXT.DESTINATIONC(CELL)—STATION)*SD*SLOT/1000000.LET BECINC(CELL)=D+TIME.VIF M.SLI3=0ACTIVATE A SETUP3 III D _SECONDSALWAYSLET STAGECCELL)=10FILE CELL IN SU3ALWAYSIF NEKT.DESTINATIONC(CELL)=DESTINATICNC(CELL)AND TYPECCELLj=;AND NIECCELL)=9;LET E1C.DLY=110*ABS.FWEXT.DESTINATIONC(CELL)—STATION)*SD * SLOT:1000000.+TIME.V—BEOINCCELLLET D=110*ABS.F(M.STATION—DESTINATIONC(CELL))*SD*SLCT:1000000.+0.0000003LET S.DELAY=TIME.V—STARTCiCELL)+DADD 1 TO CSP.COUNTLET 8=1.ALWAYSIF NEXT.DESTINATIONC;CELL=SOURCECCELL) ANDTYPEC.!CELL)=21 AND MIDC(CELL)=99LET D=21*ABS.ESOUPCEC(CELLH-DESTINATIONC(CELL))*SD*SLOT;1331:000.LET BEINC(CELL=TINE.V+DIF IITERM.BUFFER2=0ACTIVATE A TERM.PRO2 IN D .SECONDSALWAYSLET STAGECOELL.=22FILE CELL IN TERM.BUFFER2ALWAYSIF NEXT.DESTINATIONC;CELL =DESTINATIONC ELLAND TYPECCELL=22AHD MIDCDCELL)=99LET D=22*ABS.FSCJJRCEC;CELL)—DESTINATIONCCELL)*SD*SLCT:10C)I0(330.LET BE=C(CELL)=-TIME.V+DIF N.TERM.BUFFER2=0ACTIVATE A TERN.PRO2 IN D .SECONDSALWAYSLET STAGECOCELLJ=23FILE CELL IN TERM.BUFFER2ALWAYSIF NEXT.DESTINATIONC(CELL)=SOURCECCELLAND TYPEC(CELL) , 23 AND MIDCDOELL=99LET D=23*ABS.F(SCURCECCELL)—DESTINATIONC(CELL)*SD*SLOT/1000000.LET T.DELAY=TINE.V—TERMCCCELL+DLET STN=DESTINATIONC(CELL)IF REAL.F(STN);3. NEREAL.F(INT.FREAL.FSTN)/3.))SUBTRACT 1 FROM CHANNEL.UT(STH)ALWAYSSUBTRACT 1 FROM NO.TALKINGLET TRUE.TALK=INT.F(PI0LTALKING*0.43)LET BOUNDARY=INT.Fi(TRUE.TALK*64.*N.SLOTA/310000.):18./53.))ADD 1 TO CEND.COUNTLET 8=1ALWAYSOTHERWISELET K=1ALWAYSIF 8=1.DESTROY THIS CELLALWAYSIF N.BUFFERA2(STATION)>0LET CDA2(STATION) ,RCA2(STATION)LET RCA2(STATICN)=0ADD 1 TO RQ132STATION)OTHERWISELET CDA2(STATION)=-2ALWAYSIF BWB.MOD>0IF BWB.CNTR(STATION)=BWB.MOD-1LET BWB.CHTRASTATION)=0IF CDA2(STATION)=-2ADD 1 TO RCA2(STATION)OTHERWISEADD 1 TO CDA2cSTATION)ALWAYSOTHERWISEADD 1 TO BWB.CNTRA I, STATION;ALWAYSALWAYSALWAYSALWAYSIF BUSYA(SLOTA)=0 AND CDA1;STATION)>-2SUBTRACT 1 FROM CDA1 ) STATION)IF CDA1STATION)=-1LET BUSTASLOTA=1REMOVE FIRST CELL FROM BUFFERA1STATION)LET DSTA(SLOTA)=NEET.DESTINATIONC ; CELL )LET D1.DELAY=ABS.FCDESTINATIONCCELLI—STATION!*SD*SLOV1000000.+TIME.V—STABTC(CELL)DESTROY THIS CELLADD 1 TO SENT1IF N.BUFFERA1STATION)>0LET CDAUSTATION) =RCAUSTATION)LET RCAl(STATION)=0ADD 1 TO RQB1(STATION)OTHERWISELET CDA1(STATION)=-2ALWAYSIF BWB.MOD>0IF BWB.cNTRASTATION)=BWB.MOD-1LET BWB.CNTRA(STATIoN)=0IF CDA1(sTATION)=-2ADD 1 To RCA1STATION)OTHERWISEADD 1 TO CDA1(STATION)ALWAYSOTHERWISEADD 1 TO BWB.CNTRA(STATION)ALWAYSALWAYSALWAYSALWAYSIF BUSyA(sLoTA)=0 AND cDAO(STATION)>-2SUBTRACT 1 FROM CDAO(STATION)IF cDAOSTATICN)=-1LET BUSYA(SLOTA)=1REMOVE FIRST CELL FROM BUFFERAO(STATION)LET DSTA(SLOTA)=NEXT.DESTINATIONC(CELL)LET DO.DELAY=ABS.F;DESTINATIONC(CELL)-STATION)*SDIsLOT:1000000.+TIME.V-STARTC(CELL)DESTROY THIS CELLADD 1 TO SENTOIF 11.BUFFERAOSTATION)>0LET CDA0(STATION)=RCAO(STATION)LET ECA0STATIoN)=0ADD 1 TO RQBOSTATION)OTHERWISELET CEAODSTATION)=-2ALWAYSIF BWB.MOD>0IF BWB.CNTRA(STATION)=BWB.NOD-1LET BWR.cNTRASTATION)=0IF CDAO(STATION)=-2ADD 1 TO RCAO+STATIOIDOTHERWISEADD 1 To CDAOsTATION)ALWAYSOTHERWISEADD 1 TO BWB.CNTRASTATIONALWAYSALWAYSALWAYSALWAYSALWAYSIF BUSYB(SLOTB)=0IF BUsIB(SLOTB)=0 AND RCB2(sTATION)>0AND CDB2(STATIOH)=-2SUBTRACT 1 FROM RCB2(sTATION)ALWAYSIF BUSYB(SLOTB)=0 AND RCB1(sTATION)>0AND CDB1(STATION)=-2SUBTRACT 1 FROM RCB1(STATION)ALWAYSIF BUSYB(SLOTB)=0 AND RCEO(STATION)>0AND CDBO(STATIoN)=-2SUBTRACT 1 FROM RCBO(STATION)ALWAYSLET K=0.IF BUSYB(SLOTB)=0 AND CDB2(STATION)>-2SUBTRACT 1 FROM CDB2(STATION)IF CDB2(sTATION)=-1LET BUSYB(SLOTB)=1REMOVE FIRST CELL FROM BUFFERB2(STATION)LET DSTB(SLOTB)=NEXT.DESTINATIONCCELL)LET 02.DELAY=ABS.E(STATION-HEXT.DESTINATIONC(CELL))*SD*SLOT/1000000.+TIME.V- BEGINC(CELL)ADD 1 TO SENT2IF TYPEC(CELL) NE 0IF NEXT.DESTINATIONC CELL =M.STATION ANDTYPEC(CELL)=1AND MIDCCELL)=99LET El.DLY=Y1*ABS.F(NEXT.DEsTINATICNCCELL)-STATION)*sD * SLOT/1000000.+TIME.V-BEGINC(CELL)LET D=Y1 *ABs.F(M.STATION-STATION*SD I SLOT:1000000.+0.0000013LET BEGINC(IELL)=D+TIME.VIF N.SU2=0FILE CELL IN SU2ACTIVATE A SETUP2 IN D .SECONDSOTHERWISEFILE CELL IN SU2ALWAYSALWAYSIF NEKT.DESTINATIoNC(CELL)=1 AND TYPEC(CELL =1AND MIDC(CELL)=99LET E2.ELY=Y2*ABS.F(NEXT.DEsTINATIONC 2ELL)-sTATION)*SD*SLOT:1000000.+TIME.V-BEGINC(CELL)LET D=Y2*(STATION-1*SD*SLOT/1000000.LET BEGINC(CELL)=D+TIME.VIF N.SU3=0ACTIVATE A SETUP3 IN D .SECONDSALWAYSLET sTAC;EC)CELL)=3FILE CELL IN SU3ALWAYSIF NEXT.DEsTINATICNCcCELL) ,M.STATION ANDTYPEc)CELL)=2AND MIDC(CELL)=99LET E3.FiL=13*ABs.F(NEXT.CESTINATICNC)cELL)-STATION)*sD*sLoT/1000000.+TIME_V-SECINC(CELL)LET D=Y3*ABS.F(M.sTATIoN-sTATICN)*SD*sLOT 1000000.LET BEGINCCELL)=D+TIME.VIF N.sU3=0ACTIVATE A SETUP3 IN D .SECONDSALWAYSLET STACEC(CELL)=4FILE CELL IN SU3ALWAYSIF HEY)T.DEsTINATIONCCELL) ,sOURCEC)CELL) ANDTYPEC)CELL)=2AND MID),MDELL)=99LET E4.EDY=y4*ABs.F)NEXT.DEsTINATIONC)CELL)-STATION)*sD*sLOT. 1000000.BEGIIICLET DrABs.F)NEXT.DESTINATionC)CELL)-sTATIoN)*=*sLOT/1000000.LET BEGINC)CELL)=y4*D+TIME.VIF N.Su3=0ACTIVATE A 5E721,3 IN D .SECONDSALWAYSLET STACEC(CELL)=5FILE CELL IN SU3ALWAYSIF NEXT.DESTINATIONCCELL =M.STATICN ANDTYPEccELL)=5AND 1.IIDC ,•.CELL)=99LET Eb.DDI ,Y5*ABs.FWEXT.DESTINATIGNC ELL -STATION) *stp*s=,14TINF.V-BEGINC)CELL)LET D=ABS.F(NEXT.CESTIITATIONC)CELL -STATION)*SD*sLOT/1000000.LET BEGINC)CELL)=Y5*D+TIME.VIF N.SU3,0ACTIVATE A SETUP3 IN D .SECONDSALwAYSLET STAGFC)CELL)=6FILE CELL IN SU3ALWAYSIF NEXT.DESTINATICNC)CELL)=SoURcEC)CELL) ANDTyPECuCELL)=6AND MIDCCELL) , 99LET E6.DLY ,Y6 *ABS.F)NEXT.DESTINATICNC)CELL)-STATION)*sD*SLoT/1000000.+TIME.V-BEGINC)CELL)LET D,ABS.F(IEXT.DESTINATIONC)CELL)-STATION)*SD*SLOT/1000000.+0.0000003LET DEciNc)CELL)=Y6*D+TimE.VIF N.sU3 , 0ACTIVATE A SETUP3 IN D .sECONDSALWAYSLET sTA0EC)CELL)=7FILE CELL IN SU3ALWAYSIF NEXT.DEsTINATIONC(CELL) ,M.sTATION ANDTYPEC(CELL)=CAND MIDC(CELL)=99LET E7.DLY=y7*ABS.F(NEXT.DESTINATIONC)CELL)-STATION)*SD*SLOT/1000000.+TIME.V-BEGINC(CELL)LET D=ABS.F)NEXT.DESTINATIONC(CELL)-sTATION)*SD*SLOT/1000000.+0.0000003LET BEGINC(CELL)=Y7*D+TIME.VIF N.sU3,0ACTIVATE A SETUP3 IN D .SECONDSALWAYSLET STAGEC)CELL) , 8FILE CELL IN SU3ALWAYSIF NEWT.DESTINATIONC)CELL)=DESTINATICNC)CELL)AND TYPEC(CELL)=7AND MIDCCELL)=99LET E8.DLY ,Y8*ABS.F;NEXT.DESTINATICVCCELL;-sTATICSD*SLOT/1000000.+TIME.V-BEGINC(CELL)LET C=ABS.EklIFET.DESTIHATIC,NCELL -STATICM*SD*SLOT,1O000O0.LET BECINCCELL)=Y8*D+TIME.VIF N.SIJ3=0ACTIVATE A SETUP3 IN D .SECONDSALWAYSLET STAgEC)CELL)+- ;FILE CELL III SU3ALWAYSIF NEXT.DESTIMATIONC)cELL)=M.STATION ANDTYPEC)CELL(=8AND MIDC(CELL)=99LET E9.DLY=19 * ABS.F)NEXT.DESTINATIoNC(_ELL)-STATION)*SD*SLOT/1000000.+TIME.V-BECINC)CELL)LET D=ABS.F)HEXT.DESTINATIONC(CELL -sTATION)*SD*SLOT/1000000.LET BEGINC(CELL)= -19*D+TIME.VIF N.SU3=0ACTIVATE A SETUP3 IN D .SECONDSALWAYSLET STAGEO(CELL)=10FILE CELL IN SU3ALWAYSIF NEXT.DESTINATIONC(CELL)=DESTINATIONC(CELL)AND TYPECCELL)=3AND MIDC(OELL)=99LET E1J.DLY=Y10*ABS.F(NEKT.DESTINATIONCCCELL)-STATION*SD*SLOT;1000000.+TIME.V-BEC,INCHCELL)LET D=Y10 *ABS.F(M.STATION - DESTINATIONC;CELL))*SD*SLOT:1000000.+0.0000003LET S.DELAY=TIME.V-STARTC(CELL)+DADD 1 TO CSP.COUNTLET K=1.ALWAYSIF NEXT.CESTINATIONC(CELL) =SOURCEC(CELL) ANDTYPEC2ELL)=21AND MIDC(CELL)=99LET D=L1 *ABS.FSOURCEC(CELL)-DESTINATIONC(CELL)*SD*SLOT21000000.LET BEGINCCELL)=TIME.V+DIF N.TERM.BUFFER2=0ACTIVATE A TERM.PRO2 IN D _SECONDSALWAYSLET STAGECCELL)=22FILE CELL IN TERN.BUFFER2ALWAYSIF NEXT.DESTINATIONC(CELL)=DESTINATIONC;CELL)AND TYPECPCELL)=22AND MIEKCELL;=99LET D=C2 * ABS.FSOMRCEC(CELL)-DESTINATIONC(CELL )*SD*SLOT21000000.LET BEC;INCCELL)=TIME.V+DIF N.TERM.BUFFER2=0ACTIVATE A TERM.PRO2 IN D _SECONDSALWAYSLET STAGECCELL=23FILE CELL IN TEBM.BUFFER2ALWAYSIF NEXT.DESTINATIONCCELL)=SOURCEC(CELL) AIIDTYPEC.CELL)=23AND MIDCCELL)=99LET 1, C3*ABS.FSOURCEC(CELL-DESTINATIONCCELL))*SD*SLOT21000002.LET T.DELAY=TIME.V-TERMC(CELL)+DLET STN=DESTINATIONC(CELL)IF REAL.FSTIM23. NEREAL.F;INT.FREAL.F(STN)/3.))SUBTRACT 1 FROM CHANNEL.UTHSTNALWAYSSUBTRACT 1 FROM NO.TALKINGLET TRUE.TALK=INT.F(IJO.TALKING*0.43)LET BOUNDARY=INT.FHTRUE.TALK*64.*N.SLOTA/310000.):(48./53.))ADD 1 TO CEND.COUNTLET K=1ALWAYSOTHERWISELET K=1ALWAYSIF K=1.DESTROY THIS CELLALWAYSIF N.BUFFERB2STATIOIN>0LET CDB2(STATION)=RCB2(STATION)LET RCB2(STATION)=0ADD 1 TO RQA2(STATION)OTHERWISELET CDB2(STATION)=-2ALWAYSIF BWB.MOD>0IF BWB.CNTRB(STATION)=BWB.MOD-1LET BWB.CNTRB(STATION)=0IF CDB2(STATION)=-2ADD 1 TO RCB2(STATION)OTHERWISEADD 1 TO CDB2STATION)ALWAYSOTHERWISEADD 1 TO BWB.CNTRB(STATION)ALWAYSALWAYSALWAYSALWAYSIF BUSYB;SLOTBJ=0 AND CEq31STATION)> - 2SUBTRACT 1 FROM CDB1STATIOrnIF CD81(STATICN)=-1LET BUSYBSLOTB=1REMOVE FIRST CELL FROM BUFFERB1STATION)LET DSTB(SLOTB)=NEXT.DESTINATIONC(:ELL)LET Dl.DELAY=STATION-DESTINATIONCCELLH*SD*SLOT/1000000.+TIME.V-STARTC(CELL)DESTROY THIS CELLADD 1 TO SENT1IF N.BUFFERB1STATIOIN>0LET CDB1STATICKN=RCB1STATICIMLET RCB1STATIOIN=0ADD 1 TO RQA1;STATION)OTHERWISELET CD21(STATION)=-2ALWAYSIF BWB.MOD>0IF BWB.CNTRB(STATION)=BWB.MOD-1LET BWE.CNTRB(STATION)=0IF CDB1(STATION)=-2ADD 1 TO RCB1(STATION)OTHERWISEADD 1 TO CDB1(STATICN)ALWAYSOTHERWISEADD 1 TO BWB.CNTRB(STATION)ALWAYSALWAYSALWAYSALWAYSIF BUSYB.:SLOTB)=0 AND CDBO(STATION)>-2SUBTRACT 1 FROM CDBO(STATION)IF CCBO(STATION)= - 1LET BUSYB(SLOTB)=1REMC0.7E FIRST CELL FROM BUFFERBO(STATION)LET DSTB(SLOTB) =NEXT.DESTINATIONC(CELL)LET DO.DELAY , (STATION-DESTINATIONC(CELL))*BD*SLOT/1000000.+TIME.7-STARTC(CELL)DESTROY THIS CELLADD 1 TO SENTOIF N.BUFFERBO(STATION)>0LET CDBOISTATICIM=RCB0(STATION)LET RCBO (STATION) =0ADD 1 TO RQAO(STATION)OTHERWISELET CDRO(STATION)=-2ALWAYSIF BWB.MOD>0IF BWB.CNTRB(STATION)=BWB.N0D-1LET BWB.CNTRB(STATION)=0IF CDBO(STATION)=-2ADD 1 TO RCBO(STATION)OTHERWISEAEC , 1 TO CDRO(STATICM)ALWAYSOTHERWISEAEC , 1 TO BWB.CNTRB(STATION)ALWAYSALWAYSALWAYSALWAYSALWAYSLOOPWAIT SLOT _MICROSECONDSLOOPLOOPEND"The STOP process is the simulation result report. The"statistic data collected during the simulation are finally"reported by this process.PROCESS STOP.SIMLET LOADM0=(CELLO.CREATED*44.*8.)^1000000.*(STOP.TIME/DIVISION))LET LaADM1 , (CELL1.CREATED*44.*8.)/(1000000.*(STOP.TIME/DIVISION))LET LOADM2=(CELL2.CREATED*44.*8.( ; ) (1000000.*(STOP.TIME/DIVISION))LET CSP.LOADM=(CSP.CREATED*44.*2.)^1000000.*(STOP.TIME/DIVISIOM))LET CTM.LOADM=(CTM.CREATED*44.*8.)/(1000000.*(STOP.TIME/DIVISIO•))LET SIG.LOADM=((CSP.CREATECeCTM.CREATED)*44.*8.)/(1000000.*(STOP.TIME/DIVISION))LET P.LaADM , (NO.TALKINO*64.*53.*0.43)/(1000.*48.)IF (MOBILE.TOTAL.TALKED+V.LOST)>0LET PV.LOSS=V.LOST/(MOBILE.TOTAL.TALKED+V.LOST)OTHERWISELET PV. LOSS=0ALWAYSIF (T=L.TALKED+CALL.LOST)>0LET CALL.BLOCK=CALL.LOST;(TOTAL.TALKED+CALL.LOST)OTHERWISELET CALL.BLOCK=0ALWAYSIF CELLO.CREATED>0LET LOSS0=100.*OELLO.LOSS/CELLO.CREATEDOTHERWISELOSS0=0ALWAYSIF CELL1.CREATED>0LET LOSS1=100. * CELL1.LOSS:CELL1.CREATEE ,OTHERWISELET LOSS1=0ALWAYSIF CELL2.CREATED>0LET LOSS2=100.*CELL2.LOSS/ ,-ELL2.CREATEDOTHERWISELET LOSS2=0ALWAYSLET THRCUGHPTTO=SENTO*44.*8.:( STOP.TIME/DIVISIOLE*1000000.)LET THROUGHPUT1=SENT1*44. * 8.;((STOP.TIME:DIVISION(*1000000.)LET THROUCHPUT2=SENT2*44.*8./((STOP.TIME/DIVISICN)*1000000.)LET P.THROUGHPUT=P.SENT*48.*8./((STOP.TIME/DIVISION)*1000000.)PRINT 1 LINE WITH P.SENT AS FOLLOWSP.SENT=********.LET THROTC,HPUT=THROUCHPUTO+THRCUGHPUT1+THROUGHPUT2+P.THRCUGHPUTLET ISOC.CHAN=ISOC.CHAN+HO.MAIIRESF;PRINT 36 LINES WITH STATIONARY.PER.CELL,MOBILE.PER.CELL,NODEA,NODEB,N.STATION,BWB.MOD,SD,SWITCH,TOT.ERL,ERL.PER.CELL,CALL.RATE.PER.CELL,LOADM2,LOADM1,LOADMO,CSP.LOADM,CTM.LOADM,SIG.LCADM,P.LOADM,CALL.ARV.RATE, CALL.DPT.RATE, NO.TALKING,NO.TO.TALK, TIME.V,MEANDDO*HOURS.V, SDDO*HOURS.V, MINDDO*HOURS.V, MAXDDO*HOURS.V,MEAIIDDI*HOURS_V, SDDl*HOURS.V, MINDD1*HOURS.V, MAXDDl*HOURS.V,MEANDD2*HOURS.V, SDD2*HOURS.V, MINDD2*HOURS.V, MAXDD2*HOURS.V,MEANEl*HOURS.V, SDE1*HOURS.V, MINE1*HOURS.V, MAXE1*HOURS.V,MEANE2*HOURS.V, SDE2*HOURS.V, MINE2*HOURS.V, MAXE2*HOURS.V,MEANE3*HOURS.V, SDE3*HOURS.V, MINE3*HOURS.V, MAXE3*HOURS.V,MEANE4*HOURS.1, SDE4*HOURS.V, MINE4*HOURS.V, MAXE4*HOURS.V,MEANES*HOURS.V, SDES*HOURS.V, MINES*HOURS.V, MAXES*HOURS.V,MEANE6*HOURS.V, SDE6*HOURS.V, MINE6*HOURS.V, MAXE6*HOURS.V,MEANE7*HOURS.V, SDE7*HOURS.V, IIINE7*HOURS.V, MAXE7*HOURS.V,MEAME8*HOURS.V, SDE8*HOURS.V, MINE8*HOURS.V, MAXE8*HOURS.V,MEANE9*HOURS.V, SDE9*HOURS.V, MINE9*HOURS.V, MAXE9*HOURS.V,MEANE10*HOURS.V, SDE10*HOURS.V, MINE10*HOURS.V,MAXE10*HOURS.V,MEANSD*HOURS.V, SSD*HOURS.V, MINSD*HOURS.V, MAXSD*HOURS.V,MEANTME*HOURS.V, STMD*HOURS.V, MINTMD*HOURS.V, MAXTMD*HOURS.V,ISOC.CHAN1,HO.MANRES,PV.LOSS,CALL.BLOCK,BUFFER.CAPACITY,LOSS2,LOSS1,LOSSO,THROUGHPUT2,THROUGHPUT1,THROUGHPUT°, P.THROUGHPUT,THROUGHPUT,CSP.COUNT, CEND.COUNT,CSP.LOST AS FOLLOWSCall setup/Call procedure for SST/SBWM/OB Simulation Result(NAMJIAN (NAN)Stationary.per.cell=********. Mobile.per.cell=********.EN in*** .a, *** . h,Bus capacity = 155Mbps. Number of stations =^*. BWB.MODE= *.Node Distance(in slots) = ***. Erasure Node available = ***.Total voice load = *^Erlang. (****.** Erlang/cell new callrate/cell=***.**)^traffic load2=***.**Mbps. loadl=***.**Mbps.load0=***.**Mbps.csp.load=***.**Mbps. ctm.lced=***.**Mbps. sig.load=***.**Mbps.P.load:***.**Mbps. CALL.ARV.RATE:****.**call/sec.CALL.DPT.RATE:****.**call/sec.NO.TALKING: ******.NO.TO.TALK: ******.After******.^secondsof simulation:* * * * ** * * * *^*. * * * * **.** ****.** **** * * * *call setup delay^*.***** *.*****^*.*****call terminate delay *.*****^*.*****^*.*****^*.*****With ****_ isochronous channels on the MAN, of which^* arereserved for interMAN handoffsThe call blocking probability of a mobile user = *.*****The call blocking probability of total=*.*****With a buffer capacity of * each station/bus, the cell dataloss in each priority level is: levell= *.**%. level2= *.**%.level3= *.**%. The data throughput in each priority level is:level2= ****.***Mbps.levell. ****.***Mbps.leye10= ****.***Mbps.PA.throughput=*****.***Mbps.Total.throughput=*****.***Mbps.CSP.NUMBER:******. CEND.NUMBER:*******. CSP.LOST:********."STOPENDTABLE OF SIMULATION RESULTS(UNIT: milliseconds)Mean^std_dev Minimum Maximumpriority° cell delay *.*****^*.*****^*.*****^*.*****priorityl cell delay *.***** *.*****priority2 cell delaystage 1^cell delaystage 2^cell delaystage 3^cell delaystage 4^cell delaystage 5^cell delay *.*****stage 6^cell delay *stage 7^cell delay *.*****stage 8^cell delay *.***** *.stage 9^cell delay *stage 10 cell delay** * * *

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0065103/manifest

Comment

Related Items