UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

An ATM-based interworking architecture for wireless personal communications Wong, Terrence Sui Wing 1996

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

Item Metadata

Download

Media
831-ubc_1997-0085.pdf [ 4.34MB ]
Metadata
JSON: 831-1.0065321.json
JSON-LD: 831-1.0065321-ld.json
RDF/XML (Pretty): 831-1.0065321-rdf.xml
RDF/JSON: 831-1.0065321-rdf.json
Turtle: 831-1.0065321-turtle.txt
N-Triples: 831-1.0065321-rdf-ntriples.txt
Original Record: 831-1.0065321-source.json
Full Text
831-1.0065321-fulltext.txt
Citation
831-1.0065321.ris

Full Text

AN ATM-BASED INTERWORKING ARCHITECTURE FOR WIRELESS PERSONAL COMMUNICATIONS by TERRENCE SUI WING WONG B. Eng., Lakehead University, Canada, 1994 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE STUDIES DEPARTMENT OF ELECTRICAL ENGINEERING  We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH C O L U M B I A December 1996 © Terrence Wong, 1996  in presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make' it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department  or  by his  or  her  representatives.  It  is  understood  that  copying or  publication of this thesis for financial gain shall not be allowed without my written permission.  Department of  T^'^&'fy/S/J^Cr  The University of British Columbia Vancouver, Canada  Date  DE-6 (2/88)  /JW:  )  f/f  f  .  Cttet ((/>-  Abstract The growing demand for mobile communication services as well as technological advances drive the new generation of Personal Communication Networks (PCNs). New or modified networks and communication protocols are required to support associated Personal communication Services (PCS).  -  An ATM-based architecture has been proposed to support future PCS. The network would provide improved PCS performance for applications such as hand-off and call set up. CDMA technology is assumed as the air interface, with a bandwidth in each cell site of 2.5 MHz and a sectorization antenna with 3 sectors/cell. The PCS radio cell coverage size varies depending on geographic location; radii of 275 m and 981 m are proposed for microcells in metropolitan and urban areas, respectively. The topology of the proposed network is a star configuration which connects the base stations (BS) to the mobile switching centre (MSC). The proposed network which  ! interconnects 16 switching centers can support up to 3.0 million potential PCS users with a coverage of 600 K m in metropolitan area. 2  Strategies such as fast hand-off routing and signalling schemes are presented based on the virtual connection tree concept. Radio cells with a 50 percent overlapping area for improving hand-off performance are considered. Radio cell radii are calculated in different environments with various market penetrations. Cell site overload probability for voice and data users within a connection tree region is studied. Network performance in terms of end-to-end delay, cell delay variation, and buffer occupancy are investigated in depth.  Table of Contents Abstract  ii  List of Tables  vi  List of Figures  vii  Acknowledgment  x  Chapter 1 Introduction  1  1.1  Background and Motivation  1.2  Objectives  1.3  Outline of the thesis  2 ;  3  Chapter 2 ATM-Based PCS and The ATM Protocol 2.1  2.2  3  5  ATM-Based PCS.:....  5  2.1.1  General Architecture of ATM-Based PCS  5  2.1.2  ATM-Based PCS Wireless Interface  2.1.3  ATM-Based Signaling  2.1.4  Associated.and Quasi-associated Mode ATM Signaling  ,  ...5 8  Asynchronous Transfer Mode (ATM) Protocol Specifications  .....9 :..  12  2.2.1  Formal Parties Involved in ATM Standards Design  12  2.2.2  The B-ISDN Protocol Reference Model  13  2.2.3  ATM Cell Format  14  2.2.4  Transmission Path, Virtual Path, and Virtual Channel in the ATM Network.. 17  2.2.5  ATM Layer Protocol Model  2.2.6  ATM Adaptation Layer (AAL) Protocol Model  iii  .18 ....19  <  iv  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture 22 3.1  Characteristics of the PCS Radio Cell Environment  22  3.2  Variable Bit Rate for Voice Encoding in CDMA Systems  22  3.3  PCS Radio Cell Capacity  .,  23  3.3.1  Sectorization Antenna Gain  24  3.3.2  Unbalanced Traffic Loads for Cell Sites in Urban Areas  24  3.4  CDMA Based Mobile Initiated Soft Hand-off Scheme  25  3.5  ATM-Based Virtual Connection Tree Architecture to Support Soft Handoffs  27  3.5.1  Neighboring Mobile Acces s Region  27  3.5.2  Fast Routing Scheme with Virtual Connection Number (VCN) Lookup Table. 29  3.6  3.7  Mobile Initiated Handoff Signaling with Connection Tree Architecture  33  3.6.1  Handoff Signaling for Two Base Stations Connected to the Same Switch ..33  3.6.2  Two Base Stations Connected to Different Switches  34  Base Station Overload Probability  36  Chapter 4 Network Coverage, Modeling and Simulation Tool (OPNET) Specifications 40 4.1  Radio Cell Coverage  40  4.1.1  Number of Potential Users in a Connection Tree Region  40  4.1.2  Radio Cell Coverage in Metropolitan, Urban, and Suburban Areas  41  4.2  Network Simulation Model  44  4.3  Simulation Tool (OPNET) Specifications  46  Chapter 5 Simulation Results for Voice and Data Traffic 5.1  50  Simulation Results for Data Traffic Over an ATM Network 5.1.1  Delay and Buffer Occupancy Analysis of Data Traffic  50 ;  50  5.1.2 5.2  Delay Analysis of Intra and Inter-network Data Traffic  Simulation Results for Mixed Voice and Data Traffic  .54 60  5.2.1  Delay Analysis of Voice and Data Traffic with Variable Data Traffic Loading 61  5.2.2  Delay Analysis of Voice and Data Traffic with Different Voice Coding Rates  66  Delay Analysis of Inter- and Intra-Network Voice and Data Traffic  69  5.2.3  Chapter 6 Simulation Results for Voice and Data Traffic with Silence Deletion  73  6.1  The Two-States Markov Silence Deletion Model  73  6.2  Delay and Throughput Analysis of Voice Traffic with Silence Deletion  74  6.3  Inter- and Intra-Network Performance With Silence Deletion Scheme  78  6.4  Network Coverage with Various traffic and Market Penetrations  81  Chapter 7 Summary and Conclusions  83  7.1  Summary and Findings  83  7.2  Topics for Future Investigation  84  Bibliography  86  Appendix A. List of Abbreviations and Acronyms  91  Appendix B. Calculation of Confidence Interval  94  Appendix C. OPNET ATM Models  95  List of Tables Table 2.1  Summary of ATM Signaling transport architecture  Table 2.2  B-ISDN/ATM service classes  20  Table 3.1  Lookup table for two base stations connected to same switch  31  Table 3.2  Switch/VCN assignments with . two base stations connected to different switches. •  9  32  Table 3.3  Results for base station overload probability  39  Table 4.1  Number of potential voice and data users in a connection tree  41  Table 4.2  Number of potential users in different radio environments  42  Table 4.3  Radio cell radii in different environments  44  Table 5.1  Maximum buffer occupancy in central and gateway switches  53  Table 6.1  Coverage area per switching node in metropolitan areas with various market penetrations Results of confidence interval calculation  82 94  Table B . l  vi  List of Figures Figure 2.1  General architecture of PCS over ATM  6  Figure 2.2  An ATM-compatible PCS approach wireless interface.  7  Figure 2.3  ATM service and control transport network.  Figure 2.4  ATM-based quasi-associated signaling transport architecture and protocol stacks of network elements involved in signaling transport 10  Figure 2.5  ATM-based associated signaling transport and protocol stacks of network  ....8  elements involved in signaling transport  :  11  Figure 2.6  B-ISDN protocol reference model  14  Figure 2.7  ATM cell efficiency and packetization delay.  15  Figure 2.8  ATM UNI and NNI cell structure  16  Figure 2.9  Switching functions for VP and VC switches  17  Figure 2.10  End system and intermediate system of VP/VC function  18  Figure 2.11  A A L protocol sublayer model  20  Figure 2.12  CPCS-PDU and SAR-PDU formats for the A A L type 5  21  Figure 3.1  Unbalanced traffic load in rush hours  25  Figure 3.2  Channel assignment during soft handoff.  26  Figure 3.3  Hierarchical interconnection network  28  Figure 3.4  Two base stations connected to same switch.  30  Figure 3.5  Handoff involving two different switches  32  Figure 3.6  Handoff signaling for two base stations connected to same switch  .34  Figure 3.7  Handoff signaling involves two switches  35  Figure 3.8  Base station overload probability plotted against number of voice connections. ..38  Figure 3.9  Base station overload probability plotted against number of data connections  ,  vii  39  Figure 4.1  Fifty percent overlapping microcell environment  43  Figure 4.2  Simulation model  Figure 5.1  Node_0 to node_15 mean data cell delay.  51  Figure 5.2  Central switch mean buffer occupancy vs. total traffic  52  Figure 5.3  Gateway switches 0 and 2 mean buffer occupancy.  52  Figure 5.4  Node_0 to node_15 cell delay variance vs. total traffic  54  Figure 5.5  Node 0 to node 15 mean delay with various intra-network traffic  55  Figure 5.6  Node 0 to node 1 mean data packet delay with various intra-network traffic  Figure 5.7  Mean data packet delay from gateway switch 0 to nodes 2 and 3  57  Figure 5.8  Central switch mean buffer occupancy.  58  Figure 5.9  Gateway switches 2 and 3 mean buffer occupancy  58  Figure 5.10  Node 0 to node 15 cell delay variance  60  Figure 5.11  Node 0 to node 15 data packet delay with various data channels per node  61  Figure 5.12  Node 0 to node 15 voice packet delay with various data channels per node  62  Figure 5.13  Mean voice packet delay with various data channels per node.  63  Figure 5.14  Mean data packet delay with various data channels per node  64  Figure 5.15  Node 0 to node 15 data ATM cell delay variance with various data channels per  '.  node  45  56  ..65  Figure 5.16  Central switch buffer occupancy  65  Figure 5.17  Gateway switches 0 and 2 buffer occupancy.  66.  Figure 5.18  Node 0 to node 15 mean data packet delay with various voice coding rates  67  Figure 5.19  Mean data packet delay with various voice coding rates:  67  Figure 5.20  Central switch buffer occupancy with various voice coding rates  68  Figure 5.21  Mean data buffer occupancy with various voice coding rates. viii  ..69  Figure 5.22  Node 0 and node 15 mean data packet delay with various intra-network traffic. ..70  Figure 5.23  Mean data packet delay with various intra-network traffic  71  Figure 5.24  Data ATM cell delay variance with various intra-network traffic  72  Figure 5.25  Mean data buffer occupancy with various intra-network traffic  72  Figure 6.1  Two state markov silence deletion model  73  Figure 6.2  Node 0 to node 15 mean data packet delay  76  Figure 6.3  Mean data packet delay.  76  Figure 6.4  Mean data ATM cell delay variance  77  Figure 6.5  Mean data buffer occupancy.  77  Figure 6.6  Node 0 to node 15 mean data packet delay.  78  Figure 6.7  Mean data packet delay.  79  Figure 6.8  Mean data packet ATM cell delay variance  80  Figure 6.9  Mean data buffer occupancy.  81  Figure C.l  ATM source node model  95  Figure C.2  ATM network to network switch  95  ix  Acknowledgment I wish to express my great appreciation to my research supervisor Dr. Robert W. Donaldson. There is not a page in this thesis that has not benefited from his insights and careful critiques. He helped me rethink the research material and pushed me in new directions. I also wish to express my appreciation to my colleagues for their kindness and consistent help throughout my study in UBC. In addition, I wish to express special thanks to the Canadian Institute of Telecommunications Research for the financial support of this project. Finally, my thanks go to my family and friends for their encouragement and support throughout my graduate study.  Chapter 1 Introduction Wireless personal communications have been gaining popularity for the past decade. The first generation of mobile communication systems, called Advanced Mobile Phone System (AMPS) by AT&T and Nordic Mobile Telephone (NMT) by Ericsson, were introduced in 1970s [1]. In recent years, due to the high demand for personal mobile communication services, advances in technology have been developed. The air interface technique for mobile communication has migrated from an analog (FDMA or TDMA) to a digital (CDMA or TDMA) system. A digital air interface environment is expected to increase the system capacity several times over that of the analog system [2]. During the mid 1980s, a number of digital mobile communication systems were being implemented. Typical systems include the Global System for Mobile Communication (GSM) which was implemented in Europe and most Asian counties, the digital AMPS (DAMPS), implemented in North America, and the Japanese digital cellular system (JDC), implemented in Japan. With the evolution of the second generation in personal communication systems, voice will not be the only traffic in the network; data and low bit rate video traffic (such as video conferencing) transmitted or received, will be transported to orfroma portable terminal. Current mobile systems with their limited air interface bandwidth are not adequate to meet the increasing demand for mobile communication services. New spectrum from 1850 to 1990 MHz has been allocated for future PCS. The CDMA technique will be used in the PCS air interface to utilize this new spectrum more effectively. Second generation of PCS will enable communication with other PCS users regardless of their geographical location. The coverage area of the future PCS will not be limited to urban  1  Chapter 1 Introduction  2  areas, as with current mobile communication system, but also will extend to rural areas, which may be covered by low earth orbit satellite systems (LEO) such as IRIDIUM [3].  1.1 Background and Motivation With a 20 to 50 percent growth.in various parts of the world, wireless communications have drawn the serious attention of service providers [1]. The proposed PCS will support various kinds of traffic (voice, data, video) using microcells with radii of 200 to 500 meters in metropolitan areas. Small radio cells enable system capacity gains through frequency reuse. However, small cells increase hand-off traffic and signaling network complexity. The current mobile network with the centralized database approach will not meet the requirements of future PCS [1]. In previous years, the Metropolitan Area Network (MAN), based on the IEEE 802.6 standard, has been proposed to meet the requirements necessary to support PCS [4]. A thorough study of this network in terms of architecture, signaling, packet delay, and throughput analysis has been completed [5] [6] [7] [8]. Recently the focus on back-bone networks to support PCS has moved toward Asynchronous Transfer Mode (ATM) architecture. Unlike MANs, the topology of an ATM network can vary (e.g, star topology, partial and full mesh topology). ATM networks have gained significant interest from PCS researchers because of the capability to support various protocols such as TCP/ IP. An ATM adaptation layer creates universal environment for any packet format that is received from the transport layer. ATM is a combined protocol with network, data link and physical layers. With this simplification, the link layer cost can be reduced due to the increased in speed.  Chapter 1 Introduction  3  1.2 Objectives ATM networks which support PCS are expected to carry multimedia traffic. In this work, a network with ATM switches interconnected by the Synchronous Optical Fiber Network (SONET) is considered. The infrastructure of a given ATM network may vary depending on geography limitations and traffic load conditions. This thesis involves a study of the traffic characteristics and performance of a star topology ATM network. Thesis objectives can be catalogued as follows: •  To examine the utility of the virtual connection tree architecture in an ATM-based PCS.  •  To calculate the required PCS air interface bandwidth allocation and its system capacity.  •  To analyze the suitability of PCS over a star topology ATM network.  •  To compare the network performance of single-mode and multi-mode traffic conditions.  •  To analyze network congestion.  •  To test the relationship between voice and data packet delays under voice silence deletion.  1.3 Outline of the thesis Chapter 2 describes the ATM standards protocol as specified by ITU/CCITT. Items relevant to PCS are identified and discussed. Also, some general requirements and considerations for future PCS networks are identified. In chapter 3, the PCS radio cell environment with the CDMA air interface is examined. Aspects such as channel allocation and system capacity are  Chapter 1 Introduction  4  defined. A virtual connection tree architecture with applications to PCS is also considered. Chapter 4 articulates the network modeling and specifications of the software simulation tool (OPNET) used for PCS/ATM performance analysis. Simulation results for the single- and multimode traffic conditions are considered in chapter 5. A multimedia traffic condition with a silence deletion to the voice source is examined in chapter 6. Chapter 7 summarizes the results and provides some suggestions for future research.  Chapter 2 ATM-Based PCS and The ATM Protocol 2.1 ATM-Based PCS Although the initial PCS is designed primarily for voice communication, it has been recognized that the system should contain data and video traffic to support a wider range of telecommunication applications over the wireless network [10]. The new generation of PCS, based on the ATM architecture, has been standardized by the International Telegraph and Telephone Consultative Committee (CCITT). ATM-based PCS has a number of advantages over alternatives. Bursty data traffic is handled effectively by the ATM variable bit rate (VBR) scheme, while the constant bit rate (CBR) scheme can support traffic such as voice or large file transfer.  2.1.1 General Architecture of ATM-Based PCS Next-generation PCS will involve both wireline and wireless communications. Figure 2.1 shows a typical ATM network architecture. For a given ATM network, a switch's capacity at each level is designed according to the traffic load. Several base stations are multiplexed onto one ATM switch. A number of local switches are connected to a central switch which is responsible for interconnecting other wire line networks such as the public service telephone network (PSTN), the Internet, or other PCS networks. Since ATM may be the major back-bone network for future PCS, compatibility with other networks (e.g, wireless L A N or MAN) is also important.  2.1.2 ATM-Based PCS Wireless Interface Future PCS are expected to provide high quality wireless communication. The current mobile hand-off scheme with "break before connection" will not meet future needs: Advanced techniques which provide soft or seamless hand-off are proposed [11]. In the actual radio cell  5  Chapter 2. ATM-Based PCS and The ATM Protocol  Figure 2.1  6  General architecture of PCS over ATM.  environment, a coverage gap may form between two radio cells especially in downtown areas with high rise buildings. A scheme which employs a macrocell on top of several microcells has been implemented to address this problem [12]. The advantage of this scheme is not only beneficial to the hand-off traffic but also can reduce the blocking probability caused by traffic congestion in the microcell by accommodating traffic overloads. It is anticipated that PCS will eventually carry multimedia traffic. Bursting data traffic and delay-sensitive voice traffic may be handled by the same air interface. A flexible bandwidth allocation must be considered. To handle different traffic types, a given air interface packet format  Chapter 2 ATM-Based PCS and The ATM Protocol  7  would vary in terms of its length and discard priorities. Figure 2.2 shows one possible ATMcompatible PCS wireless air interface.  1240 Bytes  H Mobile Data Terminal  PCS UNI  H  48 bytes A T M payload  User data 9.6 kbits/s A T M cell  ATM NNI  PCS UNI  voice frame  ATM based backbone network  K—H Mobile Phone  PCS UNI  I—V\  I—PI  8 kbits/s  Wireless PCS Network  Base Station  ATM Backbone Infrastructure  —  //A  Wireless PCS Air Interface Cell Header UNI: User to Network Interface  Figure 2.2  A T M Cell Header  NNI: Network to Network Interface  A n ATM-compatible PCS approach wireless interface.  Mobile terminals are expected to handle various kinds of traffic (voice, data, and video) and likely would inform the network of quality of service (QOS) requirements. An air interface cell header would be added to the data unit or voice frame before it is sent over the radio frequency (RF) channel. At the base station, air interface cell header information would be translated into an ATM cell header. The payload of an ATM cell would be filled by the air interface cell. The structure of an ATM cell is discussed in section 2.2.3.  8  Chapter 2 ATM-Based PCS and The ATM Protocol  2.1.3 ATM-Based Signaling As telecommunication vendors plan to provide multi-media personal communication services, the current network capacity and its associated signaling scheme may not be able to support future needs. The proposed ATM-based PCS network differs from the current mobile network in several ways. The major differences are not limited to the differences in transmission speeds but also to their signaling management schemes. The broadband-ISDN signaling management scheme has been proposed by ITU recommendations 1.311 and Q.2010 [11] [12] as well as by the telecommunications industry [13]. To efficiently use the ATM transport capacity, both services and control signaling messages are carried by the same physical network. As shown in Figure 2.3, both the ATM control signal and user data are carried in the same physical network but are separated from each other by using different virtual connections (VCs) or virtual paths (VPs). The use of the ATM VP transport network to carry control signaling reduces network cost and operational complexity.  Intelligent nodes  Figure 2.3  Service control  O&M  A T M service and control transport network.  Traffic management  Chapter 2 ATM-Based PCS and The ATM Protocol  9  2.1.4 Associated and Quasi-associated Mode ATM Signaling The design of B-ISDN signaling is still an open issue for network management designers. The major concern when designing these signaling schemes is to enable support future PCS, high speed data services, and today's signaling systems such as SS7 and POTS. Table 2.1 summarizes the functions of associated and quasi-associated mode signaling architectures. Table 2.1  Summary of A T M Signaling transport architecture. AssiK-ialOil  Network Configuration  Mode  Signaling messages between two SSPs are transported without higher layer termination (No STPs required).  Quasi-Associated Mode Signaling messages between two SSPs are transported through B-STPs (involve B-STPs). Signaling and user messages are sent over by separate networks. One-to-one replacement of today's dedicated copper data transport.  Signaling Message Routing  Network layer routing (eg., MTP-3).  Network layer routing (MTP-3). A T M layer routing (without MTP-3 function in A T M switches).  Database Access  B-SSP to B - S C P  B-SSP to B-STP to B-SCP  The proposed quasi-associated mode ATM signaling scheme is based on today's SS7 signaling concept. Enhancement of broadband signal transfer point (B-STP) and broadband signal control point (B-SCP) capability is required for the broadband network. Figure 2.4 shows a quasiassociated signaling transport architecture and its protocol stacks of network elements involved in signaling transport. In this architecture, broadband service switching point (B-SSP) communicates with B-SCP through B-STP. The B-STP is an ATM switching system with an message transfer path (MTP-3) controller which only supports signaling traffic. In order to achieve the reliability and management capability of a broadband network, the number of B-SCPs must increase from today's signaling network. By increasing this number, the number of signaling links  Chapter 2 ATM-Based PCS and The ATM Protocol  10  between B-SSPs and B-SCPs will also be increased. Eventually, the signaling network may become complicated and the system capacity growth would become limited.  Connectivity of virtual path for signaling  Connectivity of signaling links  Higher layer signaling  1  MTP-3  1 1  AAL  vc  1  VP  VP  1  PL  PL  i  B-SSP  1  1  1  1  I  i  VPX  Virtual path for signaling  Figure 2.4  Higher layer signaling  —  —  —  —  —  MTP-3 AAL VC VP PL  B-STP Signaling links  *-  1  MTP-3  |  AAL  1  VC  1  1 1i  VP PL B-SSP VPX: Virtual path cross connect  ATM-based quasi-associated signaling transport architecture and protocol stacks of network elements involved in signaling transport.  An example of an associated signaling transport architecture is shown in Figure 2.5. In this architecture, two adjacent B-SSPs are interconnected with VPX and B-SSP by a virtual path. The STP function is distributed into each signaling node. For the associated mode, the B-SCP may include a network routing function (MTP-3), with the signaling message routing based on distributed processing.  11  Chapter 2 ATM-Based PCS and The ATM Protocol  B-SSP plane  VPX: Virtual path cross connection system. B-SSP includes part of STP (MTP-3) functions.  Figure 2.5  Virtual path for user services. p a t h  for  s  i  g  n  a  l  m  g  m  e  s  s  a  g  e  ATM-based associated signaling transport and protocol stacks of network elements involved in signaling transport.  As discussed previously, the existing SS7 network may require more than two SCPs and more physical connections between STPs and SCPs to support new services requiring access to SCPs. This could impose a potential limitation for the SS7 network whose potential scalability limitation in the existing SS7 network may be alleviated when ATM-based signaling transport is introduced. By partitioning and replicating the database one can meet, in a cost effective manner, the high capacity and reliability that will be required of B-SCPs. A cost comparison of the quasi-associated and associated mode ATM signaling schemes indicates that the quasi-associated mode scheme requires less initial setup cost. The primary benefit of the quasi-associated architecture is that it can reuse most of today's SS7 network. This architecture simply uses ATM signaling links as a one-to-one replacement of today's dedicated  Chapter 2 ATM-Based PCS and The ATM Protocol  12  copper signaling link. However, the quasi-associated architecture may not be economically efficient when long-term operation costs are considered. This is because the associated signaling transport architecture can support both user services and signaling messages in the same physical network. Since any two signaling points in an ATM network may already carried user services, the cost of adding additional signaling connections on the same connection path is small. This is particularly true when large data files or video services occupy the channel.  2.2 Asynchronous Transfer Mode (ATM) Protocol Specifications Asynchronous Transfer Mode (ATM) is a cell-based switching and multiplexing technology. It is designed for general purpose and connection-oriented transfer of various services. ATM can support both connection-oriented and connectionless traffic directly from its upper layer by using adaptation layers. The virtual channels operate in a bandwidth-on-demand fashion, with either constant bit rate (CBR) or variable bit rate (VBR) mode being used. ATM offers the potential to standardize on one network architecture defining the multiplexing and switching method; with SONET/STM, it provides the basis for the physical transmission standard in very high speed rate. ATM also supports multiple quality of service (QOS) classes for different applications with different delay and priority requirements.  2.2.1 Formal Parties Involved in ATM Standards Design There are two parties involved in B-ISDN/ATM standardization and specification. The formal international standards party is the International Telecommunications Union (Telecommunications Standardization Sector), known as CCITT. The premier formal B-ISDN/ATM standards organization in the United States is the American National Standard Institute (ANSI). The  \  Chapter 2 ATM-Based PCS and The ATM Protocol  1  3  premier formal B-ISDN/ATM standards organization in Europe is the European Telecommunication Standard Institute (ETSI). These formal parties defined the concept of B-ISDN in the late 1980s and chose the ATM technology as the basis for future standards. The research of B-ISDN/ATM was slow initially, but proceeded more rapidly when four major industry forums were formed. These include the ATM forum, the Internet Engineering Task Force (IETF), the Frame Relay forum, and the SMDS Interest Group (SIG). They are not formal standards committees but are independent groups formed by vendors, users, and industry experts who want to ensure standards for inter-operation ability. In particular, the ATM forum has provided valuable contributions to the formal standards organizations. Since 1990, ITU-T/CCITT has published the following recommendations regarding B-ISDN/ATM standards: •  1.150 B-ISDN Asynchronous Transfer Mode Functional Characteristics  •  1.356 B-ISDN ATM Layer Cell Transfer Performance  •  1.361 B-ISDN ATM Layer Specification  •  1.362 B-ISDN ATM Adaptation Layer (AAL) Functional Description  •  1.363 B-ISDN ATM Adaptation Layer (AAL) Specification  2.2.2 The B-ISDN Protocol Reference Model The initial recommendations for B-ISDN were published in 1988 (1.321) by CCITT but were not approved until 1990. These recommendations include the B-ISDN protocol reference model (PRM), as shown in Figure 2.6.  Chapter 2 ATM-Based PCS and The ATM Protocol  Figure 2.6  14  B-ISDN protocol reference model.  The user and control planes span down from the higher layer, through the AALs, to the ATM layer and physical layer. The control plane is used for connection control, including the call connection set-up and release functions. Once a connection is established; the user data are transmitted using one of the protocols in the user plane. The management plane is further divided into layer management and plane management. As shown in Figure 2.6, layer management interfaces with each layer in the control and user planes. Plane management has no layered structure and is currently only an abstract concept with little standardization. The A A L supports the higher layer functions of user and control planes. The ATM layer defines how information supplied from higher layers is to be mapped into the physical layer. The physical layer provides for transport over the network. The services offered by this layer include bit timing, network clock, and bit error rate bounds.  2.2.3 ATM Cell Format CCITT has selected the ATM cell format to be 53 bytes, with a 5-byte header and 48-byte  Chapter 2 ATM-Based PCS and The ATM Protocol  15  payload. The relatively small size of the payload field keeps the packetization delay relatively low; this delay can become significant with a large payload size and low source encoding rate. The 5-byte header was chosen based on cell efficiency. Figure 2.7 shows the efficiency and packetization delay versus the payload size. As shown in thefigure,packetization delay is seen to be directly proportional to payload size. ATM cell efficiency increases minimally for payload sizes larger than 48 bytes.  0  32  64  96  128 160 192 224 256  A T M Cell Payload Size (Bytes) Figure 2.7  0  32  64  96  128 160 192 224 256  A T M Cell Payload Size (Bytes)  A T M cell efficiency and packetization delay.  The ATM cell header identifies the destination, cell type, and cell priority. The payload contains either user or network signaling information, depending on whether the cell is used for data or signal transport. In order to accommodate traffic between a private-to-public network interface or public-to-public network interface, two ATM cell structures have been created. These are user-to-network interface (UNI) and the network-to-network interface (NNI), respectively. The UNI is the interface between the user and the network switch. The NNI is the interface between switches or between networks.  Chapter 2 ATM-Based PCS and The ATM Protocol  16  The ATM UNI and NNI cell structures are shown in Figure 2.8. The UNI cell header contains destination address in two parts. Thefirstpart is the 8 bits virtual path identifier (VPI). The second part is the 16 bits virtual channel identifier (VCI), which holds local significance only. The cell header also contains 4 bits generic flow control (GFC), 3 bits payload type (PT), and 1 bit cell loss priority (CLP). The entire header is error protected by a one byte header error check (HEC) field. A fundamental concept of the ATM is that switching occurs based upon the VPI & VCIfieldsof each cell. Switching done on the VPI only is called a virtual path connection (VPQ, while switching done on both VPI & VCI values is called a virtual channel connection (VCC). The NNI ATM cell has a format identical to the UNI format with one exception: there is no GFC field in the NNI cell; instead, the NNI uses the 4 bits GFC field to increase the VPI field to 12 bits.  GFC - Genericflowcontrol  53 bytes 48 bytes  Header  VPI - Virtual path identifier VCI - Virtual channel identifier  Payload  PT - Payload type GFC  VPI  VPI  8-5  4-1  8-5  VCI  VCI  VCI  PT  4-1  8-1  8-5  4-2  CLP 1 '  HEC 8-1  Payload 384- 1  CLP - Cell loss priority ,  HEC - Header error check  ATM UNI cell structure 53 bytes  VPI  VPI  VCI  VCI  VCI  PT  CLP  HEC  8-1  8-5  4-!  8-1  8-5  4-2  1  8-1  ATM NNI cell structure Figure 2.8  A T M U N I and N N I cell structure.  Payload 384- 1  :  Chapter 2 ATM-Based PCS and The ATM Protocol  17  2.2.4 Transmission Path, Virtual Path, and Virtual Channel in the ATM Network A transmission path can be viewed as a physical optical fiber, with one or more virtual paths. A typical virtual path has a capacity of 155.52 Mbps if SONET STS-3 is used in the physical layer. A virtual path can support many virtual channels since virtual channels are multiplexed into a virtual path in the output port of an ATM switch. In an ATM network, the switching function can be performed at the transmission path, virtual path, or virtual channel level. Switches which only perform virtual channel connections are commonly called V C switches. A VP switch is used when VCC and VPC are switching together. Figure 2.9 illustrates the basic function of the VC and VP switches.  V C Switch Transmission Path  OjJ  M~t>  Virtual Channel Virtual Path  ————  X  V C Switch Figure 2.9  V P Switch  V C Switch  Switching functions for V P and V C switches.  The VPI and the VCI in the ATM cell header are important parameters in determining path routing. VPI and VCI values must be unique to a specific transmission path. The VPI value is used when virtual path connections are switching. However, virtual channel connections are switched  18  Chapter 2 ATM-Based PCS and The ATM Protocol  using combined VPI and VCI values.  2.2.5 ATM Layer Protocol Model The ATM layer is common to all services and provides cell transfer capabilities. The ATM layer characteristics are independent of the physical medium used. The ATM layer provides for cell multiplexing, and demultiplexing, and for routing functions using the VPI and VCI fields of the cell header. Furthermore, cell flow supervision occurs at the ATM layer to ensure that the connections stay within the service quality. The ATM layer is also responsible for cell sequence integrity for each source, but no retransmission or error correction is performed at this layer. The ATM layer interfaces with the physical layer through a physical service access point (PHY-SAP) using the request and indicate primitives. The ATM entity passes one cell per request and accepts one per indicate primitive, as illustrated in Figure 2.10.  A A L Layer  A A L Layer  IATMSAP  A T M SAP|  A T M Layer A T M Entity  A T M Entity) A T M Layer  A T M Entity  |  A T M Layer er r  (Physical Layer  Physical Layer j 1  Physical Layer  Intermediate System  End System SAP - Service Access Point  #  (6)  - VP or V C Endpoint  \ PHY SAP  End System O  - VP or V C Connecting Point  Figure 2.10 End system and intermediate system of VPfVC function.  Chapter 2 ATM-Based PCS and The ATM Protocol  19  The ATM layer also interface with the adaptation layer through an ATM service access point (ATM-SAP) similar to two PHY-SAP primitives. ATM.Data.Request initiates the transfer of an ATM service data unit (ATM-SDU) and its associated SDU type to the ATM layer peer entity over an existing connection. An ATM-SDU contains 48 bytes of user data, including the A A L header. The loss priority and the SDU type parameters are used to assign the proper CLP and PTI fields to the corresponding ATM physical data unit (ATM-PDU) generated at this layer. ATM.Data.Indication indicates the arrival of an ATM-SDU over an exiting connection with a congestion indication and the received SDU type.  2.2.6 ATM Adaptation Layer (AAL) Protocol Model The B-ISDN/ATM protocol model adapts the services provided by the ATM layer to those required by the higher layers through the ATM adaptation layer (AAL). Figure 2.11 shows the logical interface of the A A L model. Services are provided to higher layers by an A A L service access point (AAL-SAP) across which primitives regarding the AAL-PDU are passed. The A A L is subdivided into the convergence sublayer (CS) and the segmentation and reassembly (SAR) sublayer. The CS sublayer is further subdivided into service specific (SS) and common part (CP) components. ITU-T recommendation 1.362 defines the basic principles and classification of A A L functions. The attribute of the service class is the timing relationship required between the source and destination.. Table 2.2 summarizes the basic classification of A A L service classes. While class A supports constant bit rate (CBR) services with end-to-end timing and a connection-oriented mode, class B supports variable bit rate (VBR) service with end-to-end timing and a connection-oriented mode. Class C supports variable bit rate (VBR) service with no  Chapter 2 ATM-Based PCS and The ATM Protocol  20  timing required; however, it is still in connection-oriented mode. Class D supports variable bit rate (VBR) service with notimingrequired, and a connectionless mode.  A A L Service Access Point (AAL-SAP)  B-ISDN Layer A A L - P D U Primitives  B-ISDN Sublayer  Service Specific Convergence Sublayer (SSCS)  CS  Service Specific (SS)  AAL SSCS-PDU Primitives  Common Part (CP)  SAR  Common Part Convergence Sublayer (CPCS)  Segmentation and reassembly (SAR)  CPCS-PDU Primitives  Segmentation & Reassembly Sublayer  SAR-PDU Primitives  A T M Service Access Point (ATM-SAP)  Figure 2.11  A A L protocol sublayer model.  Table 2.2 B - I S D N / A T M service classes. Service Class Attribute  Class A  Timing relation between source and destination Bit rate  Class C  Class D Not required  Required Constant  Connection mode AAL  Class B  Variable Connection oriented  AAL  1  AAL 2  Connectionless A A L 3/4, A A L 5  A A L 3/4, A A L 5  AALs 1-5 were initially defined by the ITU-T to directly map onto the service classes. However, AALs 1-4 have drawn less attention from the communication industry compared to the  Chapter 2 ATM-Based PCS and The ATM Protocol  21  A A L 5 model, which supports variable bit rate traffic with either a connection-oriented or connectionless mode. A A L 1 was developed for CBR services with tight end-to-end delay constraints. The services of A A L 1 include high quality audio, video, and telephony. The standardization of A A L 2 has not been completed by the ITU-T. A A L 3/4 are defined for connection-oriented or connectionless VBR services with loose end-to-end delay constraints between source and destination. A A L 5 is proposed for variable bit rate services with loose delay constraints between source and destination. The CS-PDU format for A A L 5 is shown in Figure 2.12. The CS-PDU contains the user data field, a pad field to align the resulting PDU to fill an integral number of ATM cells, a control field, a length field, and a CRC field. The total overhead is 8 bytes for each CS-PDU.  CPCS-PDU CPCS-PDU Trailer CPCS-PDU Payload  PT  SAR-PDU Payload  5 bytes header '  PAD - Padding  1 - 65,535 bytes  PT  PAD  CPCS-U1J  CPI  Length  CRC  0-47  1  1  2  4  SAR-PDU Payload  48 bytes payload  field  Length - CPCS-PDU length  U U - User-to-user indication  CRC - Cyclic Redundancy check  CPI - Common part indicator  PT - 3 bit payload type AALJndicate  Figure 2.12 C P C S - P D U and S A R - P D U formats for the A A L type 5.  VY  SAR-PDU Payload  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture 3.1 Characteristics of the PCS Radio Cell Environment Code Division Multiple Access (CDMA) is a modulation and multiple access scheme based on spread spectrum techniques. This established technology has been applied to digital cellular communications and Personal Communication Services (PCS). CDMA has some advantages which will solve some problems associate with PCS design (eg., soft handoff and radio cell soft capacity). By employing CDMA, PCS will provide services in an economic and efficient manner. Research done by Qualcomm Communications Inc. shows that by employing sectorizated antennae in a CDMA radio cell, the capacity of a PCS radio cell can increase by a factor of 2.55, with 120° per sector and 3 sectors per radio cell [17]. The associated radio cell size reduction implies that a mobile terminal would experience more handoff events; this is because the mobile terminal is required to frequent handoff between sectors (within the same radio cell) or base sites. Increased handoffs may result in call processor overloaded or call dropping. A virtual connection tree architecture helps to reduce the call dropping during handoff by distributing handoff processing from the centralized call processor to multi-level databases.  3.2 Variable Bit Rate for Voice Encoding in CDMA Systems The new generation of voice encoder/decoder for CDMA has variable rate design. Twoway voice communication between the base station and the mobile terminal by using a dynamically variable data rate. The transmitting encoder takes voice samples and generates an encoded speech packet for transmission to the receiver. The receiving decoder decodes the speech packet into voice samples. In a typical full duplex, two-way voice conversation, the duty cycle from a  22  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  23  voice source is usually less than 40%. With the variable encoding technique, it is possible to reduce the transmission rate from 8 kbit/s during speech to 1 kbit/s when there is no speech, thereby reducing substantially the transmitting power and interference to other users during nonspeech times.  3.3 PCS Radio Cell Capacity In the cellular system with frequency reuse, co-channel interference is accepted but maintained below an established level, with the goal of increasing system capacity. CDMA is inherently less sensitive to co-channel interference than FDMA or TDMA [19]. In CDMA, frequency reuse efficiency is determined by the signal to interference ratio that results from all system users within a coverage region. The total capacity is typically quite large, and the collective statistics of all users are more important that those of a single user. In this case, the net interference to any given signal is the average of all the users' received power multiplied by the number of users. As long as the ratio of received signal power to the average noise power density is greater than a threshold value, the channel will provide an acceptable signal quality. The primary parameters that determine CDMA digital cellular system capacity are processing gain, bit energy, noise power spectral density, voice duty cycle, frequency reuse efficiency, and the sectorization factor [19]:  No In (3.1), parameters are defined as follows: C - Number of channels per radio cell  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  24  BW - Spread spectrum bandwidth (assume 2.5 MHz) R - Channel rate (assume 8 and 9.6 kbit/s for voice and data, respectively) D  Eb/No - Bit energy/noise power spectral density (assume 7.0 dB) Vrj - Voice duty cycle (assume 40%) F - Frequency reuse efficiency (assume 60%) S - Sectorization gain (assume 2.55) G  3.3.1 Sectorization Antenna Gain When sectorization antennas are used in a radio cell, the interference seen is simply divided by three (assume three 120° sectors per radio cell). The capacity supportable by the total system is thereby increased by approximately 2.55. According to (3.1), a base site with a 2.5 MHz bandwidth and sectorized antennas can support up to 75 voice channels and 50 data channels.  3.3.2 Unbalanced Traffic Loads for Cell Sites in Urban Areas In the general case, an equal distribution of mobile terminals and traffic load throughout the service area is assumed. In reality, unequal radio cell loading frequently occurs. In rush hour, particularly, a cell site will reach its peak traffic load and the system blocking probability may become excessive. With CDMA, problems created by rush hour may be less problematic because the number of channels available in a given radio cell depends upon the traffic loading in neighboring cells. If neighboring radio cells carry less traffic, then the capacity of a heavily loaded radio cell can be increased. CDMA peak traffic loading is somewhat flexible. To reiterate the capacity of an individual CDMA radio cell can increase by 10 ~ 50% depending on the actual distribution of mobile terminals [19]. If some radio cells carry more traffic than normal, some neighboring radio cells would normally carry less traffic. The more  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  25  lightly loaded radio cell would contributes less interference to the heavily loaded neighbor radio cell which therefore carry more traffic than the rated peak level. Figure 3.1 illustrates the unbalanced load scenario.  Figure 3.1  Unbalanced traffic load in rush hours.  3.4 CDMA Based Mobile Initiated Soft Hand-off Scheme In most analog cellular systems, handoff decisions are made by the mobile switching controller (MSC). The call is switched from an old radio cell to a one with a break time in between, using the "break before connect" procedure. This approach, which involves temporary call dropping, is not commensurate with user expectations, and may become worse in the case of PCS, because of the much smaller radio cell radius and associated increase in handoff traffic. In future  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  26  PCS deployment, degradation due to handoff problems should be minimized; therefore, a soft handoff scheme should be employed. Figure 3.2 shows the power control of the handoff signal between the base station and the mobile terminal. This soft handoff scheme allows both the original radio cell and a new radio cell to temporarily serve the call during the handoff transition. The soft handoff scheme not only reduces the dropped call probability but also maintains the quality of service to the user.  ATM switch  cell power control bits — Figure 3.2  Channel assignment during soft handoff.  After a call is initiated, the mobile terminal continues to scan the neighboring cells to determine if the signal from another radio cell is comparable to that of the original radio cell. This can be done by searching all the pilot tones transmitted from base stations and maintaining a list of all whose pilot signals are above a pre-set threshold. This list is transmitted to the nearest data base (HLR or VLR) and is updated to add or delete channels based on pilot signal strength. Addition of a new base station indicates to the mobile terminal that the call has entered a new radio cell coverage area and that a handoff can be anticipated. The mobile terminal transmits a  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  27  control message to the current base station which indicates that the stronger signal has been received from a new base station. An ATM switch establishes a new path to the mobile terminal through the new base station while maintaining the original path. During this period, the mobile terminal communicates to both base stations and combines both signals to increase the overall received signal-to-noise ratio.  3.5 ATM-Based Virtual Connection Tree Architecture to Support Soft Handoffs 3.5.1 Neighboring Mobile Access Region Future PCS deployments are being designed to support mobility regardless of user location. Also, various services including voice, data, and video with different QOS requirements are envisioned. To meet these various QOS requirements, a fast handoff scheme must be available in a microcell environment. In general, when a user places a call, the call processor will establish a link between the source and the destination nodes according to the traffic characteristics and the QOS. However, in a personal communication environment, mobile users change their access point frequently and a mobile user's call must be rerouted whenever a connection is handed over to a new base station. Handover traffic may cause call processors to become overloaded and bottlenecked [21]. Figure 3.3 shows a hierarchical interconnection network. Mobile terminals within a local access region can hand over to other neighboring base stations without call processor outside this local region. Every time a mobile connection is added to a connection tree, a collection of virtual circuit numbers (VCN) is assigned to the associated call. Each of these virtual connection numbers defines a path between the local ATM switch and a distinct base station within the  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  28  mobile access region. A mobile terminal selects its base station from among those in its connection tree with an appropriately assigned VCN. In this way, an ATM cell with a given tetherless connection will eventually flow through the appropriate node switch and to its destination. At the node switch, the currently used V C N is translated to VCI & VPI, needed to switch the ATM cell to the appropriate wire path.  Neighbouring mobile access region ,  Figure 3.3  I y  Hierarchical interconnection network.  In the reverse direction, the V C N of an ATM cell appearing at a node switch is appropriately translated such that the ATM cell flows over the correct branch of the tree. The call processor becomes involved only when handoff occurs between two nodes. Since one node switch can serve a large area, the frequency of call processor involvement remains low and handoff transition time is thereby reduced. The goal of the connection tree architecture is to reduce the handoff complexity for  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  29  wireline networks. To complete a mobile connection, a fixed virtual connection is created from the tree's node back to a wired network port. As shown in Figure 3.3, all base stations which belong to the same node are called neighboring mobile access points. When a user makes a call request, there are two steps for call setup. First, the fixed portion of the virtual connection is established between the root of the tree and the appropriate fixed point of the wired network. The fixed portion connection is then maintained so long as the mobile stays within the same mobile access region. Second, within the mobile access region, two sets of virtual connection numbers are assigned to that mobile connection, to enable forward and reverse traffic to be distinguished. Each ATM switch inside that region has its own routing table which is updated when a new VCN is assigned.  3.5.2 Fast Routing Scheme with Virtual Connection Number (VCN) Lookup Table In the microcellular environment, frequent mobile handoffs are expected. Handoffs would involve either two neighboring base stations connected into the same switch or alternatively, different switches in the same neighboring mobile access region. In either case, QOS must be maintained during the handoff transition period. Any new path established during handoff should provide'the same performance characteristics as the original path. Figure 3.4 illustrates the case where adjacent microcells are connected to the same switch (switch A). Table 3.1 shows that switches A and B both have their own lookup table even though the soft handoff transition involves switch A only. When a mobile terminal moves from radio cell 1 to cell 2, it enters the soft handoff region. The mobile terminal initiates a handoff signal to switch A through base station 1. When switch A receives this handoff request, it updates its V C N  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  30  lookup table. Prior to handoff transition, VC_5's traffic is routed into VC_2, and V C _ l ' s traffic into VC_6. During the handoff transition, the lookup table will be updated. A l l the ATM cells coming from VC_5 will be duplicated and routed into both VC_2 and VC_4. Assuming the traffic characteristics for both base stations are similar and the distances between switch A and both stations are short, the ATM cell delay variation between these two paths should be acceptable. In the reverse direction, ATM cells from VC_1 and VC_3 will be routed into output port 2 of switch A, but only one ATM cell will be selected for forwarding to switch B depending on the delay.  ^ [  Upper level switch B  To other connection trees  VC 6  VC_5  1  2  A T M switch A 3 4  BS  .  5 6  BS 2  1 Cell 1  Figure 3.4  >>  Cell 2  Two base stations connected to same switch.  When the signal strength received by the mobile from radio cell 1 falls below threshold level, the mobile informs switch A to disconnect VC_1 and VC_2. Switch A will update the lookup table accordingly. After the handoff transition, ATM cells from VC_5 will be routed into VC_4, and VC_3 will be routed into VC_6 in the reverse direction. Only the lookup table in  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  31  switch A is updated, and no call processor is involved. This scheme will facilitate congestion avoidance in the call processor. Table 3.1 Lookup table for two base stations connected to same switch.  Switch A  VCN VC_1 VC_5 VC_1 VG_3 VC_5 VC_5 VC_3 VC_5  In port 3 1 3 5 1 1 5 1  VC_6 VC_2 VC_6 VC_6 VC_2 VC_4 VC_6  ve_4  Outport 2 4 2 2 4 6 2 6  Before Handoff During Soft Handoff After handoli  Figure 3.5 shows a handoff transition in which two base stations are connected to two different switches. In this case, the handoff transition involves 3 switches, but only switch C is required to update the lookup table during the handoff transition. The delay variation of duplicated ATM cells may become significant in this case because switches A and B are geographically separated and may have different traffic characteristics.  32  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  ( VG_9 Other connection tree  1 2  VC 10  BS - Base Station  Node Switch C 3 4  V C - Virtual Connection  5 6  VC 1  VC 4  BS 1  Cell 2  Celll  Figure 3.5  Handoff involving two different switches.  Table 3.2 Switch/VCN assignments with two base stations connected to different switches.  Switch A Switch C Switch A Switch B  Switch C  Switch B Switch C  VCN VC_1 VC_5  In port 3 1  VC_9 VC_6  1 4  VC_3 VC_7 VC_6 VC_8  3 1 ' 4  VC_9 VC_9  1 1  VC_3 VC_7  6  VC_6 VC_2 VC_5 VC_10 Same VC_8 VC_4 VC_10 VC_10 VC_5  Outport 2 4  Before Handoff  3 2 2 4 2  During Handoff  2 3 5  3 1  VC_7 VC_8 VC_4  2 4  VC_8  6 .  VC_10  2  VC_9  1  VC_7  5  Alter 1 landoff  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  33  3.6 Mobile Initiated Handoff Signaling with Connection Tree Architecture In section 3.5, the virtual connection tree architecture supporting PCS was examined. In the following sections, the handoff signaling of the two cases considered in section 3.5.2 are discussed. We assume mobile-initiated handoff.  3.6.1 Handoff Signaling for Two Base Stations Connected to the Same Switch The signaling which occurs when neighboring base stations which are connected into a switch is illustrated in Figure 3.6. All neighboring radio cells' pilot channels are scanned by the mobile terminal and this resulting information is stored in base station 1 (BS1). This information is updated while the mobile terminal is communicating to BS1. When the mobile terminal has received a signal-to-noise ratio below the lower threshold point, it will initiate a handoff request to BS1. BS1 will send a handoff request to switch A, with the identity of the potential new base station following its receipt the handoff signal from the mobile terminal. Switch A will assign a new virtual connection number to the candidate base station (BS2). Also, switch A will inform BS2 to allocate a radio channel for handoff transition. After these procedures have been completed, radio channels in both BSland BS2 are synchronized. A combined signal from BS1 and BS2 is decoded by the mobile terminal. Since the handoff request has been sent to switch A, there is a timer activated in BS 1. If the first candidate base station (BS2) is busy or its queue is too long, call handoff initiation time becomes irregular and timerl in BS1 will expire. Under this situation, BS1 will send the second potential base site's identity to switch A, and call handoff initiation procedures will be started again. If call handoff initiation procedures are completed within the expected time, timerl in BS1 will be deactivated and switch A will start sending  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  Mobile  BS1  BS2  Handoff request  34  ATMSW  Timerl activated Handoff traffic request  Lookup table in BS1 tells BS2 is a possible candidate  Call handoff Initiation  Virtual connection setun arir New channel assignment  a  _V»rtual connection <0  •s  < a  M.,  c  k  T i l | |  Synchronization between BS 1. BS2. and activation of timer2 P  —^ ^ d T O t o l o n g o u l u e i n B s , o  A  S ^ J f n d j e c o * candidate urformation.  |  During handoff [Terminate  m  1 1 m > 1 t j |  Seau-Ltoterminate v i r t u ^ ^  «  request  I  v  c  ^  Time «DJreddueto mobuestation r«nainingm.handoffregion toojong,  J  Call termination  Lf rnrresnonding V C  Handoff signaling  Figure 3.6  User data flow  Alter due to unexpected delay  Handoff signaling for two base stations connected to same switch,  packets to both base stations. If the mobile terminal receives a signal-to-noise ratio from BS1 which is below the lower threshold point, the mobile terminal will send a connection termination request to BS1. This request will be transferred to switch A, and the lookup table will be updated. In some cases, the mobile terminal may stays in the handoff region too long. If the mobile terminal spends too much time in the handoff region, timer 2 in BS1 will expire, the connection between the mobile terminal (MT) and BS1 will be terminated, and only BS2 will communicate with the mobile terminal.  3.6.2 Two Base Stations Connected to Different Switches When base stations involved in a handoffare connected to different switches, a longer  35  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  Mobile  BS1  SW A  BS2 _BS1 sends rossihl. ^L-'-.tr  SW B  SW C  information to SW A  to £  i1  I  p  a  V'^  c k e t d e r < ^ ^ f ^  r send  Syn requesf"  o v n reouSJ  Time expired due to unexpj :cted delay in SW_B or BS2  _Sew„d^andJda Js m | ^ a t J o 1 will wjj, be send te  Terminate communication request  n  ~  lilMconnect, V C termination Ack  Handoff Signaling  Figure 3.7  Alter due to unexpected delay  User data flow  Handoff signaling involves two switches,  handoff time as well as more complex signaling procedures are required. While the mobile terminal is registered in BS1, it scans all the possible candidates and stores this information in BS1. When switch A receives a handoff request from BS1, the location of BS2 can be found in switch_A's data base. Switch A will send a handoff request to the upper level switch or switches where two connection trees are commonly joined (switch C in Figure 3.5). Switch C will send a request for a new virtual connection to the lower level switch (switch B in Figure 3.5) which is closest to BS2. After switch B receives the VC request from switch C, it will send a request for  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  36  new radio channel assignment to BS2. Also, it is required to send back a VC establishment Ack signal to switch C, which will then update its lookup table according to the information received from both switches A and B. For the forward path, incoming ATM cells will be duplicated in switch C and send to switches A and B according to the lookup table. On the reverse path, ATM cells from BS1 and BS2 are sequenced. A lower delay ATM cell from a pair of ATM cells will be selected at switch C. When BSl's signal-to-noise ratio falls below the lower threshold level, a call termination request will be sent to BS 1 from the mobile terminal.  3.7 Base Station Overload Probability Consider a connection tree where the radio cell air interface is the major network capacity bottleneck. It is important to know the probability that a radio cell will become overloaded. Under overload, QOS can no longer guarantee to either current or new users. By assuming that any mobile user is equally likely to be found within any radio cell; the complexity of computing the overload probability can be made manageable. Consider that there are a maximum of N mobile calls in a connection tree and that each mobile is equally likely to be in communication with any particular test base station within a local region. The probability P that there are / connections ;  established with any base station is as follows:  (3.2)  (3.3)  Where  (3.4)  37  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  1 Since B is the number of base stations within the same connection tree, - is the probability that a given mobile is communicating with a given base station. If each base station can support up to m virtual connections without effecting the QOS, then its overload probability P can 0  be calculated as following:  (3.5)  As discussed in section 3.3, a radio cell with a 2.5 MHz bandwidth can support 75 voice channels and 50 data channels without affecting the QOS. Assuming that data channels are fully occupied, a simulation was completed for voice channels with different numbers of base stations as members of a connection tree. A similar idea was used to simulate data channels with varying numbers of base stations in a local region. Figure 3.8 shows the base station overload probability vs. the maximum number of voice connections with 10, 15 or 20 connected base stations. With 10 base stations, the overload probability increases significantly compared to the other two curves. With 20 base stations, the overload probability is 10" when 972 voice channels are admitted to a connection tree. With 10 3  base stations connected, only 539 connections can be admitted if we want to keep the overload probability at 10 or less. In either case, the number of voice calls actually carried is less than 75 per base station.  Chapter 3  CDMA Radio Cell Environment and Virtual Connection Tree Architecture  38  10 Each BS has 75 voice connections. Number of base stations per connection tree  B = 20 B = 15 B = 10  •s o  S3 PQ  0  100  200  300  400  500  600  700  800  900 1000 1100 1200 1300 1400  Maximum number of voice connections Figure 3.8  Base station overload probability plotted against number of voice connections.  Figure 3.9 shows the base station overload probability vs. the number of data channels in a connection tree region. This simulation is based on 50 data channels per base station, with 10, 15 or 20 base stations in a connection tree. The overload probability shown in this figure represents how a mobile data user will affect the new radio cell's QOS under handoff from the current radio cell to the new one. The results indicate that for a 20 base stations region, 733 data calls can be admitted with a 0.01 overload probability. In this case, the connection tree admits 267 data calls less than its actual capacity.  Chapter 3 CDMA Radio Cell Environment and Virtual Connection Tree Architecture  39  o  0  100  200  300  400  500  600  700  800  900  1000  Maximum number of data connections Figure 3.9  Base station overload probability plotted against number of data connections.  Table 3.3 Results for base station overload probability. Maximum number of connections (approximation values) Overload probability  Base station = 10  Base station =15  Base station = 20  0.01 (voice channel)  539  800  1067  0.001 (voice channel)  489  728  972  0.01 (data channel)  367  550  733  0.001 (data channel)  325  483  650  Chapter 4 Network Coverage, Modeling and Simulation Tool (OPNET) Specifications 4.1 Radio Cell Coverage 4.1.1 Number of Potential Users in a Connection Tree Region Microcells are expected to play a major role in the next generation of wireless networks, including those which support PCS. In order to study the traffic characteristics of the microcell environment, the number of potential PCS users in a radio cell and its coverage area have to be determined. In section 3.7, radio cell capacity in a connection tree architecture was discussed. Table 3.3 shows the maximum number of voice and data channels supportable by a connection tree without having a noticeable negative effect on the QOS. The number of voice and data channels calculated in Table 3.3 provide a reasonable accurate estimate of radio cell capacity, because mobile terminals' handoff traffic has been included. The total number of potential voice and data users in a connection tree can be calculated using the Erlang B formula. Assuming each voice user makes two calls per hour of two minute's duration, and that each data user transmits 0.1 messages per second with 10 kbits per message, the number of potential users in a connection tree is calculated. The air interface transmission rates for voice and data are 8 kbps and 9.6 kbps, respectively.  Table 4.1 shows the number of potential mobile users in a connection tree region with a 0.01 air interface blocking probability. Results for both 0.01 and 0.001 overload probabilities were obtained. The 0.01 overload probability can support more mobile users with fairly high user QOS guaranteed. A connection tree with 15 base stations can support up to 12,863 voice users and 5,026 data users with 0.01 air interface blocking and overload probabilities. Table 4.1 shows  40  Chapter 4 Network Coverage, Modeling and Simulation Tool (OPNET) Specifications  41  that a connection tree can support almost 2.6 times more voice users than data users. Table 4.1 Number of potential voice and data users in a connection tree. Number of potential users with 0.01 air interface blocking probability Overload Probability  Base station = 10  Base station = 15  Base station = 20  0.01 (voice channel)  8550  12863  17302  0.001 (voice channel)  7720  11670  15722  0.01 (data channel)  3291  5026  6768  0.001 (dalit channel)  2895  4389  5976  4.1.2 Radio Cell Coverage in Metropolitan, Urban, and Suburban Areas The objective of microcellular networks is to increase the frequency reuse factor by reducing the radio cell radius. This significant gain in spectrum reuse accommodates more users in metropolitan centers such as shopping centers, railway stations, or downtown street blocks. These locations have similar mobile traffic characteristics, including a relatively large mobile user density in a radio cell and relatively small occurrence of handoff transitions (pedestrians). Therefore, a radio cell with a small radius is more suitable for use in metropolitan areas. Urban and suburban areas would have a larger radio cell size. Suburban area cells are usually placed along the highway; a larger cell size allows more handoff transition time for vehicles, which reduces call dropping. Previous research has estimated the number of potential users in different environments [25]. Table 4.2 provides a user density summary.  Chapter 4 Network Coverage, Modeling and Simulation Tool (OPNET) Specifications  42 '  Table 4.2 Number of potential users in different radio environments. Potential Users in K m  2  Environments  Metropolitan area  Urban area  Suburban ;uva  Pedestrians  13,000  650  15  Private tars  1000  400  65  Public transport passengers  2,600  260  15  Total  16,600  1310  95  Assuming 15 base stations per connection tree, the total number of voice and data users supported by a connection tree is 17,889. If the same connection tree is used in a metropolitan area with 75 percent market penetration, it can cover up to 1.4 Km . In urban and suburban areas 2  the coverage would be 18 and 256 Km , respectively, with 75 percent market penetration. Previous research shows that overlapping cell coverage in a microcellular environment affects handoff performance [26]. A microcell requires more than a 50 percent overlap with adjacent cells in order to maintain a reasonable quality of reception in handoff regions [26]. Based on a 50 percent overlapping area, a radio cell's radius in different environments is investigated. A radio cell model with 50 percent inner and outer areas is shown in Figure 4.1. The microcell radius in this case is determined by the distance between the cell centre and the cell boundary where the signal strength drops below -90 dBm [27]. The hexagon radio cell scheme is used to calculate the cell coverage area and cell radius. Assuming the inner circle area is half of the outer circle area and the hexagon radio cell distance x is placed on the midway between the inner and outer radius, the relationship between R, r, and x can be described as follows:.  Chapter 4 Network Coverage, Modeling and Simulation Tool (OPNET) Specifications  43  R = Outer radius of a radio cell. r = Inner radius of a radio cell. x = Distance from centre to hexagon cell edge.  Figure 4.1  Fifty percent overlapping microcell environment.  R = j2-r  (4.1)  2x = R + r  (4.2)  Table 4.3 shows the radio cell radii in three different environments and different levels of market penetration. These results are based on a connection tree with 15 base stations and a 50 percent radio cell overlap, with market penetrations of 30, 50, and 75 percent. The smallest radio cell radius (192 meters) occurs when a connection tree supports 75 percent market penetration in a metropolitan environment. The cell radius increases to 356 meters when the market penetration decreases to 30 percent. In general, a microcell's radius is smaller than 1 km [25]; therefore, urban area radio cells are either microcells or macrocells, depending on market penetration. Suburban areas have fewer potential users and employ macrocells. With 30 percent market penetration in a suburban area 4701 m radio cells are proposed. These would be primarily along the highway and  Chapter 4 Network Coverage, Modeling and Simulation Tool (OPNET) Specifications  44  would provide services mainly to vehicle users. Table 4.3 Radio cell radii in different environments. Radio cell radius in meters Environmenl  75  market penetration  50 % market penetration  M)  market penetration  Metropolitan area  192  275  356  I Jrhan area  800  981  1266  Suburban area  2973  3642  4701  4.2 Network Simulation Model The network simulation model defined in Figure 4.2 was used to investigate the multimedia traffic characteristics and interworking performance for a city similar in size to Greater Vancouver. A network architecture with a star topology is employed. The star topology architecture is similar to that of the local telephone network. By using the existing telephone network resources, the initial setup cost for a local ATM-based PCS network could be reduced. The proposed local ATM network is interconnected with 155.52 Mbps transmission links (SONET STS-3). Each node in Figure 4.2 represents a local region with a number of connection trees. The coverage area of a single node depends on market penetration and percent of intra-node traffic. Assuming localized traffic characteristics, most of the traffic in node 0 ~ 4 remains in the gateway switch_0 region, with similar traffic localization in the other 3 regions. The transmission link between gateway switch_0 and the central switch is not easily congested unless the network capacity is reached. Simulation results for the delay and throughput of this network will be presented in chapters 5 and 6.  Chapter 4 Network Coverage, Modeling and Simulation Tool (OPNET) Specifications  '  '  \  ^Jj  Node A T M Switch  Figure 4.2  -*  *~  155.52 Mbits/s link  Simulation model  Network simulations are based on the following assumptions: 1. The transmission rate on each link is 155.52 Mbps. 2. The network carries voice and data traffic. 3. There are equal numbers of voice and data connections in each node. 4. Traffic is equally distributed in each node.  Chapter 4 Network Coverage, Modeling and Simulation Tool (OPNET) Specifications  46  5. The source rates for voice and data are 8kbps and 9.6kbps, respectively. 6. Inter-arrival time of data traffic at each node is exponentially distributed, with peak cell rate equal to 5 times that of the average rate. 7. Data packet size is 9920 kbps. 8. Priority is given to ATM voice cells. 9. The buffer size of the local switch, gateway switch, and central switch are all different. 10. Intra-network traffic has its origin and destination confined to one neighborhood region.  4.3 Simulation Tool (OPNET) Specifications OPNET is a Unix-based software package capable of simulating large communication networks. It is a hierarchical object-oriented modeling system with three levels used in defining a network model. The network model defines the topology location of the network nodes. The node model defines the internal structure of a node. The process model defines the communication processes which run in the various nodes. OPNET offers various tools and editors to accomplish a complex network modeling. The network editor is used to create a network model which consists of communication nodes connected together by a duplex or by point-to-point links. The node editor is used to create a node model which specifies the internal structure of a communication node. A communication node is. defined by connecting blocks called modules with packet streams. These connections allow for the exchange of data packets between modules. The process editor is used to create the process models that run in node modules. The process model is  Chapter 4 Network Coverage, Modeling and Simulation Tool (OPNET) Specifications  47  represented in both graphical and textual format. The graphical structure is in the form of finite state machine diagrams, composed of states and transitions. The textual specifications describe the precise behavior of the process model and are coded in C programming language. The simulation tool is used to execute simulations. A number of simulations can be executed simultaneously with different parameters and simulation times. The analysis tool is used to evaluate the simulation results. Specified statistics can be collected by using the probe editor. The results of an output file are displayed graphically in the analysis tool. The ATM process and node modules are fully developed according to the ITU/CCITT recommendations. A standard ATM switch in OPNET consists of an AAL, ATM management, an ATM layer, an ATM translation, and ATM switch node modules. The function of an A A L node module is to segment the upper layer PDU into A A L PDU and to reassemble A A L PDU into upper layer PDU. The segmentation and reassembly rate (SAR: ATM cells/sec) can be specified in the A A L node module, which consists of three processes. The first is the ams_aal_disp, which responds to the VCC setup and release. The second is the ams_saal, which responds to the A A L signaling. This process communicates to both the upper layer (the traffic generators in this case) and the destination A A L to ensure that the VCC is set up properly. The third process in the A A L node module is the ams_aal5_conn, which was developed based on aal_5 and is responsible for the transfer of A A L PDU between the A A L and the ATM layer. There are two sets of timer in the A A L node modules. Thefirsttimer tracks the initial VCC setup time; if a VCC cannot set up within the time frame, a negative acknowledgment is sent to the traffic source. The second timer tracks the idle time of a connected VCC which will be released if the idle time period is longer than the timer duration. Both timers' durations can be  Chapter 4 Network Coverage, Modeling and Simulation Tool (OPNET) Specifications  48  adjusted in the ams_aal_interfaces.h file. The function of the ATM management (ams_atm_management) module is to handle the routing of messages. This process also creates the call signaling process. When a request to start a new VCC arrives from the A A L module, the process ams_atm_call is created. When an ATM setup message is received, the process destination (ams_atm_call_dest) or intermediate switch (ams_atm_call_net) are created depending on whether this node is the final destination or an intermediate node for the call. The ATM layer (ams_atm_layer) module serves as an interface between the A A L module and the rest of the ATM network. An integer-type attribute specifies the ATM address for each ATM layer module. This value must be unique for all ATM layer modules in the network model. For all intermediate switches, the ATM layer module address should be set to -1, which denotes a network switch. The ATM translation (ams_atm_trans) module translates the incoming VPI and VCI values of the ATM cells received by the switch into the outgoing VPI and VCI values. This process module expects one of two cell types, either a control cell for ATM signaling and routing or an ATM data cell. If a control cell arrives at the translation module, it is sent to the ATM management module with a processing delay of 0.1 nano seconds. An ATM data cell is sent to the ATM layer with a processing delay of 0.1 nano seconds if this node is the final data cell destination. Otherwise, it is sent to the ATM switch module with a processing delay of 0.01 nano seconds for delivery to other network nodes. The ATM switch module receives ATM cells from either the ATM layer or the ATM  Chapter 4 Network Coverage, Modeling and Simulation Tool (OPNET) Specifications  49  translation modules. This process performs the switching function, enqueues the cells in an output port buffer, and sends them to the transmitter. All the ATM cells received from the input port are enqueued regardless of the transmission data rate. An ATM cell is dequeued from the output port if the transmitter is available. Each output port has a buffer for each QOS class. There are two schemes for dequeueing ATM cells from the output buffer, namely the round-robin and priority_x methods. For the round-robin method, all QOS classes have equal priority for dequeueing. For the priority_x (A, B, C, D) method, only the specified priority QOS class' buffer will dequeue; the lower priority QOS classes' buffers will dequeue when the priority_x's queue is empty.  Chapter 5 Simulation Results for Voice and Data Traffic 5.1 Simulation Results for Data Traffic Over an ATM Network Results in this section are based on an ATM network that carries data traffic only. The objective here is to study the impact of the upper layer data packet length on ATM network performance. The SAR rate is set to 40,500 cells/sec for each node in the simulation model. Sufficient data buffers are assigned in each ATM switch, which results in no ATM cell loss. The amount of ATM cell loss directly affects upper layer protocol performance. For example, consider a TCP data packet with 4,984 bytes being sent over an ATM network. This TCP packet is segmented into 104 ATM cells when it arrives at the A A L layer. During the transmission, loss in any of these 104 ATM cells would result in the TCP data packet not being reconstructed in the upper layer. Cell retransmission is excluded in the standard ATM protocol; thus the entire TCP data packet would have to be retransmitted.  5.1.1 Delay and Buffer Occupancy Analysis of Data Traffic Data packet sizes of 1240, 2488, and 4984 bytes are used for these simulations. End-toend delays, data buffer occupancies, and various levels of inter- and intra-network traffic are studied. Figure 5.1 shows the node_0 to node_15 mean data packet delay with 25 percent intranetwork traffic. The total traffic on the abscissa represents the sum of total traffic generated by all 16 source nodes. As shown in Figure 5.1, end-to-end delays are independent of packet size when the traffic level is well below the network capacity. They become more distinguishable when the total traffic reaches the network limit. For a data packet size of 1240 bytes, the end-to-end delay increases significantly when the total traffic is beyond 700 Mbps. The network can support much less traffic with data packet sizes of 2488 and 4984. The reason a larger packet size results in a  50  Chapter 5 Simulation Results for Voice and Data Traffic  51  smaller network capacity is that larger packet size increases the data traffic burstiness. This effect may result in a higher buffer occupancy over a short period of time and a significant increase in end-to-end delay.  at  •a  o  a •o o  Z 100  200  300  . 400  500  600  900  1000  Total traffic (Mbps) Figure 5.1  Node_0 to node_15 mean data cell delay.  The observation that a larger packet size may introduce higher buffer occupancy is actually demonstrated in Figures 5.2 and 5.3. In Figure 5.2, the mean buffer occupancy of the central switch is plotted for three different packet sizes. Notice that the increasing slopes of these three buffers are shaped differently. When the total traffic is equal to 700 Mbps, a packet size of 4984 bytes requires almost six times more buffers than a packet size of 1240 bytes.  Chapter 5 Simulation Results for Voice and Data Traffic  0  100  200  300  400  500  600  700  800  900  1000  700  800  900  1000  Total traffic (Mbps) Figure 5.2  0  Central switch mean buffer occupancy vs. total traffic.  100  200  300  400  500  600  Total traffic (Mbps)  Figure 5.3  Gateway switches 0 and 2 mean buffer occupancy.  Chapter 5 Simulation Results for Voice and Data Traffic  53  Figure 5.3 shows the buffer occupancy for gateway switches 0 and 2. Since the network configuration is symmetrical with equally distributed traffic, the buffer occupancies of these two switches are expected to be similar. As shown in Figure 5.3, each dashed line matches to its corresponding solid line. This indicates that switches in the network are functioning properly and that the simulation results are accurate. Mean data buffer occupancies indicate the data queue size's behavior; however, the maximum buffer occupancy values are more important when cell loss ratio in an ATM network is considered. These maximum buffer occupancies are reported in Table 5.1. These results represent the maximum values shown on each of the simulation curves in Figures 5.1, 5.2 and 5.3. Table 5.1  Maximum buffer occupancy in central and gateway switches. Maximum buffer occupancy (ATM cells)  Packet si/e (bytes)  Central swiich  Gateway switch 0  Gateway switch 2  1240  3347  1895  2057  24SK  4471  2187  2217  4984  4501  3178  3265  With a 25 percent intra-network traffic, the bottleneck of this simulation model occurs at the transmission links between the gateway and central switches. Therefore, delay between these paths accounts for much of the delay shown in Figure 5.1. Another important parameter by which to measure congestion in the network is the cell delay variation (CDV). This parameter denotes the delay difference for each ATM cell received by a destination node. As shown in Figure 5.4, CDV increases with network loading in a similar way for each packet size. 3.162 ms of CDV is recorded when a 4984 bytes packet size is used. An examination of CDV values for packet sizes of 1240, 2488, and 4984 bytes indicates that CDV is  Chapter 5 Simulation Results for Voice and Data Traffic  54  nearly 100 times greater for the 4984 byte packets than for those with 1240 bytes.  -5  0  100  200  300  400  500  • 600  700  800  900  1000  Total traffic (Mbps) Figure 5.4  Node_0 to node_15 cell delay variance vs. total traffic.  5.1.2 Delay Analysis of Intra and Inter-network Data Traffic Section 5.1.1 has presented performance analysis for a star topology network with 25 percent intra-network traffic using various data packet lengths. Results show that the network can support up to 700 Mbps when a 1240 bytes data packet size and a 25 percent level of intranetwork traffic are used. Since congestion occurs in the transmission links between the gateway and central switch, a higher percentage of intra-network traffic can increase the overall network capacity. In this section, the network performance for 25, 50, and 75 percent intra-network traffic levels using a 1240 bytes data packet size is reported. Figure 5.5 shows the end-to-end delay from node 0 to node 15. The dashed lines in this  Chapter 5 Simulation Results for Voice and Data Traffic  55  figure represent the ATM cell delay. This delay is measured by the amount of time it takes for an ATM cell to be sent from the source node's ATM layer to the destination node's ATM layer. The ATM delay includes switching delay, propagation delay, and queueing delay. Segmentation and reassembly delays are excluded.  0  200  400  600  800  1000  1200  1400  1600  1800 2000  2200  2400  Total traffic (Mbps) Figure 5.5  Node 0 to node 15 mean delay with various intra-network traffic.  The solid lines in Figure 5.5 represent the data packet (AAL) delay from node 0 to node 15. The data packet delay is measured by the beginning of segmentation at the source node's A A L and to reassembly at the destination node's AAL. The difference between ATM cell delay and data packet delay is the time required for the segmentation or reconstruction of a data packet. Delays between the upper layer and the A A L are not included in these simulations. The total network traffic increases from 700 Mbps to 2100 Mbps when intra-network traffic percentage increases from 25 to 75. The total traffic increases by only 40 percent when intra-network traffic  Chapter 5 Simulation Results for Voice and Data Traffic  56  increases from 25 to 50 percent; however, the increase exceeds 100 percent (from 1000 Mbps to 2100 Mbps) when intra-network traffic increases from 50 to 75 percent.  0.5  Data packet size 1240 bytes  t °- r 4  0>  •O  0  0  75% intra traffic 50% intra traffic 25% intra traffic  & t©-  200  800  400  600  1000  1200  1400  1600  1800 2000  2200  2400  Total traffic (Mbps) Figure 5.6  Node 0 to node 1 mean data packet delay with various intra-network traffic.  Figure 5.6 shows the intra-network delay between node 0 and node 1 with various levels of intra-network traffic. It is expected that no significant changes in intra-network delay will occur with different percentages of intra-network traffic. Simulations show that the intra-network delay for a data packet size of 1240 bytes varies from 0.1 ms to 0.3 ms depending on the traffic load. The simulation curve for the 25 percent intra-network traffic level stops around 800 Mbps because transmission links between the gateway switch and the central switch become congested. A similar situation occurs for the 50 percent intra-network traffic level. Figure 5.7 shows the data packet delay from gateway switch 0 to nodes 2 and 3. Simula-  Chapter 5 Simulation Results for Voice and Data Traffic  57  tion results show that delays between nodes 2 and 3 are similar at various intra-network traffic levels. The reason for investigating the delays between these two nodes is that both may involved in ATM cell duplication when the mobile terminal is handed off softly between nodes 2 and 3. For example consider that two neighboring cell sites are connected to different nodes, one to node 2 and another to node 3. An ATM cell must be duplicated in gateway switch 0 when the caller or callee is handed off from node 2 to node 3. If delays between these two nodes are significantly different, the mobile terminal may not be able to combine the signal from both handoff base stations.  0  200  400  600  800  1000  1200  1400  1600  1800 2000 2200 2400  Total traffic (Mbps) Figure 5.7  Mean data packet delay from gateway switch 0 to nodes 2 and 3.  58  Chapter 5 Simulation Results for Voice and Data Traffic  Total traffic (Mbps) Figure 5.8  0  Central switch mean buffer occupancy.  200  400  600  800  1000  1200  1400  1600  Total traffic (Mbps) Figure 5.9  Gateway switches 2 and 3 mean buffer occupancy.  1800  2000  2200  2400  Chapter 5 Simulation Results for Voice and Data Traffic  59  As seen in Figure 5.8, central buffer occupancies at various levels of intra-network traffic are similar when the total traffic is below 400 Mbps. The buffer is filled much more rapidly at the 25 percent intra-network traffic level when the simulation exceeds 400 Mbps. This is because more traffic is being transmitted to the other regions compared to the 50 and 75 percent levels of intra-network traffic. In Figure 5.9, solid and dashed lines represent the mean buffer occupancies for gateway switches 2 and 3, respectively. Since traffic is equally distributed in each node, the rates of buffer growth in these two switches are expected to be similar. Simulations confirm this similarity. Buffer occupancies have been confined to relatively low values because simulation at higher values requires excessive workstation memory. Cell delay variation between node 0 and node 15 is plotted in Figure 5.10. The solid lines and dashed lines in thisfigurerepresent the delay variation of the data packet and data ATM cell, respectively. Notice that the data ATM cell delay variation is somewhat smaller than the data packet delay variation at low throughput values. However, these two curves merge when the system becomes congested with traffic. At low traffic level, the major delay that occurs is from the segmentation and reassembly procedures; therefore, data packet delay variation is distinguishable from data ATM cell delay variation. At higher traffic levels, the major delay is from the queuing and dequeuing of ATM cells. Since both ATM cell and data packet delays include queueing delay, these two curves are expected to join when queueing delay dominates.  60  Chapter 5 Simulation Results for Voice and Data Traffic  -6  0  200  400  600  800  1000  1200  1400  1600  1800 2000  2200  2400  Total traffic (Mbps) Figure 5.10 Node 0 to node 15 cell delay variance.  5.2 Simulation Results for Mixed Voice and Data Traffic PCNs will carry both voice and data traffic. CDMA is one of the technologies proposed for the radio cell air interface. The proposed coding rate for voice traffic is 8 kbps, with a constant bit rate [21]. A data channel with a 9.6 kbps transmission rate has been proposed by Qualcomm [19]. In this section, the performance of voice and data traffic with various parameters will be examined. Simulations are based on various voice coding rates and levels of intra-network traffic. Voice traffic is assigned to the highest priority in the network and no data ATM cell is transmitted unless the transmitter voice queue is empty. A dynamic bandwidth allocation scheme is used for voice and data traffic in all VP links. Although voice traffic has the highest priority, the data traffic can occupy any unused voice bandwidth. As with pervious simulations, adequate buffer size is assigned to data traffic; however,-a relatively small buffer queue size is assigned to the voice  Chapter 5 Simulation Results for Voice and Data Traffic  61  traffic because of its high priority for service.  5.2.1 Delay Analysis of Voice and Data TVaffic with Variable Data Traffic Loading ,  Number of voice channels per node Figure 5.11 . Node 0 to node 15 data packet delay with various data channels per node.  Figure 5.11 shows the node 0 to node 15 mean data packet delay with various numbers of data channels per node. With 1120 data channels per node, the end-to-end delay rapidly increases when simulation reaches 4000 voice channels per node. When the number of voice channels increases from 4000 to 5000, data packets with 1120 data channels per node will experience more than a 20 ms queueing delay. For traffic with 2000 data channels per node, data packet delay grows steeply when the network reaches 3300 voice channels per node. This is because the data queue congests much faster when more data traffic is present in the network.  Chapter 5 Simulation Results for Voice and Data Traffic  70  I  I  I  8kbps for voice channel 9.6kbps for data channel  •a  60  62  I—  number of data channels/node 2950 — A — 2000 1 1120 —©—  'o S 50 c  o (0  o  z 40  500  1000  1500  2000  2500  3000  J.  3500  4000  J_  4500  5000  5500  Voice channels per node Figure 5.12 Node 0 to node 15 voice packet delay with various data channels per node.  Figure 5.12 shows the end-to-end voice packet delay between node 0 and node 15. Simulation results show that voice traffic has similar delays regardless of the data traffic loading in the network. The voice traffic delay varies from 50 to 60 us when the number of voice channels per node increases from 500 to 5000. Voice traffic should not be affected by other traffic in the network since voice traffic has the highest priority in the network. Voice intra-network traffic delays are shown in Figure 5.13. Results show that voice traffic delay varies from 28 to 30 us when the voice traffic is transmitted from node 0 to node 1. Voice traffic delay increases by less than 2 us when the total number of voice channels per node increases tenfold. Voice traffic delay for nodes 2 and 3 from gateway switch 0 is plotted in the samefigure.One sees that voice traffic experiences a similar delay in the same node, regardless of the amount of data traffic in the network. Note in the bottom part of the figure that voice traffic  63  Chapter 5 Simulation Results for Voice and Data Traffic  delays from gateway switch 0 to nodes 2 arid 3 are similar, except in the case where there are 2950 data channels per node. This problem may be caused by the switching delay in the input port of the switch because a switch may require more switching time when the network carries heavy burst traffic.  30 29 28 27 26  -  25  -  •8  24  -  D O  23  -  a o.  'S >  8kbps for voice channel 9.6kbps for data channel number of data channels/node n_0 to n_l g_0 to n_2 2950 — B - - D - 2000 A 1120  22 21  B —  20  a-—  - i  —  g_0 to n_3  •  o A  .e--&--Ba •A-  •6—e~©«r  -^-i  19 18  0  500  1000  1500  ± 2000  2500  3000  3500  4000  4500  5000  5500  Number of voice channels per node Figure 5.13 Mean voice packet delay with various data channels per node.  Data intra-network traffic delays are plotted in Figure 5.14. Delay for data traffic varies more than for voice traffic. Figure 5.15 shows the data ATM cell delay variation between node 0 and node 15. The characteristics of delay variation curves are similar to those for end-to-end delay; however, delay variation values provide additional information about network congestion. Data buffer occupancies are shown in Figures 5.16 and 5.17 for various number of voice  64  Chapter 5 Simulation Results for Voice and Data Traffic  and data channels. These curves show data buffer occupancies with respect to number of voice channels per node. The trade-off between the number of voice and data channels is evident. When the number of data channels increases from 1120 to 2000, data buffers are filled much more rapidly for a fixed voice channel value. In particular, the last two simulation points of the case where there are 2950 data channels per node show that the buffer size increases almost four times with only a small change in the number of voice channels per node.  150 140 130 120 3  110  CX  100  •tao I  8kbps for voice channel 9.6kbps for data channel number of data channels/node n_0 to n_l g_0 to n_2 g_0 to n_3 2950 — B — '--a-- . • 2000 —e— --o--G 1120 —A—A 500  1000  1500  2000  2500  3000  3500  4000  Number of voice channels per node Figure 5.14 Mean data packet delay with various data channels per node.  4500  5000  5500  Chapter 5 Simulation Results for Voice and Data Traffic  -3  Number of voice channels per node Figure 5.15 Node 0 to node 15 data A T M cell delay variance with various data channels per node.  Voice channels per node Figure 5.16 Central switch buffer occupancy.  Chapter 5 Simulation Results for Voice and Data Traffic  66  Voice channels per node Figure 5.17  Gateway switches 0 and 2 buffer occupancy.  5.2.2 Delay Analysis of Voice and Data Traffic with Different Voice Coding Rates The voice coding rate affects the data traffic delay and the number of voice and data calls in the network. In this section, the investigation is based on a voice coding rate of 8, 16, and 32 kbps. Data traffic delay and network capacity are examined. As in previous simulations, voice calls remain active throughout the simulation, and 9.6 kbps is used as the data channel transmission rate. Figures 5.18 to 5.21 show the data traffic delay and buffer occupancy measured at different nodes. In Figure 5.18, data traffic is reduced almost by half when the voice coding rate increases from 8 kbps to 16 kbps. An increase in the voice coding rate also affects the intranetwork traffic delay. Data traffic delay from node 0 to node 1 (n_0 to n_l) increases almost 15 us when the voice coding rate changes from 8 kbps to 16 kbps.  Chapter 5 Simulation Results for Voice and Data Traffic  2U  T  T  18 h  7950 vo*ce channels 9.6 kbps for data channel  16  Voice coding rate 32 kbps — A — 16 kbps 1 8 kbps —©—  14  I •3  T  12 10  !0  S  8  o  6  c  •8 o Z  4 2 0  0  500  1000  1500  2000  X  2500  3000  3500  4000  Number of data channels per node Figure 5.18 Node 0 to node 15 mean data packet delay with various voice coding rates  03  7950 voice channels 9.6kbps for data channel Voice coding rate n_0 to n_l g_0 to n_2 32 kbps — B — Q 16 kbps —e— o 8 kbps A A-----  T3  to  1000  1500  2000  2500  3000  3500  4000  Number of data channels per node Figure 5.19 Mean data packet delay with various voice coding rates.  4500  0 ton 3  5000  5500  Chapter 5 Simulation Results for Voice and Data Traffic  68  Figures 5.20 and 5.21 show the buffer size required for data traffic with three different voice coding rates. The data buffer for a voice coding rate of 8 kbps fills much more slowly than for a voice coding of 32 kbps. The rapid change in buffer occupancy implies that the data packets may experience an associated higher cell delay variation.  1UUU  U  900 800  h  o  700 o  600  "a  500  '•3  400  £  300  3  1950 voice channels 9.6kbps for data channel Voice coding rate 32 kbps —& 16 kbps 1 8 kbps — e —  200 a  U  100 n  0  I  500  •  i  1000  1  —n  1500  it  2000  1  2500  i  3000  Data channels per node Figure 5.20 Central switch buffer occupancy with various voice coding rates.  i  3500  I  4000  Chapter 5 Simulation Results for Voice and Data Traffic  0  500  1000  1500  69  2000  2500  3000  3500  4000  Data channels per node Figure 5.21  Mean data buffer occupancy with various voice coding rates.  5.2.3 Delay Analysis of Inter- and Intra-Network Voice and Data Traffic We have extended our investigation into voice and data traffic by considering the relationship between the intra-network traffic and the capacity of a star topology network. We assume that voice and data traffic each has same percentage of intra-network traffic. The data traffic delay and network capacity with various levels of intra-network traffic are shown in Figures 5.22 to 5.24. The number of voice channels per node and the voice coding rate are fixed throughout these simulations. Results demonstrate that the number of data channels accommodated per node increases by eight times when the intra-network traffic level changes from 50 to 75 percent. However, the number of data channels increases only slightly when the level of intra-network traffic changes from 25 to 50 percent. Data traffic delays from node 0 to node 1 (n_0 to n_l) and gateway 0 to node 2 (g_0 to n_2) are more distinguishable with 25  70  Chapter 5 Simulation Results for Voice and Data Traffic  percent intra-network traffic. When intra-network traffic changes from 25 to 75 percent, these curves become similar. This is because the queueing delay in gateway switch 0 is dominating the intra-network delay.  20  * •—>  1  18  7500 voice channels (8kbps) 9.6 kbps for data channel  16 h A  Percent of intra-traffic  14  75 % 50% — 25% —  1  B  —  A  —  3 -o  6 o a  Z 3000  4000  5000  6000  7000  Number of data channels per node Figure 5.22 Node 0 and node 15 mean data packet delay with various intra-network traffic.  Chapter 5 Simulation Results for Voice and Data Traffic  0  1000  2000  3000  4000  71  5000  6000  7000  8000  9000  Number of data channels per node Figure 5.23 Mean data packet delay with various intra-network traffic.  Cell delay variation for various nodes appears in Figure 5.24. The delay variations of node 0 to node 15 and node 0 to node 1 are significantly different for 25 and 50 percent intra-network traffic levels. With 75 percent intra-network traffic, the delay variation between these curves vanishes, because data buffer occupancies in each switch are well below congestion levels. Figure 5.25 shows the buffer occupancy in different switches. For 75 percent level of intranetwork traffic, buffer occupancies remain in low level until the data traffic exceeds 8000 channels per node. Gateway and central switch buffers congest at a similar traffic level with 75 percent intra-network traffic. In this case, similar levels of traffic pass through the central and gateway switches.  72  Chapter 5 Simulation Results for Voice and Data Traffic  1000  T  7800 voice channels (8kbps) 9.6kbps for data channel Percent of intra-network traffic central_sw g_sw_0  900 800 700 o  3 O o  h  600  h  500  h  o Stt  400  I  200 100  75% '50% 25 %  &— —  B  ~W  T  i_swJ2 A  —  — l  •  - —H- - '  +  hf 1000  2000  3000  4000  5000  6000  7000  Data channels per node Figure 5.25 Mean data buffer occupancy with various intra-network traffic.  8000  9000  Chapter 6 Simulation Results for Voice and Data Traffic with Silence Deletion In the previous chapter, exponential arrival rate with constant bit stream Is assumed for voice calls. In fact, neither the caller nor callee speaks continuously throughout the call duration. There are silence periods between speech segments when one or both parties don't speak. Use of these silences to support other calls increases the number of voice and data calls in each node. In section 3.4, the radio cell air interface capacity has been examined with a consideration of silence deletion. The network simulation model used in this section is the same as that in the previous chapter, but the conventional voice traffic source has been replaced by one with silence deletion.  6.1 The Two-States Markov Silence Deletion Model The two-states markov silence deletion model consists of a talk state and a silence state. During the simulation, each active voice call switches independently between silence and talk periods. According to [39], the mean time for talk and silence periods is 1.5 and 2.25 seconds, respectively. Both talk and silence periods have exponential distributions.  Figure 6.1  Two state markov silence deletion model.  Consider Figure 6.1, where - and - are the mean times for talk and silence periods,  73  Chapter 6 Simulation Results for Voice and Data Traffic with Silence Deletion  74  respectively. The probability that a call stays in the talk or silence state can be calculated by using a two state markov matrix. Assuming it and n are the stationary probabilities that a call stays in t  s  talk and silence state, respectively, the transition probability matrix can be formed as follows: Talk Talk  Silence  -al  ,  e  Silence  -al  l-e  ,  -bt  -bt  l-e  e  -at  n,-e  +K -(l-e s  -bt  ,,  n •e  -bu ) =K  t  —at.  + % • (1 - e K +K = 1  t  ) =%  s  t  1  .".71, 1  =  ~  b  t  1 -at -e  „  i-e  —xt  Assuming • l-e :.k. = ' a+b  s  s  -bt  -e  ~xt 0.4  if xt« 1 .-.re. = 0.6 s  The results show that the probability of a call staying in a talk state is 0.4; the probability of a call staying in silence state is 0.6.  6.2 Delay and Throughput Analysis of Voice Traffic with Silence Deletion The mixed traffic environment where voice calls with silence deletion co-exist with data calls is examined in this section. The results for data and voice traffic are plotted in Figures 6.2 to 6.5. A comparison with the results of Chapter 5 follows. With the same network configuration as in Chapter 5, it is expected that this network can carry the same amount of traffic (Mbps); however, this network can support more users by increasing the number of voice and data calls that it carries. When there are 1950 voice calls per  Chapter 6 Simulation Results for Voice and Data Traffic with Silence Deletion  '  75  node, the network can carry 2000 more data channels per node compared to the network without the silence deletion scheme (Figure 5.18 and Figure 6.2). Also, it can support almost 1700 more data channels per node when silence deletion is used and there are 7800 voice channels per node in the network (Figure 5.22 and Figure 6.2). A comparison of intra-network traffic delays with and without the silence deletion scheme demonstrates that, with silence deletion, traffic experiences less delay even when the network carries more data traffic (Figure 5.23 and Figure 6.3). Notice that data cell delay variations are stable for. the traffic with the silence deletion scheme, compared to the traffic without silence deletion scheme (Figure 6.4 and Figure 5.24). The fact that similar data cell delay variations are obtained in these simulations, regardless of the number of voice channels per node, implies that the voice with silence deletion scheme has no direct effect on a data buffer occupancy. For example, if a large number of ATM voice cells arrive at a voice queue over a short period of time, data buffer to overflow may occur because the transmitter must dequeue the voice cells in this period; however, this situation does not seem to occur very often. Data buffer occupancies for different switches are plotted in Figure 6.5. Results show that congestion at the central switch data buffer is more likely to occur than the gateway switches when traffic is equally distributed in each network node. The central switch's queue size increases substantially when each network node carries 5000 data channels and 1950 voice channels. The other two sets of curves in Figure 6.5 show much smaller gap between queues at the central and gateway switches. Voice traffic will not introduce a large queueing data traffic delays if a sufficient high transmission rate is used.  Chapter 6 Simulation Results for Voice and Data Traffic with Silence Deletion  0  1000  2000  3000  4000  76  5000  6000  7000  Number of data channels per node Figure 6.2  0  Node 0 to node 15 mean data packet delay.  1000  2000  3000  4000  5000  Number of data channels per node Figure 6.3  Mean data packet delay.  6000  7000  Chapter 6 Simulation Results for Voice and Data Traffic with Silence Deletion  77  A  10  r  I  _J  0  1000  I  1  2000  3000  1  4000  1  5000  Number of data channels per node Figure 6.4  Mean data A T M cell delay variance.  Figure 6.5  Mean data buffer occupancy.  1  6000  1  7000  Chapter 6 Simulation Results for Voice and Data Traffic with Silence Deletion  . 78  6.3 Inter- and Intra-Network Performance With Silence Deletion Scheme Network performance at various levels of inter- and intra-network traffic without voice silence deletion has been studied in section 5.2.3. In this section, the investigation extends to a network with 75 percent intra-network traffic and silence deletion. As in section 5.2.3, 7800 voice calls are simulated initially in the network. Voice traffic is then increased in steps of 3900 voice calls per node in the network. The results are shown in Figures 6.6 to 6.9. Figure 6.6 shows the mean data packet delay between nodes 0 and 15. By comparing this figure with Figure 5.2.1, one sees that the number of data channels per node increases by almost 3500 when 7800 voice calls per node are simulated. The end-to-end delay between nodes 0 and 15 is less than 0.5 ms when the network carries 15600 voice and 6000 data calls per node.  Figure 6.6  Node 0 to node 15 mean data packet delay.  Chapter 6 Simulation Results for Voice and Data Traffic with Silence Deletion  79  The mean intra-network delays in various paths are shown in Figure 6.7. The end-to-end delay between node 0 and node 1 is 139 us when 7800 voice and 3000 data calls per node are simulated in the network. This delay increases to 294 us when the number of data calls per node in the network increases to 12000. The curves in Figures 6.3 and 6.7 have different shapes especially at high level of voice traffic. The curves in Figure 6.7 increase more steeply when compared with those in Figure 6.3.  T  T  I  75 % intra-network traffic voice (8kbps) with silence deletion 9.6kbps for data channel number of voice channels per node . n_0 to n_lg_0 to n_2 g_0 to n_3 15600 A - - A - — A — 11700 -B --Q-- — a —  320 300 280 260  +  7800  - — I  1  240 220  4  h  200 180 160 140 120 100  0  1000  2000 3000 4000 5000  6000 7000  8000 9000 1000011000 1200013000  Number of data channels per node Figure 6.7  Mean data packet delay.  The cell delay variation of data traffic is shown in Figure 6.8. Results show that the differences in cell delay variation from nodes 0 to 15 and nodes 0 to 1 are relatively low with 75 percent intra-network traffic. The central and gateway switches all congest at similar levels of traffic. The maximum cell delay variation is 2.366* 10" sec when 15600 voice and 8000 data calls are 3  simulated in the network.  Chapter 6 Simulation Results for Voice and Data Traffic with Silence Deletion  80  10  io t s  o  75 % intra-network traffic 7800 voice channels (8kbps) 9.6kbps for data channel number of voice channels/node n_0 to n_15 n_0 to n_l g_0 to n_2 15600 A----- - A - — A — 11700  o  0  1000 2000 3000 4000 5000 6000 7000 8000 9000 1000011000 1200013000 Number of data channels per node  Figure 6.8  Mean data packet A T M cell delay variance.  The buffer occupancy in different switches is plotted in Figure 6.9. Unlike Figure 6.5, which the central switch congests much earlier than the gateway switches, the central and gateway switches all congest at similar traffic levels with 75 percent intra-network traffic. Buffer occupancies in Figure 6.9 are lower than in Figure 6.5; simulations show in a fairly low values of a end-toend delay (Figure 6.6).  Chapter 6 Simulation Results for Voice and Data Traffic with Silence Deletion  81  Data channels per node Figure 6.9  Mean data buffer occupancy.  6.4 Network Coverage with Various traffic and Market Penetrations As the results of Chapter 5 and 6 demonstrate, the proposed network with a star topology architecture can support varying number of voice and data calls, depending on the traffic conditions. In the simulation model, there are sixteen switching nodes in the network, and each switching node is connected with a number of connection trees at the sub-level. In each simulation, the number of voice and data calls per node varies. This implies that the coverage area for a given node may also vary in a different simulation. The coverage area per switching node in metropolitan areas are summarized in Table 6.1.  Chapter 6 Simulation Results for Voice and Data Traffic with Silence Deletion  Table 6.1  82  Coverage area per switching node in metropolitan areas with various market penetrations.  Traffic condilion  Number of connection irees/node  Number of users/node  25 '/, intra-network tiallic wilh silence deletion  3.47  62074.83  75 % intra-network traffic with silence deletion  10.40  186045.60  Coverage area per switching node in metropolitan area (Km ) 2  with various market penetration Traffic condilion  75 %  50%  30%  25 % intra-network traffic wilh silence deletion  4.99  7.48  12.46  75 'A inira-neiwork traffic with silence deletion  14.94  22.42  37.36  The results in Table 6.1 are based on 8 kbps voice coding with silence deletion and a 9.6 kbps transmission rate for the data channel. Each connection tree supports 17889 users with 15 base stations (Table 4.1). The coverage area per switching node in metropolitan areas is calculated based on the values stated in Table 4.2. For a 25 percent intra-network traffic with 30 percent market penetration, the proposed network provides up to 200 Km coverage in metropolitan areas. If a higher percentage of intra-network traffic is assumed, the network coverage area increases substantially. The proposed network can cover over 600 Km in metropolitan areas with a 75 percent level of intra-network traffic, 30 percent market penetration, and the silence deletion scheme.  Chapter 7 Summary and Conclusions To meet anticipated growth and high demand for mobile and personal communication services, a new architecture while based on an ATM network to interconnect wireless base sites has been proposed. CDMA technology is assumed to be the PCS air interface. Results obtained from this research are applicable to other air interfaces. Issues specific to PCS such as soft handoff signaling and fast resource allocation have been examined. Voice and data traffic over the ATM network has been examined based on simulations.  7.1 Summary and Findings Comparison of the associated and quasi-associated signaling modes was completed, based on cost and extendability. Results show that the initial setup cost for quasi-associated mode is smaller than for the associated mode, because the signaling network architecture for the quasiassociated signaling mode is similar to that for the SS7 network. However, the associated signaling mode allows for both user and signaling traffic to be carried by the same physical network, which reduces signaling network complexity. A PCS air interface in a microcell environment has been studied intensively. A radio cell with a 2.5 MHz bandwidth can support 75 voice channels and 50 data channels. To improve handoffs, a virtual connection tree architecture is employed in the network between a node switch and a base station. Cell site overloaded probability in the connection tree model has been calculated. Results show that up to 71 and 73 percent of voice and data channels, respectively, can be occupied in a connection tree with 0.01 cell site overloaded probability (Table 3.3). The capacity of a connec-  83  Chapter 7 Summary and Conclusions  84  tion tree with different numbers of base stations has also been examined (Table 4.1). The potential users in a connection tree with 15 base stations are selected for calculating the cell radii in different environments. Cell radii from 192 m to 356 m are proposed for use in metropolitan areas. The cell radii in urban and suburban areas increase substantially, because of the smaller number of potential users. Data traffic with different packet lengths has been simulated in Chapter 5. Increasing data packet size increases both system delay and buffer occupancy. The maximum buffer occupancy of the central switch is 4501 ATM cells when a data packet length of 4980 bytes is used. Voice delay remains stable in all simulations regardless of the amount of network data traffic. Typical values of intra-network and inter-network voice traffic delays are 29.5 us and 64 us, respectively, when 5000 voice channels per network node are simulated. The number of data channels available is inversely proportional to the voice coding rate. The network capacity increases significantly when the level of intra-network traffic increases from 50 to 75 percent. With silence deletion scheme, the proposed network can support almost 1800 more data channels per node with 7800 voice channels per node and 25 percent intra-network traffic. The potential coverage of the simulated 16-nodes network in metropolitan area is 600 K m when 75 percent intra-network traffic and 30 2  percent market penetration are considered.  7.2 Topics for Future Investigation Voice source with silence deleted result in lower mean coding rates that indirectly increases the air interface capacity in terms of number of users in a radio cell. However, the low . voice coding rate introduces high packetization delay to an ATM cell (Figure 2.8). An ATM cell that carries multiple voice calls may alleviate this problem, but there are the following concerns  Chapter 7 Summary and Conclusions  85  with this approach in an ATM network: 1) The addresses associated with multiple voice calls mustfitinto a single VCI field in an ATM cell header; 2) a destination node must be able to distinguish an ATM cell that carries one voice call from one with multiple voice calls; 3) the routing of cells with multiple voice calls requires consideration and analysis; and 4) handoffs associated with multi-call cells are complex and require analysis. Another issue that may be of interest is wide area network performance of a mesh type internetworking architecture. Such architecture provides more than one path from source to destination, and provides for improved reliability as well as some means for reduced delay resulting from localized congestion.  Bibliography [1]  V. O. K. Li, X. Qiu, "Personal Communication Systems (PCS)," IEEE Pro., Vol. 83, No. 9, pp. 1210 - 1243, Sept. 1995.  [2]  D. Goodman, "Second Generation Wireless Information," IEEE Trans. Vehic. Technol., Vol. 40, No. 2, pp. 366 - 374, May 1991.  [3]  D. E. Sterling, J. E. Haleid, "The Iridium System - A Revolution Satellite Communications System Developed with Innovative Application of Technology," Vol. 2 , pp. 436 - 440, MILCom'91, Mclean, VA, USA.  [4]  V. C. M . Leung, N . Qian, A. D. Malyan, R. W. Donaldson, "Call Control and Traffic Transport for Connection-Oriented High Speed Wireless Personal Communications Over Metropolitan Area Networks," IEEE J. Select. Area in Commun., Vol. 12, No. 8, pp. 1376 - 1388, Oct. 1994.  [5]  C. M . Cobbold, "Voice Transport, Capacity and Signaling Improvements for Integrated Wireless Personal Communications Over Metropolitan Area Networks," Master's thesis, UBC Dept. EE. Sept. 1995.  [6]  W. Y. L. Wong, "Delay-Throughput Analysis of Intra-and Intre-MAN Voice and Data Integrated Traffic in IEEE 802.6 M A N Based PCN," Master's thesis, UBC Dept. EE. Feb. 1995.  [7]  N . Qian, "Call Control Signaling for Personal Communications Over Interconnected Metropolitan Area Networks," Master's thesis, UBC Dept. EE. April 1993.  [8]  A. D. Malyan, L. J. Ng, V. C. M . Leung, R. W. Donaldson, "Network Architecture and Signaling for Wireless Personal Communications," IEEE J. Select. Area in Commun., Vol. 11, No. 6, pp. 830 - 841, Aug. 1993.  [9]  D. J. Goodman, G. P. Pollini, K. S. Meier-Hellstern, "Network Control for Wireless Communications," IEEE Commun. Mag., pp. 116 -124, Dec. 1992.  [10]  D. Raychaudhuri, N . D. Wilson, "ATM-Based Transport Architecture for Multi-services Wireless Personal Communication Netowrks," IEEE J. Select. Area in Commun., Vol. 12, No. 8, pp. 1401- 1414, Oct. 1994  86  Bibliography  87  [11]  P. Seite, "Soft Handoff in a DS-CDMA Microcellular Netowrk," IEEE 44th Vehic. Technol. conf, Stockholm, Vol. 1, pp. 530 - 534, 1994.  [12]  C. L I, L. J. Greenstein, R. D. Gitlin, "A Microcell/Marcocell Cellular Architecture for ' Low and High Mobility Wireless Users," IEEE J. Select. Area in Commun., Vol. 11, No. 6, pp. 885 - 891, Aug. 1993.  [13]  T. H. Wu, N . Yoshikai, H. Fujii, "ATM Signaling Transport Network Architectures and Analysis," IEEE Commun. Mag., pp. 90 - 99, December 1995.  [14]  J. Luetchford, T. H. Wu, N. Yoshikai, "Network Common Channel Signaling Evolution," in Processing , ISS 95, vol. 2, pp. 234 - 238, April 1995.  [15]  D. McDysan, D. Spohn, "ATM Theory and Application," New York, McGraw-Hill, 1994.  [16]  R. O. Onvural, "Asynchronous Transfer Mode Networks Performance Issues," Boston, Artech House, 1994.  [17]  A. Salmasi, K. S. Gilhousen, "On the System Design Aspects of Code Division Multiple Access (CDMA) Applied to Digital Cellular and Personal Communications Network," 41st IEEE Vehic. Technol. Conf., St. Louis, MO, USA, pp. 75 - 62, May 1991.  [18]  S. Nanda, "Teletraffic Models for Urban and Suburban Microcells: Cell Sizes and Handoff Rates," IEEE Trans. Vehic. Technol., Vol. 42, No. 4, pp. 673 - 682, Nov. 1993.  [19]  Qualcomm Inc., "An Overview of the Application of Code Division Multiple Access (CDMA) to Digital Cellular Systems and Personal Cellular Networks," Research Report, March 1992.  [20]  A. S. Acampora, M . Naghshineh, "An Architecture and Methodology for MobileExecuted Handoff in Cellular ATM Networks," IEEE J. Select. Area in Commun., Vol. 12, No. 8, pp. 1365 - 1375, Oct. 1994.  [21]  M. J. McTiffin, A. P. Hulbert, T. J. Ketseoglou, W. Heimsch, G. Crisp, "Mobile Access to an ATM Network Using a CDMA Air Interface," IEEE J. Select. Area in Commun., Vol. 12, No. 5, pp. 900 - 908, June 1994.  [22]  A. Merchant, B. Sengupta, "Assignment of Cells to Switches in PCS Networks," IEEE/ A C M Trans, on Net. Vol. 3, No. 5, pp. 521 - 526, Oct. 1995.  Bibliography  88  [23]  P. T. S. Yum, K. L. Yeung, "Blocking and Handoff Performance Analysis of Directed Retry in Cellular Mobile System," IEEE Trans. Vehic. Technol., Vol. 44, No. 3, pp. 645 650, Aug. 1995.  [24]  A. S. Acampora, J. F. P. Labourdette, "A Traffic-Handling Comparison of Centralized and Distributed ATM Switching Systems," IEEE Trans. Commun., Vol. 43, No. 6, pp. 2070 2076, June 1995.  [25]  J. Cosmas, B. Evans, C. Evci, W. Herzig, H. Persson, J, Pettifor, P. Polese, R. Rheinschmitt, A. Samukic, "Overview of the Mobile Communications Programme of RACE II," IEE Commun. Mag., Vol. 7, No. 4, pp. 155 - 167, Aug. 1995.  [26]  A. J. M . Ransom, "Handoff Considerations in Microcellular System Planning," Toronto, Vol. 2, PIMRC'95.  [27]  H. H. Xia, H. L. Bertoni, L. R. Maciel, A. Lindsay-Stewart, R. Rowe, "Microcelluler Propagation Characteristics for Personal Communicatons in Urban and Suburban Environments," IEEE Trans. Vehic. Technol., Vol. 43, No. 3, pp. 743 - 752, Aug. 1994.  [28]  J. Nguyen, R. Tarn, T. Watts, "Personal Communications Serivce - Concept and Architecture," 41st IEEE Vehic. Technol. Conf., St. Louis, MO, USA, pp. 75 - 62, May 1991.  [29]  N. D. Wilson, R. Ganesh, K. Joseph, D. Raychaudhuir, "Packet CDMA Versus Dynamic TDMA for Multiple Access in an Integrated Voice/Data PCN," IEEE J. Select. Area in Commun., Vol. 11, No. 6, pp. 870 - 884, Aug. 1993.  [30]  O. T. W. Yu, V. C. M . Leung, "Connection Architecture and Protocols to Support Efficient Handoffs Over an ATM/B-ISDN Personal Communicatons Network," Research Report, Dept. EE. UBC, April 1995.  [31]  D. J. Goodman, "Cellular Packet Communications," IEEE Trans. Commun., Vol. 38, No. 8, pp. 1272 - 1280, Aug. 1990.  [32]  K. S. Meier-Hellstern, G. P. Pollini, "Network Protocols for the Cellular Packet Switch," IEEE Trans. Commun., Vol. 42, No. 2/3/4, pp. 1235 - 1244, Feb./March/April 1994.  [33]  B. Jabbari, "Intelligent Network Concepts in Mobile Communications," IEEE Commun. Mag., Vol. 30, No. 2, pp. 64 - 69, Feb. 1992.  Bibliography  89  [34]  T. F. La-Porta, M. Veeraraghavan, P. A. Treventi, R. Ramjee, "Distributed Call Processing for Personal Communications Services," IEEE Commun. Mag., Vol. 33, No. 6, pp. 66 75, June 1995.  [35]  Q. A. Zeng, K. Mukumoto, A. Fukuda, "Performance Analysis of Mobile Cellular Radio System with Priority Reservation Handoff Procedures," 44th Vehic. Technol. Conf., Stockholm, Vol. 3, pp. 1829 - 1833, 1994.  [36]  A. Jalali, P. Mermelstein, "Effects of Diversity, Power Control, and Bandwidth on the Capacity of Microcellular CDMA Systems," IEEE J. Select. Area in Commun., Vol. 12, No. 5, pp. 952 - 961, June 1994.  [37]  J. Shapira, "Microcell Engineering in CDMA Cellular Networks," IEEE Trans. Vehic. Technol., Vol. 43, No. 4, pp. 817 - 825, Nov. 1994.  [38]  W. T. Chen, C. F. Huang, S. C. Ding, "A Scheme for QOS Control in ATM Switching System," 45th Vehic. Technol. Conf., Vol. 3, pp. 196 - 200, 1995.  [39]  K. Apostolidis, "A Reservation Protocol for Packet Voice and Data Intergration in Unidirectional Bus Network," IEEE Trans. Commun., Vol. 41, No. 3, pp. 478 - 485, March 1993.  [40]  C. A. Johnston, "Architecture and Performance of HIPPI-ATM-SONET Terminal Adapters," IEEE Commun. Mag., Vol. 33, No. 4, pp. 46 - 51, April 1995.  [41]  F. Vakil, H. Saito, "On Congestion Control in ATM Networks," IEEE LTS. Vol. 2, No. 3, pp. 55 -65, Aug. 1991.  [42]  B. G. Kim, P. Wang, "ATM Netowrk: Goals and Challenges," IEEE/ACM Commun. Vol. 38, No. 2, pp. 40 - 44, Feb. 1995.  [43]  C. Liu, H. T. Mouftah, "Virtual Call Admission Control - A Strategy for Dynamic Routing Over ATM Networks," 45th Vehic. Technol. Conf., Vol. 2, pp. 201 - 205,1995.  [44]  M . Rahnema, "Overview of the GSM System and Protocol Architecture," IEEE Commun. Mag., Vol. 31, No. 4, pp. 92 - 100, April 1993.  [45]  J. H. Chao, N. Uzun, "An ATM Queue Manager with Multiple Delay and Loss Priorities," Globecom'92, IEEE Global Telecommun. Cof., Orlando, FL, USA, Vol. 1, pp. 308 - 313, 1992,.  Bibliography  [46]  90  R. Franz, K. D. Gradisching, M . N . Huber, R. Stiefel, "ATM-Based Signaling Network Topics on Reliability and Performance," IEEE J. Select. Area in Commun., Vol. 12, No. 3, pp. 517 -525, April 1994.  Appendix A. List of Abbreviations and Acronyms AAL  ATM Adaptation Layer  ACK  Acknowledgment  AMPS  . Advanced Mobile Phone System .  ANSI  American National Standard Institute  ATM  Asynchronous Transfer Mode  B-ISDN  Broadband - Integrated Services Digital Network  BS  Base Station  CBR  Constant Bit rate  CCITT  Consultative Comrnittee International Telegraphy and Telephone  CDMA  Code Division Multiple Access  CDV  Cell Delay Variation  CLP  Cell Loss Priority  ETSI  European Telecommunication Standard Institute  FDMA  Frequency Division Multiple Access  GFC  Generic Flow Control  GSM  Global System for Mobile Communication  HEC  Header Error Check  HLR  Home Location Register  IETF  Internet Engineering Task Force  ITU  International Telecommunication Union  MAN  Metropolitan Area Network  MSC  Mobile Switching Center  91  Appendix A. List of Abbreviations and Acronyms  MTP  Message Transfer Part  MUX  Multiplexer  NNI / UNI  Network to Network Interface / User to Network Interface  PCS  Personal Communication Services  PDU  Protocol Data Unit  PL  Physical Layer  POTS  Plain Old Telephony Service  PRJVI  Protocol Reference Model  PSTN  Public Service Telephone Network  PT  Payload Type  QOS  Quality of Service  SAP  Service Access Point  SAR  Segmentation and Reassembly  SCP  Signal Control Point  SDU  Service Data Unit  SIG  SMDS Interest Group  SMDS  Switched MultiMegabit Data Service  SONET  Synchronous Optical Fiber Network  SS7  Signaling Scheme 7  SSP  Service Switching Point  STP  Signal Transfer Point  TCP / IP  Transmission Control Protocol / Internet Protocol  TDMA  Time Division Multiple Access  92  Appendix A. List of Abbreviations and Acronyms  VBR  Variable Bit Rate  VCC / VPC  Virtual Channel Connection / Virtual Path Connection  VCN  Virtual Circuit Number  VLR  Visitor Location Register  VPX  Virtual Path Cross Connection  Appendix B. Calculation of Confidence Interval The confidence interval shows the acceptance region of a data set. If the sample size is small, t-test is most frequently used to calculate the confidence interval. The acceptance region can be expressed as follows:  Mean (sample)  (B 1)  x=  where x and n are data value and number of samples, respectively.  Standard deviation (sample)  s =  V  (B2)  n-1  The confidence interval (small sample):  x-\t  l a / 2  Jn)  <[i<x +  (B 3)  \t --= a/2  Where t and a are the distribution and the level of significance, respectively. A set of data based on 1950 voice channels (with silence deletion) and 1000 data channels per network node is examined. The 99 percent confidence intervals of voice delay, data delay, and buffer occupancy are calculated and shown in Table B. 1.  Table B. 1 Results of confidence interval calculation. Mean value  Confidence interval  Percentage of error (%)  Node 0-15 voice A A L delay  49.357 us  [49.242us <ux49.471 us]  0.231  Node 0-15 data A A L delay  148.834 us  [146.242us <u< 151.427 us]  1.742  Node 0-1 voice AA1 delay  28.924 us  [28.870 us <u< 28.978 us]  0.187  Node 0-1 data AA1 delay  102.158 us  [ 100.910 us <|x< 103.405 us]  Node sw_0 buffer  2.7816 A T M cells  [ 2.6567 < | X < 2.907]  4.49  Central switch buffer  3.477 A T M cells  [ 3.417 <n< 3.536]  1.723  94  . 1.222  Appendix C. OPNET ATM Models  

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.831.1-0065321/manifest

Comment

Related Items