Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Quality of service enhancements in IEEE 802.11 wireless LANs 2003

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
ubc_2003-0589.pdf
ubc_2003-0589.pdf [ 2.68MB ]
Metadata
JSON: 1.0065459.json
JSON-LD: 1.0065459+ld.json
RDF/XML (Pretty): 1.0065459.xml
RDF/JSON: 1.0065459+rdf.json
Turtle: 1.0065459+rdf-turtle.txt
N-Triples: 1.0065459+rdf-ntriples.txt
Citation
1.0065459.ris

Full Text

Quality of Service Enhancements in IEEE 802.11 Wireless LANs by George Wai Wong B.A.Sc, The University of British Columbia, Canada, 2001 A THESIS SUBMITTED IN PARTIAL F U L F I L L M E N T OF T H E REQUIREMENTS FOR T H E D E G R E E OF M A S T E R O F APPLIED SCIENCE in T H E F A C U L T Y OF G R A D U A T E STUDIES D E P A R T M E N T OF E L E C T R I C A L AND C O M P U T E R ENGINEERING We accept this thesis as conforming to the required standard T H E UNIVERSITY O F BRITISH C O L U M B I A September 2003 © George Wai Wong, 2003 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 Electrical and Computer Engineering The University of Brit ish Columbia Vancouver, Canada Date: September 26 , 2003. Abstract One of the most important of the many developments in telecommunications is the convergence o f voice, video, and data communications within the Internet Protocol (IP) suite. IP was originally designed to support best-effort data services. The development o f IP-based multimedia networking applications has imposed Quality of Service (QoS) requirements on the IP network, especially with respect to real-time traffic. The Internet Engineering Task Force (IETF) is currently working on QoS differentiation at the LP layer; however the result is sub-optimal without lower layers' support. Wi th the increasing use of wireless Internet services over 802.11 wireless L A N s , it is essential to focus on QoS differentiation support at the 802.11 M A C layer. To improve the current 802.11 M A C protocol, the I E E E Task Group E was formed and is defining QoS enhancements for the 802.11 M A C layer by introducing an Enhanced Distributed Coordination Function (EDCF) . The E D C F provides prioritization enhancement of the 802.11 Distributed Coordination Function (DCF). The objective of this thesis is to propose and evaluate a novel packet retransmission algorithm called Age-Dependent Backoff to improve the QoS performance of E D C F . The A D B algorithm dynamically varies the persistent factor associated with the contention window, based on real-time packet queue age and lifetime. A D B maintains the backward compatibility property o f E D C F and involves relatively easy implementation at low cost. Extensive simulation results obtained using O P N E T software show that E D C F with the A D B retransmission algorithm provides low values for delay, jitter, and drop rate for real- time traffic without sacrificing the throughput o f best-effort data traffic. A D B is a viable, novel and low cost means to improve the QoS performance of E D C F . i i Table of Contents Abstract i i Table of Contents i i i List of Tables v i List of Figures v i i Acknowledgments ix Chapter 1. Introduction 1 1.1 Background on I E E E 802.11 Wireless L A N s 1 1.2 Motivations 3 1.3 Thesis Contributions 4 1.4 Thesis Outline 5 Chapter 2. Overview of IEEE 802.11 Wireless LANs 7 2.1 Introduction to the I E E E 802.11 Standard 7 2.1.1 802.11 Topology 8 2.1.2 802.11 Protocol Architecture 10 2.2 802.11 D C F Protocol 12 2.3 New 802.1 l e E D C F Protocol 14 Chapter 3. Age-Dependent Backoff Algorithm for 802.1 le EDCF 18 3.1 Backoff Behaviour of 802.11e E D C F 18 3.2 Age-Dependent Backoff Algorithm 19 3.3 Implementation Complexity 21 3.3.1 Implementation of the A D B Algorithm 21 i i i 3.3.2 Discussion of the Implementation Complexity 24 Chapter 4. Design of Simulation Models 25 4.1 Overview of O P N E T Simulation Tools 26 4.2 802.11 Model Design 28 4.2.1 802.11 Physical Layer Mode l 28 4.2.2 802.11 M A C Layer Model 30 4.3 Simulation Models for Upper Layer Protocols 36 4.3.1 Network Layer Model : Internet Protocol 36 4.3.2 Transport Layer Models: T C P and U D P 37 4.4 Traffic Source Models 40 4.4.1 Voice Model 41 4.4.2 Video Conferencing Model 42 4.4.3 F T P Fi le Transfer Model 43 Chapter 5 Performance Analysis of the Proposed ADB Algorithm in an Ad-Hoc Network and a Hotspot 45 5.1 Ad-Hoc Network Scenario 46 5.1.1 Overview of the Ad-Hoc Network Scenario 46 5.1.2 Simulation Results from the Ad-Hoc Network Scenario 47 5.1.3 Discussions of the Ad-Hoc Network Results 51 5.2 Hotspot Scenario 52 5.2.1 Overview of the Hotspot Scenario 52 5.2.2 Simulation results from the Hotspot Scenario 54 5.2.3 Discussions of the Hotspot Results 61 5.3 Conclusion from the Simulation Results 62 iv Chapter 6. Summary and Conclusions 64 6.1 Summary of the Work 65 6.2 Future Work 66 References 67 v List of Tables Table 4.1. Parameters for the M A C layer 35 Table 4.2. Default settings o f the T C P parameters 39 Table 4.3. Parameters for video conferencing 43 Table 4.4. Parameters for F T P file transfers 44 v i List of Figures Figure 2.1. I E E E reference model comparing with the OSI reference model 8 Figure 2.2. Independent basic service set 9 Figure 2.3. Extended service set 9 Figure 2.4. 802.11 protocol layers 10 Figure 2.5. 802.11 D C F M A C protocol operation 12 Figure 2.6. RTS /CTS/da ta /ACK and N A V settings 14 Figure 2.7. Service differentiation by different A IFS values 15 Figure 2.8. Transmission architecture of E D C F vs. D C F 16 Figure 3.1. PF[TC] vs. Age 20 Figure 3.2. F low chart for the implementation of the A D B retransmission algorithm.. 23 Figure 4.1. Hierarchical structure for O P N E T models 26 Figure 4.2. Finite state machine for 802.11 M A C layer model with E D C F and A D B . . 33 Figure 4.3. Two-state model of a voice user 41 Figure 5.1. Network topology of an ad-hoc scenario 47 Figure 5.2. Voice packet delay in an ad-hoc network 48 Figure 5.3. Voice packet jitter in an ad-hoc network 48 Figure 5.4. Voice packet drop rate in an ad-hoc network 49 Figure 5.5. Video packet delay in an ad-hoc network 49 Figure 5.6. Video packet jitter in an ad-hoc network 50 Figure 5.7. Video packet drop rate in an ad-hoc network 50 Figure 5.8 Total F T P data throughput in an ad-hoc network 51 vi i Figure 5.9. Network topology of a hotspot scenario 53 Figure 5.10. Upl ink and Downlink voice packet delay in a hotspot 55 Figure 5.11a. Upl ink voice packet j itter in a hotspot 56 Figure 5.11b. Downlink voice packet jitter in a hotspot 56 Figure 5.12a. Upl ink voice packet drop rate in a hotspot 57 Figure 5.12b. Downlink voice packet drop rate in a hotspot 57 Figure 5.13. Video packet delay in a hotspot 58 Figure 5.14a. Upl ink video packet j itter in a hotspot 59 Figure 5.14b. Downlink video packet jitter in a hotspot 59 Figure 5.15a. Upl ink video packet drop rate in a hotspot 60 Figure 5.15b. Downlink video packet drop rate in a hotspot 60 Figure 5.16. Total F T P data throughput in a hotspot 61 v i i i Acknowledgments I wish to thank most sincerely my supervisor, Dr. Robert W . Donaldson, for his inexhaustible patience, encouragement and valuable assistance in connection with this thesis. I would like to express my thanks to my colleague, Victor Fong, for participating in technical and non-technical discussions with me. I would also like to thank O P N E T technologies, Inc. for providing us a free academic version of O P N E T Modeler. Finally, I am most grateful to my parents not only for their support but also for their constant encouragement. ix Chapter 1 Introduction In this introductory chapter, background material on I E E E 802.11 wireless L A N s is reviewed in section 1.1. The motivations for this thesis work are presented in section 1.2. The thesis contributions are summarized in section 1.3. Finally, the outline of the thesis is provided in section 1.4. 1.1 Background on IEEE 802.11 Wireless LANs Wireless L A N products first appeared iri the late 1980s, marketed as substitutes for traditional wireline L A N s [1]. The idea was to use wireless L A N s to avoid the cost of installing L A N cables and to ease the task of relocating computer stations. The history of wireless L A N development has been fraught with proprietary non-standard technologies which cause wireless L A N products being non-interoperable among vendors. Wi th the ratification of the IEEE 802.11 wireless L A N standard, wireless L A N s have emerged from being proprietary implementations to become open solutions for providing mobility as wel l as essential network services where wireline installations are impractical [2]. The original I E E E 802.11 standard [3], published in 1997, supports data rates up to 2 Mbps in the 2.4-GHz industrial, scientific, and medical (ISM) band. In 1999, The I E E E 802.11 Working Group published its enhanced versions, I E E E 802.11a [4] and 802.11b [5], that extend the data rate to 54 Mbps in the 5-GHz unlicensed national information infrastructure (UNII) band and 11 Mbps in the 2.4-GHz I S M band, respectively. Recently, the I E E E has approved the final specification for IEEE 802.1 l g [6], which is backwards 1 compatible with 802.11b and which boosts the bandwidth capability to 54 Mbps in the 2.4- G H z I S M band. Although most of the growth in 802.11 wireless L A N s has been around the 802.11b standard, pundits predict that the emergence of the 802.1 l g standard wi l l dominate the 802.11 wireless L A N market because of the full backward compatibility with 802.1 l b and the high data rate of 54 Mbps [7]. Despite not being finalized until recently, many vendors have already been shipping 802.1 l g products based on the drafts of the 802.1 l g standard since late 2002. Fol lowing the approval of 802.1 l g , the formation of IEEE 802.11 Task Group N began. The data rate of 802.1 In has not been determined yet but is expected to be at least 108 Mbps or possibly beyond up to 320 Mbps. The I E E E 802.11 wireless L A N standard is based on the definition of the medium access control ( M A C ) protocol and the physical layer (PHY) specifications. The P H Y defines frequency bands, data rates, and other details of the actual wireless transmission. Above the P H Y is the M A C layer, which regulates access to the shared wireless channel so that station transmissions do not interfere with one another. The 802.11 M A C layer was designed to be common among different 802.11 P H Y s such as 802.11a, 802.11b and 802.1 l g . The 802.11 M A C layer protocol is specified in terms of coordination functions that determine when a wireless station is allowed to transmit data over the wireless medium. The mandatory Distributed Coordination Function (DCF) uses a carrier-sense multiple access/collision avoidance ( C S M A / C A ) protocol for sharing the wireless medium. This C S M A / C A protocol reduces the probability of collisions among stations which share the medium by using a random backoff time when the medium is busy. Once the medium is idle, a random backoff time defers a station from transmitting a frame, minimizing the chance of inter-station collisions. 2 802.11 wireless LANs are today's most deployed wireless LANs and are expected to play a major role in next-generation wireless communications. The increasing use of wireless Internet services has created a strong demand for public Internet access over wireless LANs. It is envisioned that 802.11 will soon become one of the most common wireless Internet access technologies. 1.2 Motivations The world of communications has undergone many changes over the last few years. One of the most important changes is the convergence of voice, video, and data communications within the Internet Protocol (IP) suite. IP was originally designed to support data services such as file transfer, e-mail and remote terminal, which are tolerant of delay and jitter. Voice and video services, as opposed to data services, require a minimum transmission rate and suffer significantly from high delay and jitter. The development of IP-based multimedia networking applications has imposed additional requirements on the IP network, creating a need for end-to-end Quality of Service (QoS) support. Although the Internet Engineering Task Force (IETF) is currently working on service differentiation at the IP layer, the result is sub-optimal without lower layers' support. In recent years, there has been a substantial increase in the deployment of 802.11 wireless LANs for wireless Internet services. Wireless Internet Service Providers (WISPs) are appearing everywhere, deploying 802.11 hotspots in coffee shops, hotels, and airports. With the growing popularity and acceptance of 802.11 wireless LANs, it is essential to focus on service differentiation support at the 802.11 M A C layer. 3 To improve the current 802.11 M A C protocol to support applications with QoS requirements, the I E E E 802.11 Task Group E was formed and is defining QoS enhancements for the 802.11 M A C protocol. The 802.1 l e draft introduces Enhanced Distributed Coordination Function (EDCF) and Hybrid Coordination Function (HCF) , both of which are currently under discussion [8]. E D C F is a prioritization enhancement of the 802.11 D C F using the Virtual D C F mechanism [9]. E D C F is the contention-based channel access mechanism for H C F which is based on a poll ing mechanism similar to the 802.11 Point Coordination Function (PCF) [10]. H C F allows the Hybrid Coordinator (HC) , typically located at the access point (AP), to initial contention-free periods at any given time during a contention period. The success of 802.11 wireless L A N s is based on D C F , which is a distributed protocol with minimum management and maintenance costs. The dynamic nature of ad-hoc networks makes it difficult to use a poll ing protocol relying on centralized control to maintain connection, reservation, and scheduling states. Wi th these observations, we argue that a distributed control of wireless channel results in a more productive use of wireless resources than a centralized control. Therefore, we w i l l focus on the improvement of E D C F in this thesis. 1.3 Thesis Contributions The thesis extends the current work on providing QoS enhancements in 802.11 wireless L A N s studied by I E E E 802.11 Task Group E. In particular, we address the improvement of the new 802.1 l e E D C F by proposing an efficient retransmission algorithm called Age-Dependent Backoff (ADB) [11], which dynamically adjusts the persistent factors 4 based on the ages of the real-time packets in the transmission queues and the lifetimes of the real-time packets. The complexity of implementing the A D B algorithm is relatively low. A D B requires minor modifications in the computation of C W minimizing the migration effort from the 802.1 le E D C F and provides backward compatibility to the 802.11 D C F . The performance of E D C F with the A D B retransmission algorithm is evaluated using the O P N E T simulation tools. We test the A D B algorithm in two typical environments, an ad- hoc network and a hotspot, in which 802.11 wireless L A N s might most probably be applied. The ad-hoc network is used mostly in situations where users need to deploy a network to start communication quickly without the benefit of a fixed network irifrastructure. The hotspot is usually deployed by a WISP, providing wireless Internet access services for public use. In both environments, voice, video and data services are active simultaneously. Two simulation scenarios are implemented using O P N E T to model these two common environments. We study the improvements in service differentiation under the environments when the A D B retransmission algorithm is employed in the new 802.1 l e E D C F protocol. The simulation results indicate that using A D B in E D C F provides low values for delay, jitter and drop rate for real-time traffic without sacrificing the throughput of the best- effort traffic in a wide range of traffic loads and network configurations. 1.4 Thesis Outline The subsequent chapters o f this thesis are organized as follows. In Chapter 2, we review the 802.11 standard and introduce the new 802.1 l e E D C F protocol proposed by the 802.11 Task Group E. In Chapter 3, we describe our proposed A D B retransmission algorithm 5 for E D C F and discuss the complexity of A D B . The simulation model is presented in Chapter 4 and the performance analysis of 802.1 l e E D C F with A D B is studied in Chapter 5. Finally, a summary and conclusions are provided in Chapter 6. 6 Chapter 2 Overview of IEEE 802.11 Wireless LANs Before studying the details of anything, it often helps to get a general "lay of the hand." A basic introduction is often necessary when studying networking topics because the number of acronyms can be overwhelming. The 802.11 standard uses a significant number of acronyms, which makes the introduction to be important. This chapter introduces the acronyms used throughout the entire thesis and provides an overview of the I E E E 802.11 wireless L A N standard. The introduction to the 802.11 wireless L A N standard appears in Section 2.1. The 802.11 D C F , a mandatory 802.11 M A C protocol, mentioned briefly in the previous chapter is presented thoroughly in Section 2.2. Finally, Section 2.3 describes the new 802.1 le E D C F protocol proposed by the 802.11 Task Group E for supporting service differentiation at the M A C layer of 802.11 wireless L A N s . 2.1 Introduction to the IEEE 802.11 Standard The 802.11 wireless L A N standard, officially called " I E E E Standard for Wireless L A N Medium Access Control ( M A C ) and Physical Layer (PHY) Specifications," defines over-the-air protocols necessary to support networking in a local area. A s with other I E E E 802-based standards (such as 802.3 Ethernet and 802.5 Token Ring), the primary service o f the 802.11 standard is to deliver M A C Service Data Units (MSDUs) between peer Logical L ink Controls (LLCs) . A s a refresher for readers, the I E E E Standards Committee subdivided the Data L ink layer in the open systems interconnection (OSI) reference model developed by the International Organization for Standardization (ISO). The result of this subdivision depicted in Figure 2.1 split the Data L ink layer into a M A C layer and a L L C layer [12]. L L C 7 is the highest layer of the IEEE 802 reference model, providing addressing and data link control, and is independent of the network topology, the transmission medium, and the M A C protocol. The 802.11 standard specifies M A C protocols and P H Y specifications for wireless connectivity of fixed, portable, and mobile stations moving at pedestrian or vehicular speed within a local area. Wireless cards and access points typically provide functions of the 802.11 standard. OSI Reference IEEE 802 Reference Model Model Data Link Layer Logical Link Layer (LLC) Medium Access Control (MAC) Physical Layer Physical Layer Figure 2.1. IEEE reference model comparing with the ISO reference model 2.1.1 802.11 Topology The basic building block of a 802.11 wireless L A N is a basic service set (BSS), which consists of stations that execute the same M A C protocol and compete for access to the same shared wireless medium. A n independent B S S (IBSS) is a standalone B S S that has no backbone infrastructure and consists of at least two wireless stations as shown in Figure 2.2. This type of network configuration is often referred to as an ad-hoc network. A B S S may be connected to a backbone distribution system through an access point (AP) to form an 8 extended service set (ESS) as illustrated in Figure 2.3. The term infrastructure network is used informally to refer to the E S S . The BSSs are like cells in a cellular network and the A P s are analogous to base stations in the cellular network [13]. This type of configuration satisfies the needs of large coverage networks of arbitrary size and complexity. Figure 2.2. Independent basic service set Figure 2.3. Extended service set 9 2.1.2 802.11 Protocol Architecture Figure 2.4 depicts the protocol architecture o f the 802.11 standard. The channel access for the wireless stations in a B S S is under the control of the M A C layer with two sublayers, namely Distributed Coordination Function (DCF) and Point Coordination Function (PCF). D C F is a contention-based M A C protocol providing asynchronous data transfer on a best-effort basis and is a mandatory M A C protocol supported by all stations. The access control in ad-hoc networks uses D C F only. Infrastructure networks can operate using just D C F or a coexistence of D C F and P C F . Details of the D C F operation wi l l be described in section 2.2. 802.2 Logical link control (LLC) t MAC layer Physical layer I Contention-free service * Contention service Point coordination function (PCF) 1 Distributed coordination function (DCF) 2.4-GHz 2.4-GHz Infrared frequency- direct- hopping sequence spread spread spreatrum sprectrum Data rates of 1 and 2 Mbps 802.11 5-GHz 2.4-GHz 2.4-GHz orthogonal direct- orthogonal frequency- sequence frequency- division spread division multiplexing sprectrum multiplexing Data rates Data rates Data rates of 6, 9,12, of 5.5 and up to 18, 24,36, 48, 11 Mbps 54 Mbps and 54 Mbps 802.11a 802.11b 802.1 lg Figure 2.4. 802.11 protocol layers 10 P C F is a centralized M A C protocol providing contention-free multiple access service for the transmission of real-time traffic by poll ing stations in turn. Studies have shown that it is difficult for P C F to achieve high efficiency and to satisfy the time requirement for real-time traffic, [14], [15] and the cooperation between D C F and P C F leads to poor performance [16]. P C F is not supported by commercially available wireless cards and A P s . The P H Y specifies the actual transmission details and has been the focus of much work in the past few years. The 802.11 standard specifies several P H Y s . The initial standard approved in 1997 included frequency hopping spread spectrum (FHSS) and direct-sequence spread spectrum (DSSS) delivering data rates of 1 and 2 Mbps in the 2.4-GHz I S M band. The standard also defined an infrared option operating at 1 and 2 Mbps via passive ceiling reflection; one reason that the infrared option has never gained wide market support might be because infrared transmissions cannot penetrate walls. Most of the early 802.11 wireless L A N s are F H S S due to the relative implementation simplicity. In 1999, the IEEE issued the second and third P H Y s , I E E E 802.11a and 802.11b, at roughly the same time. 802.1 l a operates in the 5 -GHz UNII band at data rates up to 54 Mbps using orthogonal frequency division multiplexing ( O F D M ) ; 802.11b operates in the 2.4-GHz I S M band with 5.5 and 11 Mbps data rates using DSSS. The I E E E has recently approved the final specification for 802.1 l g which adopts 802.1 la ' s O F D M but runs 54 Mbps in the 2.4- G H z I S M band and is backwards compatible with 802.11b. Most 802.11 wireless L A N s deployed today comply with the 802.1 l b standard. 11 2.2 802.11 D C F Protocol The fAindamental access method of the 802.11 M A C protocol is D C F , which supports asynchronous data transfer on a best-effort basis and is the only possible M A C protocol in 802.11 ad-hoc networks. The 802.11 D C F M A C protocol operation is depicted in Figure 2.5. Immediate access when medium is free >= DIFS -DIFS- Busy Medium •DIFS- — PIFS -SIFS+ • Contention Window - n —i—i—i—i—i— Backoff-Window _i i i i i Next Frame Defer Access h Slot time Select slot and decrement backoff as long as medium is idle Figure 2.5. 802.11 D C F M A C protocol operation For a station to transmit a M A C protocol data unit ( M P D U ) , it must sense the medium to determine i f another station is transmitting and must ensure that the medium is idle for the specified D C F Interframe Space (DIFS) duration. A station may transmit a pending M P D U when it determines that the medium is idle for a time interval greater than or equal to the DIFS period. If the medium is sensed busy, the station continues to sense the channel for an additional random time after detecting the channel as being idle for the DIFS duration. The additional random time period is selected from C W . The size of C W , bounded by the maximum value CWmax, is doubled after each unsuccessful transmission to reduce the 12 coll ision probability. Fol lowing each successful transmission, C W is reset to the minimum value CWmi„. This is the well-known binary exponential backoff (BEB) algorithm. The backoff time, backoff time, can be expressed as follows [17]: backoff _time = randInt(0,mm(CWTnin x2reny ,CWmax))xslot _time (1) In (1), randlnt(a, b) generates a random integer in the range from a to b uniformly, min(c, d) gives the smaller value of c and d, retry is the number of retransmission attempts, and slot time is a time duration specified by the physical layer parameters. During backoff, the station decreases its backoff counter by one i f the medium is idle for a slot time period and freezes the backoff counter when the medium is busy. When the backoff counter reaches zero, the station wi l l transmit its M P D U immediately. When a destination station receives the M P D U successfully, it returns an Acknowledgement ( A C K ) frame to the source station after a Short Interframe Space (SIFS) duration. D C F offers an optional means of transmitting data frames that requires the transmission of Request To Send (RTS) and Clear To Send (CTS) frames prior to the transmission of the actual data frames. The R T S / C T S transmission scheme can alleviate the hidden terminal problem and can reduce the transmission time wasted as a result of a coll ision due to the longer frame size of the actual M P D U . The R T S and C T S frames include information regarding the transmission time of the next data frame and the corresponding A C K frame. The Network Al location Vector ( N A V ) maintained by each station is an 13 indicator of time periods when other stations close to the transmitting station and hidden stations close to the receiving station w i l l not commence any transmissions. D C F with the R T S / C T S transmission scheme and the N A V settings of other stations are shown in Figure 2.6. SIFS H—H Source RTS DATA SIFS <—W Destination CTS SIFS SIFS DATA SIFS ^ >\ Other NAV (RTS) I i i i i i i Contention Window i i i i i i NAV (CTS) Defer Access Figure 2.6. RTS /CTS/da ta /ACK and N A V settings Backoff after defer 2.3 New 802.1 le EDCF Protocol The new 802.1 l e E D C F medium access scheme is governed by a distributed mechanism very similar to 802.11 D C F . Service differentiation is achieved through the introduction of Traffic Categories (TCs). Each T C has a different transmission queue and each transmission queue has a different interframe space (Arbitrary Interframe Space - AIFS[TC]) , a different set of contention window limits (CW m i „ [TC ] and C W m a x [ T C ] ) , and a different persistent factor (PF[TC]). 14 Figure 2.7 illustrates the service differentiation accomplished by using different A IFS values. Each T C within a station starts a backoff independently after detecting the channel as idle for an AIFS[TC] duration. In the E D C F retransmission scheme, the size of the new CW[TC] after an unsuccessful transmission is determined by expanding/reducing the size of the previous CW[TC ] by a factor of PF[TC] , whereas in legacy 802.11 D C F , C W is always double after every unsuccessful transmission, i.e. PF=2. A s in legacy D C F , the C W [ T C ] never exceeds its maximum bound C W m a x [ T C ] . A random backoff counter is chosen from the interval [0, CW[TC] ] in the case of AIFS[TC]>DIFS and from [1, CW[TC]+1] in the case of AIFS[TC]<DIFS [18]. When the backoff counter of a T C reaches zero, the station transmits a pending M S D U from the corresponding transmission queue. A short A I F S , a small size of C W limits, and a low PF value are associated with high priority packets, enabling them to start contenting the channel earlier and to complete the backoff sooner, thus offering a high probability of winning the contention race. DIFS/AIFS N H Busy Medium AIFS[j] <4 AIFS[ i ] < •) DIFS/AIFS PIFS SIFS H H Contention Window T 7 1 Backoff Window _z l l— Next frame Defer Access < ' • "Slot time Select slot and decrement backoff as long as medium is idle Figure 2.7. Service differentiation by different A IFS values 15 Within a station, up to eight transmission queues are realized as virtual stations. See Figure 2.8. Should the backoff counters o f two or more parallel TCs in a single station count down to zero at the same time, a scheduler, which resides in the station, resolves the virtual coll ision by allowing the highest priority T C among the virtually collided TCs to transmit its M P D U [19]. The other virtually collided TCs execute the retransmission mechanism independently, as i f a coll ision had occurred. 802.11 one backoff instance transmission attempt PO PI 802.1 le: up to 8 independent backoff instances P2 P3 P4 P5 P6 Scheduler (resolves virtual collision) transmission attempt Figure 2.8. Transmission architecture of E D C F vs. D C F P7 1 r ! • 1 r r r r 1 r r r r Backoff j Backoff Backoff Backoff Backoff Backoff Backoff Backoff Backoff DIFS | AIFS[0] AIFS[1] AIFS[2] AIFS[3] AIFS[4] AIFS[5] AIFS[6] AIFS[7] CW 1 CW[0] CW[1] CW[2] CW[3] CW[4] CW[5] CW[6] CW[7] PF =2 | PF[0] PF[1] PF[2] PF[3] PF[4] PF[5] PF[6] PF[7] ] r r r 1 T 1 r 1 r 1 f 1 f Since 802.1 le is a draft standard presently under review, many issues are still unsolved and are expected to change [8]. We assume that E D C F described here w i l l not undergo any major modifications. The PF parameter mentioned in this section was proposed 16 in.a previous version of the drafts but has been removed from the current draft [20]. In the current version of E D C F , after each unsuccessful transmission, CW[TC ] is doubled while remaining less than C W m a x [ T C ] , i.e. PF[TC] = 2.0 which is equivalent to the conventional B E B algorithm. 17 Chapter 3 Age-Dependent Backoff Algorithm for 802.1 le EDCF The previous chapter has reviewed the 802.11 standard and has introduced the new 802.1 l e E D C F protocol, proposed by the 802.11 Task Group E, to provide service differentiation at the M A C layer of 802.11 wireless L A N s . The retransmission mechanism of the E D C F protocol causes large delay and jitter for real-time traffic. We now propose a retransmission algorithm called Age-Dependent Backoff (ADB) to alleviate packet delay, jitter and drop rate for real-time traffic. The proposed A D B algorithm for the new 802.1 l e E D C F is presented in this chapter. The backoff behavior of 802.1 le E D C F is described in Section 3.1. Details of A D B are provided in Section 3.2. Finally, Section 3.3 discusses the complexity of implementing the A D B algorithm. 3.1 Backoff Behavior of 802.1 le E D C F In legacy 802.11 D C F , the B E B algorithm doubles C W with every transmission retry to reduce the coll ision probability in the next retransmission by providing greater transmission spacing among stations with pending M P D U s . C W becomes extremely large after successive retransmissions, which cause long delays and large jitter. To reduce delay and jitter, a smaller C W should be employed in the next retransmission thereby providing a better chance for retransmitted packets to access the medium. The wait for new arrivals would increase [21]. In a previous version of the 802.1 le draft, E D C F utilizes a multiplier, P F , to govern the adjustment of C W after an unsuccessful transmission. PF should be less than 1 for time- sensitive applications. However, collisions are a result of congestion and a wider C W is 18 desirable to alleviate congestion. Reducing the C W size for every retransmission causes heavy congestion leading to more collisions. Values of PF between one and two would be preferable. Use of different PFs for different TCs contributes to service differentiation. The current version of E D C F employs the B E B algorithm to resolve collisions. L ike D C F , B E B causes large delay and jitter values that are problematic for real-time traffic with time-bounded requirements. The situation worsens when channel conditions are bad or when the traffic load becomes heavy. 3.2 Age-Dependent Backoff Algorithm To alleviate the problem of the retransmission mechanism employed in 802.1 l e E D C F protocol, we now present the proposed Age-Dependent Backoff (ADB) algorithm for high priority real-time traffic. The idea of A D B is to dynamically adjust PF based on two factors, namely the age of a real-time packet in the transmission queue and the lifetime of the real- time packet. The relationship between the new C W , newCW[TC], and the old C W , oldCW[TC], after a coll ision is shown in (2). newCW[TC] = ((oldCW[TC] +1) x PF[TC]) -1 (2) where PF[TC] is given in (3). PF[TC] = 2 Age+ 2 (3) LT[TC] 19 Age is the packet's age in the transmission queue and LT[TC] is the lifetime of the packet. Figure 3.1 shows the value of PF[TC] as the age of a real-time frame is varied. The newCW[TC] never exceeds the parameter C W m a x [ T C ] but can be less than C W m j n [ T C ] to provide differentiation between retransmitted packets and new arrivals. Real-time packets are obsolete i f they are not received by recipients within their lifetime. Packets with queuing delay longer than the lifetime wi l l eventually be discarded by their applications and should not contend for the medium. Therefore, packets with Age > LT are discarded before attempting transmission to save bandwidth and to prevent causing additional delay to other packets. PF[TC] P-Age LT[TC]/2 LT[TC] Figure 3.1. PF[TC] vs. Age It can be seen in (2) and (3) that the new C W is expanded by the factor 1 < PF < 2 in the first half of a packet's lifetime and compressed by the factor 0 < PF < 1 in the second half. During the first half of the lifetime, A D B allows the backoff to increase gradually while 20 avoiding high coll ision probability, but at the same time precluding a huge increase of delay and jitter. During the second half of the packet's lifetime, the backoff decreases slowly from the expanded C W to raise transmission probability, thereby preventing packets from being dropped. It should be emphasized here that the A D B algorithm is used together with the B E B algorithm in 802.11 wireless L A N s . The A D B algorithm is for the high priority real-time traffic with time-bounded requirements, while the B E B algorithm is for the best-effort traffic with tolerance of large delay and jitter. A D B provides higher access priority than B E B , since A D B expands C W by a factor between 1 and 2 or reduces the C W by a factor between 0 and 1, whereas B E B always expands C W by a factor of 2. A D B contributes to service differentiation by allowing a real-time frame with an age close to its lifetime to have higher priority for channel access than does a new arrival real-time frame. 3.3 Implementation Complexity The A D B retransmission mechanism is easy to implement. It requires minor modifications in the computation of C W thereby minimizing the migration effort from the new 802.1 le E D C F mode. Our A D B retransmission strategy provides backward compatibility to the 802.11 D C F protocol. 3.3.1 Implementation of the ADB Algorithm Implementing the A D B algorithm is relatively simple. To keep track of the ages of 21 the frames in each transmission queue, every pending frame in the queue requires a time stamp to record its arrival time from the upper layer. A temporary timestamp field can be appended to the 802.11 M A C frame to hold the arrival time. The timestamp field should be removed just before the actual transmission to keep the protocol backwards compatible with the 802.11 standard. The variable Age in (2) is calculated by the current time o f the system minus the time stamp of the frame at the head of the queue. If Age > LTVTC], the corresponding frame wi l l be dropped and the next frame in the queue wi l l be evaluated. The process continues until a frame with Age < LT\TC] is encountered or until there are no frames in the queue. If the queue holds no frames, the corresponding queue wi l l not contend for the access of the channel. The newCW[TC] is determined using (2) and (3). The smaller value of newCW[TC] and C W m a x [ T C ] is chosen to be the C W size. After the new C W size has been determined, the value of oldCW[TC] is updated and wi l l be used for the calculation of the next C W size. The f low chart shown in Figure 3.2 illustrates the implementation of the A D B algorithm. 22 S T A R T A collision occurred in previous transmission and the channel is idle for a period of AIFS[TC]. Need to adjust the new C W. Calculate Age Age = current time - timestamp YES YES Drop the frame NO Calculate PF\TC\ PP\TC] = (2/LT[TC]) x Age r Calculate newCW\TC\ newCW[TC\ = ((oldCW[TC\+\) x PF[TC\) - 1 r NO END The corresponding queue will not contend for the access of the channel if newcw\rc\ > cw^irq ? Check to see if the new CW is larger^ than the maximum size^ YES END Starts backoff Set newCW\TC\ to the maximum size newCW[TC\ = CWmax[TC] NO r Update oldCW\TC\ oidcwyrc] = newCW[TC] 1 r Figure 3.2. Flow chart for the implementation of the A D B retransmission algorithm 23 3.3.2 Discussion of the Implementation Complexity The implementation of the A D B algorithm described in the previous section is practical with current software and hardware technologies and requires very few additional resources. Most of the modern microprocessors are equipped with real-time clocks offering timer functions; these can be used to complete the timestamp field and to provide system current time. The calculation of Age requires one subtraction. One comparison is needed to check i f Age > LT\TC] and one register is required to buffer the parameter LT\TC\ for each transmission queue. The calculation of PF[TC] in (3) requires one division, one multiplication, and one addition and the calculation of newCW[TC] in (2) introduces one addition, one subtraction, and one multiplication. The additional resources mentioned above are necessary to implement the A D B algorithm. The addition, subtraction, multiplication, division, and comparison operations can be done using digital hardware circuits or software instruction codes. The A B D algorithm can be implemented with minor modification of the existing hardware or the embedded software of the wireless cards. 24 Chapter 4 Design of Simulation Models The discussion in the last chapter has presented the problem of the retransmission mechanism employed in the new 802.1 le E D C F protocol and has introduced our proposed A D B algorithm for real-time traffic in 802.1 l e E D C F . To evaluate the performance o f A D B against B E B , several verification methods can be used. One method is to implement the A D B algorithm in wireless cards and to then measure the delay, jitter and drop rate of real-time packets from an actual network. This is definitely a high-cost and time-consuming approach. A n alternative efficient method is to use computer simulation. We fol low this widely used method, based on O P N E T simulations, to evaluate the performance of A D B against B E B . In this chapter, we review the O P N E T simulation tools in section 4.1. Section 4.2 describes the simulation models for the 802.11 standard. Section 4.3 discusses the simulation models for the upper layer protocols. Section 4.4 presents the voice, the video, and the data traffic source models used in our work. 4.1 Overview of OPNET Simulation Tools Optimized Network Engineering Tools (OPNET) , licensed by M i l 3 , Inc., is a piece of engineering software capable of simulating large communication networks with detailed protocol modeling. It is widely accepted by researchers to model and simulate complex network scenarios, communication protocols, and traffic models. Whi le O P N E T has many pre-defined models, it allows users to build new models by using finite state machines 25 together with the Proto-C language, a combination of general C/C++ facilities and an extensive library of built-in and high-level subroutines known as Kernel Procedures. OPNET allows users to define network topologies, nodes, and links that describe a network. A simulation can be executed and the results are then analyzed for the network. OPNET has three main types of tools, namely the Model Development tool, the Simulation Execution tool and the Result Analysis tool. These three types of tools are used together to model, simulate and analyze the network. OPNET defines a network model using a hierarchical structure shown in Figure 4.1. The highest level is the network domain, which is constructed from the node domain and the link domain. The node domain is specified in terms of the process domain. Network Domain Node Domain Link Domain Process Domain Figure 4.1. Hierarchical structure for OPNET models 26 The Network Domain defines a network topology described in terms of subnetworks, nodes, l inks, and geographical context. It consists of the nodes and the links which can be deployed within the geographical context. O P N E T provides fixed nodes as well as point-to- point and bus links. In addition, it offers mobile and satellite nodes, and wireless links. The Node Domain provides for the modeling of communication devices that can be deployed and interconnected at the Network Domain. The device models are called nodes and may correspond to various types of computing and communicating equipment in the real world such as routers, switches, bridges, workstations, servers, etc. The Process Domain allows users to create processes using finite state machines and the Proto-C language. These processes are used to express the behavior of protocols, algorithms, and applications. O P N E T comes with built-in model libraries for some common protocols and enables users to focus on the modeling of user-defined protocols. The wireless L A N simulation model is shipped as part of the standard O P N E T model and is based on the I E E E 802.11 standard. Specific parts of the wireless L A N model have been simplified or omitted in view of the fact that it is intended primarily for the performance estimation of the 802.11 D C F protocol. We implement the new 802.1 l e E D C F protocol and the proposed A D B algorithm by modifying the existing wireless L A N model. The detail description of the modification wi l l be presented in the next section. 27 4.2 802.11 Model Design Our 802.11 L A N model is implemented in O P N E T by modifying the existing wireless L A N model to support the new 802.1 le E D C F protocol and our proposed A D B retransmission algorithm. In this section, we describe the 802.11 model from the bottom to the top, i.e. from the physical layer to the M A C layer. 4.2.1 802.11 Physical Layer Model The primary function of the 802.11 physical layer is to transmit a sequence of bits over the wireless medium. The IEEE 802.11 High Rate Direct Sequence Spread Sprectrum (HR-DSSS) physical layer, commonly referred to as 802.1 l b , is a rate extension to the 802.11 D S S S standard. It operates in the 2.4GHz I S M band and includes complementary code keying ( C C K ) to achieve additional data rates of 5.5 and 11 Mbps. 802.11b is the most common wireless L A N implementation today and is interoperable with 802.11 D S S S implementations. We design our 802.11 physical layer model to be similar to the 802.1 lb . The existing wireless L A N model provides three choices for the physical layer configuration in the IEEE 802.11 standard: Frequency Hopping, Direct Sequence, and Infra Red. The wireless L A N model supports data transfer at 1, 2, 5.5 and 11 Mbps. A l l control packets are transmitted at a data rate of 1 Mbps as specified by the standard. These data rates are modeled as the speed of the transmitter and receiver connected to the 802.11 M A C layer. Although the wireless L A N model does not simulate the actual physical layer of the IEEE 28 802.11 specification, the physical layer configuration are needed by the M A C protocol to determine the parameters such as SIFS, DIFS and C W limits. In our study, the modulation technique is relatively unimportant to the service differentiation in 802.11 wireless L A N s . However, the effect of wireless channel noise must be considered due to its impact on bit and packet errors which cause retransmissions. Because of transmission impairments, such as noise and collisions, bit errors can disrupt the sequencing of frames. A station may send a data frame and never receive an A C K . Stations initiating the exchange of frames have the responsibility of error recovery. This recovery involves the retransmission of frames after SIFS sec , i f no response is heard from the destination station. Therefore, a frame corrupted by noise or which has collided with another frame causes execution of the retransmission protocol. The O P N E T Wireless Module allows users to specify the quality of a wireless link based on a number of factors such as the distance between the transmitter and the receiver, the power of the transmitting signal, the transmitter and the receiver antenna gains, the background and the interference noises, etc. O P N E T ' s simulation kernel relies on a 14-stage computational pipeline which uses these factors to emulate the characteristics of the wireless communication channel. The bit error rate can be measured statistically during the simulation. The existing O P N E T wireless L A N model uses the default wireless model settings with some minor modification to conform to the 802.11 standard. We specify the noise figure 29 parameter, the transmitting power, and the distance between the transmitting and receiving stations to achieve a bit error rate of 10"9. We set the error threshold to be zero. Hence, any errors in a packet are discarded by the M A C layer. In order to model the 802.1 l b physical layer, we set the channel data rate at 11 Mbps with direct sequence spread spectrum physical characteristic. With this direct sequence physical characteristic, we specified SIFS = 10 u.s, DIFS = 50 u,s, C W m j n = 31, and C W m a x = 1023. 4.2.2 802.11 M A C Layer Model Based on the existing wireless L A N model in O P N E T , we modify the 802.11 D C F protocol at the M A C layer to develop the new 802.1 l e E D C F protocol with the proposed A D B algorithm. The existing wireless L A N model is intended primarily for DCF-based M A C performance estimation. Many important features of the 802.11 M A C are implemented. The following is a summary of the important features included in the wireless L A N model [22]. • Access Mechanism: Carrier sense multiple access and coll ision avoidance ( C S M A / C A ) D C F access scheme as defined in the standard. • Frame exchange sequence: Data and A C K frame exchange to ensure the reliability of data transfer. Optional R T S / C T S frame exchange is available for media reservation. 30 • Backoff and Deference: The B E B algorithm and the interframe spacing: SIFS, DIFS, and EIFS for D C F implementation. The values for the interframe spacing are selected based on the selection o f physical characteristics. • Channel Data Rate: Date rates of 1, 2, 5.5 and 11 Mbps • Recovery mechanism: Short and Long retry counters as defined in the standard. • Fragmentation and Reassembly: Optional data frame fragmentation based on the size of the data packet received from the higher layer. The fragments are reassembled at the destination station. • Duplicate Packet Detection: Any duplicated packets are discarded by the M A C layer. • Access Point Functionality: Only a wireless L A N router can be configured as an access point in an infrastructure B S S network. In order to form an E S S such that stations within various BSSs can communicate with each other, an IP network needs to be configured. • Buffer Size: Data arrived from a higher layer to the M A C layer is stored in a buffer. The buffer size is limited to the maximum value set by the user. Higher layer packets are dropped, once the maximum buffer size is reached. Some additional features are necessarily added to make that the existing wireless L A N model supports E D C F and A D B . The wireless L A N model has only one transmission queue buffering frames waiting for transmission. A l l packets coming from the upper layer are forwarded directly to that transmission queue. The packets placed in the queue wi l l be transmitted only i f the backoff counter reaches zero and the wireless channel has been idle for 31 DIFS sec. The model has only one backoff counter and one interframe spacing period, DIFS, for all data frame transmissions. Mult iple transmission queues, backoff counters, C W limits and interframe spacing periods are necessary for implementing the E D C F protocol. Packets arriving from the upper layer are forwarded to the corresponding queues based on the traffic types. To simulate a wireless L A N carrying voice, video, and data traffic, we implement three transmission queues for three different traffic types. In order to simplify the implementation, each station is allowed to generate and receive only one type of bi-directional traffic and therefore, using destination M A C addresses is enough to determine the traffic types. A scheduler, which resides in an A P , is implemented to resolve the virtual coll ision when two or more TCs in the access point (AP) count down to zero at the same time. To implement the A D B algorithm, each pending frame is given an arrival timestamp. A l l packets from the upper layer are sent to the corresponding transmission queues with their timestamps which are used to calculate Age, as explained in section 3.3.1. The implementation of A D B in the simulation model is as presented in section 3.3.1. The finite state machine for modeling the 802.11 M A C layer with E D C F and A D B is shown in Figure 4.2. A transition between the B K O F F N E E D state and the I D L E state is added to the finite state machine. This transition is used when the ages of all the packets in the transmission queues are greater than their lifetimes. The expired real-time packets are then dropped from the queues since packets with ages greater than their lifetimes wi l l 32 eventually be discarded by the applications. The station should not contend for the access of the channel and the finite state machine should go back to the I D L E state. ^̂fc< | IHTT 1| »|B33_IHIT| »f IDLE \[* f (̂ d«fault) <FRM_EHD_TO_IDLE> ^ w_(IpLE_AFTER_CFP ) / C ARCEL_DEF_EVEHT; h (VAIT_FOR_FRAME) Tr- | ^ P ^ ^ ^ T ^ P ^ ^ (FRAME_TIWEOUT || FRAME_RCVD) (default) v̂-""" (READ¥JTO_TRAHStolT 6A f&BHJVM_I3_IDLE) y (B ACK_T 0_DEFER)/wlar\_s cite dulêde f e reno e ( ) (PERF 0RM_BACKOFF ^ <" "(FRM_EHD_T 0_DEFER) ! (default) (HIS SED_DEAELIHE)\ \ (DEFEBEflCE_0FF) && !MISSED_DEADLIHE; ^ m b p _ / E E D ] (TRAH SMIT_FRAME &£. (TPJU1 SMIS SI OH_C OBPLETE ) 'fas SED_DEADLXHE ) (BACK_TO_IDLE) \ (PERFORM TRANSMIT)/wlanjperf oim._transifcit_ch«ck () |B ACKOPF jl HTRBHSMIT | (default) (READY_TO_TRflH SMIT MEDIVM_I S_IDLE £•& 0 fP_»P_»G dium_C ontro 1 -= OP C_B 0 OLIHT _DI SAB LED ) .̂'̂  (de£*ult)/wl»n_ijat«rrupts_in£o (); Figure 4.2. Finite state machine for 802.11 M A C layer model with E D C F and A D B The INIT and the B S S I N I T states are used to initialize all the variables, statistics, memory allocations and other entities. When a packet from the upper layer arrives and the channel is idle, the state machine wi l l go directly to the T R A N S M I T state to transmit the packet. If a packet arrives but the channel is busy, the state machine w i l l move to the D E F E R state to defer for a period of D IFS/AIFS after the channel becomes idle. After the deference described above, the state machine wi l l jump to the B K O F F N E E D state. A C W size is then calculated using the B E B or the A D B algorithm. 33 Once the C W size is determined, the state machine w i l l go to the B A C K O F F state to udergo the backoff operation. When a backoff counter reaches zero, the state machine w i l l move to the T R A N S M I T state to transmit the packet. After finishing the transmission, the transmitting station wi l l wait for an A C K frame from the receiving station to confirm the success of the frame reception. The state machine wi l l jump to the F R M E N D state and then to the W A I T F O R R state to wait for the A C K frame. If the A C K frame arrives successfully, it w i l l go either to the I D L E state through the F R M E N D state i f there are no pending frames in the transmission queues or to the D E F E R state through the F R M E N D state to start the next transmission i f the transmission queues are not empty. Table 4.1 shows the simulation parameters for the M A C layer protocol. In our simulation scenarios, detailed in Chapter 5, we assume that the simulation network consists of one B S S with no hidden stations, i.e. stations can hear each other in the B S S . The R T S / C T S optional is disabled since it has been indicated that the R T S / C T S mechanism produces very limited advantages with no hidden stations [23]. Moreover, most of the default settings o f commercially available wireless cards are without the R T S / C T S mechanism. The short retry limit gives the maximum number of transmission attempts for frames whose size are less than or equal to R T S threshold, while the long retry limit specifies the maximum number o f transmission attempts for frames whose size exceeds the R T S threshold. When the R T S / C T S is disabled, the short retry limit is used for all transmissions. Frames that could not be transmitted successfully after reaching the short retry limit are discarded by the M A C layer. 34 Parameters Values Slot time 20 p.s SIFS 10 us DIFS 10u.s + 2 x 2 0 u.s = 50 \is AIFS[1] (voice) 10|j.s + 2 x 2 0 p.s = 50 p,s AIFS[2] (video) 10 u.s + 3 x 20 u,s = 70 [is AIFS[3] (data) 10p.s + 5 x 2 0 p,s=110p,s rCWmin, CWmax l (DCF) T31, 10231 [CWmin[ l ] , CWmaxI/l]] (voice) r 7 , 3 i i [CWmin[2], CWmax[2H (video) [15,63] rCWmin[3], CWmax[311 (data) [15,255] Short Retry Limit 255 Long Retry Limit 255 Table 4.1. Parameters for the M A C layer The A D B algorithm is used for the retransmission mechanism for real-time traffic with time-bounded requirements. The lifetime of a real-time packet must be specified to allow A D B to calculate the new C W size after every unsuccessful transmission. We now discuss the lifetimes of voice and video real-time packets. The main performance parameters for voice traffic are packet delay, jitter, and loss. The end-to-end delay for IP telephony ranges from 300 to 1000 ms. User tolerance o f delay varies significantly. Demanding users require delay of 200 ms or less, while more patient users wi l l accept a delay of 300 to 800 ms [24]. Since the wireless L A N represents only a single hop of an end-to-end connection, we consider 25 ms as the maximum acceptable value for the voice packet transfer delay over the wireless L A N and let LT for voice packets be 25 ms. 35 We assume that 75 ms is the maximum allowable video packet delay at a single wireless hop; thus LT = 75 ms for video packets. 4.3 Simulation Models for Upper Layer Protocols The existing O P N E T wireless model includes node models which correspond to various real world wireless L A N equipment types such as wireless stations and wireless routers. By replacing the original M A C layer with our modified M A C layer presented in the previous section, we developed wireless station and router node models supporting the new 802.11 E D C F protocol and the proposed A D B algorithm. The upper layer protocols such as the Internet Protocol (IP), the Transport Layer Protocol (TCP), the User Datagram Protocol (UDP) and the application layers are unmodified. In this section, we introduce the basic functions and parameters chosen for these upper layer protocols from bottom to top, i.e. from the network layer to the application layer. 4.3.1 Network Layer Model: Internet Protocol IP is a connectionless network level protocol interconnecting multiple networks, possibly of different types. In the Internet Protocol Suite, IP provides best-effort delivery services to T C P and U D P and is situated beneath the transport layer o f the OSI seven-layer model [13]. In turn, IP relies on the services of the data link layer, such as the 802.3 Ethernet and the 802.11 wireless L A N , to relay packets to other IP modules. The packets that are created and forwarded by IP modules are called datagrams. Because IP connects different types of networks that may support different maximum transfer 36 units (MTUs) , fragmentation is performed by the IP modules whenever they attempt to forward a datagram over an interface whose datagram length exceeds its M T U . Reassembly functions are only performed on datagrams when they reach their final destination. The M T U of 802.11 wireless L A N s is 2304 bytes. We configure the IP modules to perform fragmentation when forwarding a datagram with size larger than 2304 bytes over an 802.11 interface. Since the size of the data packets from the IP layer cannot be larger than the 802.1 l ' s M T U , the fragmentation and reassembly option at the 802.11 M A C layer can be disabled. 4.3.2 Transport Layer Models: T C P and UDP Two transport layer protocols, T C P and U D P , are build on the best-effort service, provided by IP, to support a wide range of applications. In this section we discuss the properties of T C P and U D P and the parameters chosen in our simulations. U D P is a very simple but unreliable, connectionless transport layer protocol that provides only two additional services beyond IP: demultiplexing and error checking on data. IP delivers packets to a host but does not know how to deliver them to the specific application in the host. U D P adds a mechanism that distinguishes among multiple applications within the host. IP only checks the integrity of its header. U D P can optionally check the integrity of the entire U D P datagram. If a datagram is found to be corrupted, it w i l l be simply discarded and the source U D P entity wi l l not be notified. 37 T C P is a transport layer protocol, providing a connection-oriented, reliable, in- sequence, byte-stream service. T C P offers f low control that allows receivers to control the rate at which the sender transmits information so that the receiver buffers do not overflow. T C P also provides congestion control which ensures that the sender does not waste resources by sending more traffic than the network can successfully carry to the receiver. Similar to U D P , T C P supports multiple applications in a host and error checking on T C P segments. T C P and U D P both have the ability to establish a host-to-host communication channel for delivering packets between applications running in two different stations. The main difference is that T C P provides reliability, and f low and congestion control services, while U D P trades of f those services to improve performance. Application protocols can choose T C P or U D P at the transport layer. The Fi le Transfer Protocol (FTP) uses T C P to ensure that an exact copy of a file is delivered to the recipient. Real-time applications such as voice and video conferencing use U D P because they can tolerate some errors or data loss. O P N E T ' s T C P module is based on the T C P specified in R F C 793 and R F C 1122. The model incorporates some important features such as end-to-end reliability based on acknowledgment, retransmissions triggered by exponentially backed-off timers, f low control based on the availability of remote buffering resources, reordering of data that arrives out of sequence, connection establishment and closing through three-way handshake protocols, and "Slow-start" congestion avoidance and control [25]. The model allows users to control the T C P behavior by selecting the T C P parameters. We choose the default setting o f the T C P parameters. Some of the important default parameter settings are shown in Table 4.2. 38 Maximum Segment Size 2272 bytes Receive Buffer 8769 bytes Max imum A C K Delay 0.200 sec Slow-Start Initial Count 1 M S S Fast Retransmit Enabled Fast Recover Reno Karn 's Algorithm Enabled Initial R T O 3.0 Min imum R T O 1.0 Max imum R T O 64 R T T Gain 0.125 R T T Deviation Gain 0.25 R T T Deviation Coefficient 4.0 Timer Granularity 0.5 sec Persistence Timeout 1.0 sec Table 4.2. Default settings of the T C P parameters The maximum segment size is set to 2272 bytes so that the underlying 802.11 network can carry traffic without any fragmentation. With the fast retransmit algorithm enabled, after receiving three duplicate A C K s , the T C P sender infers that a segment is lost and re-transmits the missing segment without waiting for its retransmit timer to expire [26]. This allows T C P to detect and retransmit a lost segment more quickly than for time-out retransmission. T C P Reno employs Karn's algorithm [27], which requires round-trip estimates to be updated only for the A C K s of those segments that have been transmitted only once. T C P uses frequent measurements of the round trip timer (RTT) to dynamically adjust the retransmission timeout (RTO). When the R T O expires before an A C K is received, retransmission occurs starting with the first byte of unacknowledged data. The R T T gain and R T T deviation gain are constants between 0 and 1 and control how fast the R T O adapts to changes in network traffic. The R T T gain and the R T T deviation gain are recommended to be 0.125 and 0.25 respectively 39 [28]. For further information regarding the default setting, please refer to the TCP Model Description of OPNET user's manual [25]. In OPNET's UDP model, there are no parameters to specify the behavior of UDP since UDP itself is a very simple protocol. If the application layer sends a packet with size larger than the maximum payload size of a UDP datagram, 65535 bytes - 8 bytes = 65527 bytes, the packet will be dropped. 4.4 Traffic Source Models We use OPNET's standard network application model for the traffic source models. The standard network application model is a simple, general model of client and server or a peer-to-peer network application. Its behavior can be modified through parameter settings to make it emulate a wide variety of network applications. It does not, however, model the behavior of any particular application in detail. The parameters of the standard network application model are selected by carefully emulating the environment in which 802.11 M A C would most probably be applied. With this in mind, we choose numerical parameters in such a way that the traffic sources reflect as closely as possible the network load conditions of an actual real-life situation. The simulations consider three types of traffic sources: voice, video conferencing and FTP traffic. Each wireless station runs only one session and all sessions are bi-directional, i.e. each station is the source of an uplink flow and the sink of a downlink flow for the session it 40 runs. Voice and video conferencing traffic have higher priority than F T P traffic. Section 4.4.1 and section 4.4.2 describe the voice and video conferencing models, respectively. The best-effort F T P file transfer model is presented in section 4.4.3. 4.4.1 Voice Model Voice traffic is known to have a two-state O N / O F F behavior, where voice users are either in talkspurt or silence as shown in Figure 4.3. For efficient usage of wireless bandwidth, silence suppression is employed. Hence, voice packets are generated in talkspurt (ON), while no packets are generated during silence (OFF). Both the duration of talkspurt and of silence fol low the exponential duration with the mean duration equal to 1 and 1.35 seconds respectively [29]; thus a voice source is in talkspurt mode approximately 42.6% of the time. Each voice station runs only one bi-direction voice session over UDP/ IP . Figure 4.3. Two-state model of a voice user The G.729 speech codec is selected to model good-quality voice calls, with 8-kbps coding rate, 10-byte packets generated in every 10 ms during talkspurt, 10 ms processing delay, and 5 ms lookahead delay. The processing delay is the delay required to run the 41 encoding algorithm on a single frame and the lookahead delay is the amount of time needed to delay the speech for look ahead calculations for the encoder frame. The encoder buffers data for this extra time to enable prediction, while compressing data. The one-way latency of the encoder is the sum of the frame size, processing delay, and lookahead delay and therefore, the total codec delay for G.729 is 25 ms [30]. 4.4.2 Video Conferencing Model The video conferencing application in O P N E T ' s standard network application model is used to generate streaming video frames across the wireless L A N . It allows users to control the properties of streaming video through the application parameters. Table 4.3 shows the parameters selected for modeling a low-quality video conferencing application characterized by a relatively low bit rate of 128 kbps for both the uplink and downlink Parameters Description Values Incoming Stream Interarrival Time Time between frames generated within a video conferencing session from the destination 20 frames/sec Outgoing Stream Interarrival Time Time between frames generated within a video conferencing session from the source 20 frames/sec Incoming Stream File Size Average size of an incoming video frame Exponentially distributed with a mean of 800 bytes Outgoing Stream Fi le Size Average size of an outgoing video frame Exponentially distributed with a mean of 800 bytes Table 4.3. Parameters for video conferencing 42 UDP transport protocol is used for the video conferencing application. A video station runs one bi-directional video session over UDP/IP. 4.4.3 FTP File Transfer Model An FTP application enables file transfers between a client and a server. The FTP application in OPNET's standard network application model is used to generate best-effort data traffic across the network. FTP has two basic commands for transferring a file: Get and Put. The Get command triggers the file transfer from the server and the Put command sends a file from the client to the server. FTP uses the connection-oriented TCP transport protocol. A new transport connection is opened for each file transfer. Table 4.4 shows the parameters chosen in the FTP application. Parameters Description Values Command Mix (get/total) Ratio of the Get (download) commands to the total number of commands (sum of Get and Put commands) 50% Inter-Request Time Time between subsequent file requests Exponentially distributed with a mean of 0.02048 second File Size Size of a file being transferred 1024 bytes Table 4.4. Parameters for FTP file transfers 43 Using the above parameter values, we assume that the upload and the download streams are identical with 2 0 0 kbps of FTP traffic in each stream. The rate at which files are requested is independent of the response received; that is, the second request can initiate without the first response being received. 44 Chapter 5 Performance Analysis of the Proposed ADB Algorithm in an Ad-Hoc Network and a Hotspot The previous chapter has introduced the OPNET simulation models. We will use these models to evaluate the performance of the new 802.1 le EDCF protocol with our proposed ADB retransmission algorithm. The behavior of EDCF without ADB is presented for comparison. This chapter presents simulation parameters and scenarios, results from the simulations, and discussions of the results. Two typical environments of 802.11 wireless LANs are considered, namely the ad-hoc network and the hotspot scenarios. In the ad-hoc network scenario, different services such as voice, video and data services may be active simultaneously in an ad-hoc network. The ad- hoc network is an independent BSS in a restricted space where stations can hear each other. The hotspot scenario is conceived where voice, video, and data users are connected to wireline networks through an AP in a single BSS. A hotspot is a specific type of infrastructure mode installation in which a commercial entity provides wireless Internet access to the public. Any reference herein to "hotspot" installations would apply equally to other infrastructure mode installations. Section 5.1 describes the ad-hoc scenario and section 5.2 presents the hotspot scenario. Both sections provide simulation results as well as discussions concerning the results. All results presented in this chapter are the average of results collected from simulating each of 45 the scenarios for 180 seconds for each of 5 different seed numbers. Section 5.3 provides a conclusion regarding the simulation results. 5.1 Ad-Hoc Network Scenario This section presents the ad-hoc network scenario. The overview of the ad-hoc network scenario is introduced in section 5.1.1. The simulation results are given in section 5.1.2. The discussions are provided in section 5.1.3. 5.1.1 Overview of the Ad-Hoc Network Scenario One environment in which 802.11 wireless LANs is likely to be applied is in an ad- hoc network. Such situation is appropriate for, assemblies, outdoor activities, rescue operations, or major disaster areas, where users need to deploy networks to start communication immediately without the benefit of a fixed network infrastructure. Figure 5.1 shows the topology of the ad-hoc network. For simplicity, we assume that there are no hidden stations and that the distance between every pair of transmitting and receiving stations is the same. The bit error rate of each wireless link is therefore assumed to be constant and equal. In the ad-hoc network, voice, video conferencing and best-effort FTP file transfer services are active simultaneously. We conceive that the ad-hoc network is an independent BSS with 10 voice, 4 video and n FTP client and server stations where n is variable. Since 802.1 le EDCF is designed to 46 be backwards compatible to 802.11 DCF, we assume that half of the FTP client and server stations are using the 802.11 DCF while the other half are using the new 802.1 le EDCF protocol with the BEB algorithm with PF value of 2.0. We simulate the ad-hoc network to compare, for real-time traffic, the performance of our ADB algorithm with that of the generalized BEB algorithm with PF = 1.5 and 2.0. Best-effort FTP data traffic always employs the conventional BEB algorithm where PF = 2.0. Independent basic sevice set n/4 FTP clients ^ (802.1 le EDCF) n/4 FTP servers (802.lie EDCF) 200 kbps of FTP traffic per flow ^ n/4 FTP clients (802.11 DCF) 802.11 ad-hoc network Figure 5.1. Network topology of an ad-hoc scenario 5.1.2 Simulation Results from the Ad-Hoc Network Scenario This section presents the results of the ad-hoc network scenario. Voice packet delay, jitter, and drop rate are shown in Figures 5.2, 5.3, and 5.4 respectively. We define the jitter as 47 the delay variance and the drop rate as the percentage of packets with delay longer than their lifetime. The lifetime of a voice packet is assumed to be 25 ms. 12 16 20 number of FTP stations 24 28 -ADB • EDCF (PF=2.0) • EDCF (PF=1.5) Figure 5.2. Voice packet delay in an ad-hoc network 0.0007 12 16 20 number of FTP stations 24 28 -ADB • EDCF (PF=2.0) • EDCF (PF=1.5) Figure 5.3. Voice packet jitter in an ad-hoc network 48 12 1 6 2 0 number of FTP stations - A D B - E D C F ( P F = 2 . 0 ) — A — E D C F ( P F = 1 . 5 ) Figure 5.4. Voice packet drop rate in an ad-hoc network Figures 5.5, 5.6, and 5.7 show delay, jitter and drop rate, respectively, for video packets. The lifetime of a video packet is assumed to be 75 ms and therefore, a video packet is dropped if its age exceeds 75 ms. 8 1 2 1 6 2 0 number of FTP stations 2 4 2 8 - A D B - E D C F ( P F = 2 . 0 ) — A — E D C F ( P F = 1 . 5 ) Figure 5.5. Video packet delay in an ad-hoc network 49 Figure 5.6. Video packet jitter in an ad-hoc network The main performance characteristic of the best-effort F T P traffic is measured by throughput. The total F T P traffic throughput for the ad-hoc network is shown in Figure 5.8. number of FTP stations - A D B • E D C F ( P F = 2 . 0 ) • E D C F ( P F = 1 . 5 ) Figure 5.7. Video packet drop rate in an ad-hoc network 50 7 0 0 0 0 0 g- 6 0 0 0 0 0 5 5 0 0 0 0 0 Q. o> 4 0 0 0 0 0 1 3 0 0 0 0 0 •is 2 0 0 0 0 0 « 1 0 0 0 0 0 o 0 4 8 12 16 2 0 2 4 2 8 number of FTP stations — • — A D B - » - E D C F ( P F = 2 . 0 ) - A - E D C F ( P F = 1 . 5 ) Figure 5.8. Total FTP data throughput in an ad-hoc network 5.1.3 Discussions of the Ad-Hoc Network Results Our simulation results demonstrate that EDCF with the ADB algorithm outperforms EDCF with the generalized BEB algorithm for PF = 1.5 and 2.0 in all aspects of network performance that we have tested. The voice packet delay, jitter and drop rate are significantly lower when using ADB as illustrated in Figures 5.2, 5.3, and 5.4. Relative to BEB, ADB also demonstrates considerable improvements in the video packet delay, jitter, and drop rate as depicted in Figures 5.5, 5.6 and 5.7. The relative improvements in voice and video traffic become more noticeable and pronounced as the number of FTP stations increases. ADB enhances the performance of the 51 voice and video traffic without sacrificing the throughput of the best-effort FTP traffic; as well, ADB prevents the FTP traffic from being starved as shown in Figure 5.8. ADB dynamically adjusts the change of CW sizes based on the ages and lifetimes of voice and video packets to avoid long delay and to prevent excessive collisions under heavy traffic load. As a result, the delay, jitter and drop rate of both voice and video packets are improved considerably. These enhancements in voice and video packet performance do not adversely affect the FTP traffic throughput. Accordingly, we can conclude that ADB enables more efficient use of the wireless channel and improves the overall performance of the ad-hoc network. 5.2 Hotspot Scenario The subject of this section is the hotspot scenario. An overview of this scenario appears in section 5.2.1. Simulation results are given in section 5.2.2, and discussions are provided in section 5.2.3. 5.2.1 Overview of the Hotspot Scenario A hotspot is another environment where 802.11 wireless LANs would likely be deployed. A hotspot is a geographic area covered by wireless networks, with Internet access being available to those devices with wireless network cards. 52 Most of the hotspots today are covered by 802.1 lb wireless LANs to provide wireless Internet access services for public use. The network topology associated with the hotspot scenario is depicted in Figure 5.9. Figure 5.9. Network topology of a hotspot scenario The hotspot scenario is developed to study the performance of the ADB algorithm in a hotspot environment. For simplicity, we assume that there is only one AP in the hotspot and that there are no hidden stations in a BSS. The distance between each wireless station and the AP is the same and constant, in which case the bit error rate of each wireless link is assumed to be identical. Voice, video conferencing and best-effort FTP file transfer services are active simultaneously in the hotspot. Because users in the hotspot environment rarely make phone 53 or video conferencing calls, or rarely transfer files to their neighbors, all types of traffic generated within the BSS are delivered to their peers or servers in the wireline network through the AP. As well, all types of traffic destined for wireless stations are also via the AP from the wireline network. The upstream and the downstream traffic loads for all types of services are assumed to be the same. We assume that the hotspot consists of 6 voice, 2 video, and n FTP client stations. Since 802.1 le EDCF is designed to be backwards compatible to 802.11 DCF, we assume that half of the FTP clients are using the 802.11 DCF while the other half are using the new 802.1 le EDCF protocol with the BEB algorithm with PF value of 2.0. We simulate the hotspot scenario to compare the A D B algorithm for real-time traffic against the generalized BEB algorithm with PF values of 1.5 and 2.0 for real-time traffic. Again, best-effort FTP data traffic employs the conventional BEB algorithm with PF = 2.0. 5.2.2 Simulation Results from the Hotspot Scenario This section presents results from simulations of the hotspot scenario. Voice packet delay is shown in Figure 5.10. Uplink and downlink voice packet jitter is shown in Figure 5.11a and 5.11b respectively, where jitter is defined as the delay variance. The uplink and downlink voice drop rate is shown in Figure 5.12a and 5.12b, respectively. The lifetime of a voice packet is assumed to be 25 ms. 54 0 . 0 2 1 0 1 2 14 1 6 1 8 number of FTP clients 2 0 2 2 - • — A D B - U p l i n k • E D C F ( P F 2 . 0 ) - U p l i n k - A — E D C F ( P F = 1 . 5 ) - U p l i n k —o— A D B - D o w n l i n k • • - E D C F ( P F = 2 . 0 ) - D o w n l i n k — A - E D C F ( P F = 1 . 5 ) - D o w n l i n k Figure 5.10. Uplink and Downlink voice packet delay in a hotspot 55 1 . 2 0 E - 0 3 1 0 1 2 14 1 6 1 8 number of FTP clients - A D B - D o w n l i n k - E D C F ( P F = 1 . 5 ) - D o w n l i n k • E D C F ( P F = 2 . 0 ) - D o w n l i n k (b) Figure 5.11. (a) Uplink voice packet j itter in a hotspot. (b) Downlink voice packet jitter in a hotspot 56 1 0 1 2 14 1 6 number of FTP clients 1 8 2 0 2 2 - A D B - U p l i n k • E D C F ( P F = 2 . 0 ) - U p l i n k • E D C F ( P F = 1 . 5 ) - U p l i n k (a) 1 0 1 2 14 1 6 1 8 number of FTP clients 2 0 2 2 - A D B - D o w n l i n k - E D C F ( P F = 1 . 5 ) - D o w n l i n k • E D C F ( P F = 2 . 0 ) - D o w n l i n k (b) Figure 5.12. (a) Uplink voice packet drop rate in a hotspot (b) Downlink voice packet drop rate in a hotspot 57 Figure 5.13 shows video packet delay. Figure 5.14a and 5.14b show uplink and downlink video packet jitter. Figure 5.15a and 5.15b show uplink and downlink video packet drop rates. 0 . 2 5 1 0 1 2 14 1 6 18 2 0 2 2 number o f FTP clients - A D B - U p l i n k • E D C F ( P F = 2 . 0 ) - U p l i n k — A — E D C F ( P F = 1 . 5 ) - U p l i n k — o— A D B - D o w n l i n k — • - E D C F ( P F = 2 . 0 ) - D o w n l i n k — A - E D C F ( P F = 1 . 5 ) - D o w n l i n k Figure 5.13. Video packet delay in a hotspot 58 0 . 0 0 9 1 0 12 14 1 6 number of FTP stations 1 8 2 0 2 2 - A D B - U p l i n k • E D C F ( P F = 2 . 0 ) - U p l i n k • E D C F ( P F = 1 . 5 ) - U p l i n k (a) 1 2 14 16 number of FTP clients 2 2 - A D B - D o w n l i n k - E D C F ( P F = 1 . 5 ) - D o w n l i n k • E D C F ( P F = 2 . 0 ) - D o w n l i n k (b) Figure 5.14. (a) Uplink video packet jitter in a hotspot (b) Downlink video packet jitter in a hotspot 59 1 0 12 14 1 6 1 8 number of FTP clients 2 0 2 2 - A D B - U p l i n k • E D C F ( P F = 2 . 0 ) - U p l i n k • E D C F ( P F = 1 . 5 ) - U p l i n k (a) number of FTP clients - A D B - D o w n l i n k - E D C F ( P F = 1 . 5 ) - D o w n l i n k • E D C F ( P F = 2 . 0 ) - D o w n l i n k (b) Figure 5.15. (a) Uplink video packet drop rate in a hotspot (b) Downlink video packet drop rate in a hotspot 60 The main performance characteristic of data traffic is measured by throughput. The total throughput of the FTP data traffic in the hotspot network is shown in Figure 5.16. •3- 7 0 0 0 0 0 ~ 6 0 0 0 0 0 3 £ 5 0 0 0 0 0 o 4 0 0 0 0 0 Z 3 0 0 0 0 0 1 0 2 0 0 0 0 0 t 1 0 0 0 0 0 co o 0 -I , , , , , , , 1 6 8 1 0 1 2 14 1 6 1 8 2 0 2 2 number of FTP clients — • — A D B - » - E D C F ( P F = 2 . 0 ) - A - E D C F ( P F = 1 . 5 ) Figure 5.16. Total FTP data throughput in a hotspot 5.2.3 Discussions of the Hotspot Results Figures 5.10, 5.11 and 5.12 show that, relative to EDCF without ADB, voice packet delay, jitter and drop rate improve significantly as the number of FTP stations increases when EDCF incorporates ADB. Figures 5.13, 5.14 and 5.15 demonstrate that with ADB, video packet delay, jitter and drop rate all improve considerably when many best-effort FTP stations are active. 61 Like the simulation results from the ad-hoc scenario, those from the hotspot scenario indicate that when the A D B retransmission algorithm is employed in the new 802.1 l e E D C F protocol, the delay, jitter, and packet drop rate of the voice and video traffic is reduced without sacrificing the throughput of best-effort F T P traffic (see Figure 5.16). The results suggest that A D B enhances QoS performance of E D C F in the hotspot. The E D C F protocol provides service differentiation, which is an important improvement over 802.11 D C F . E D C F presents delay asymmetry, giving an advantage to uplink transmission as shown in Figures 5.10 and 5.13. This is because the aggregate downlink real-time traffic sent by the A P must complete for the channel in equal terms with all the stations that want to transmit in the uplink direction. The A D B algorithm reduces delay asymmetry by giving priority to real-time packets with high age values. Downlink real-time traffic having long queuing delay wi l l be assigned to use small PF value as adjustment to their C W sizes. 5.3 Conclusion from the Simulations Results Results from both the ad-hoc and hotspot scenarios suggest that the A D B retransmission algorithm used in the new 802.1 le E D C F protocol offers improvement, relative to that from use of the B E B algorithm (with real-time traffic) in QoS differentiation, under a wide range of traffic loads, in both ad-hoc networks and in hotspots. The performance of the real-time traffic is enhanced without reducing the throughput of the best- effort traffic. 62 Service differentiation becomes more noticeable and more pronounced as the number n of best-effort stations increases. ADB can also reduce delay asymmetry at the hotspot. The difference between the uplink and the downlink delay is reduced significantly when the ADB algorithm is employed. There is a penalty for using ADB to reduce delay, jitter, and packet drop rate of real- time traffic while maintaining throughput levels of best-effort data traffic. The best-effort traffic delay and jitter increase, in some cases substantially, when the ADB algorithm is employed. However such delay and jitter increase, not presented here, is not regarded as problematic for best-effort data traffic. 63 Chapter 6 Summary and Conclusions The increasing use of wireless and Internet communications has created a strong demand for public Internet access over 802.11 wireless L A N s . IETF is currently working on service differentiation at the IP layer, but the result is sub-optimal without lower layers' support. Since 802.11 wireless L A N s appear everywhere, it is essential to focus on service differentiation support at the 802.11 M A C layer. To improve the current 802.11 M A C protocol to support applications with QoS requirements, the I E E E 802.11 Task Group E was formed and is defining QoS enhancements for 802.11 M A C . The 802.1 l e draft introduces the E D C F protocol, which is a prioritization enhancement of the 802.11 D C F protocol. In the current version of E D C F , the B E B algorithm is employed in E D C F to resolve collisions; however B E B causes long delay and large jitter that are unfavorable for real-time packets with time- bounded requirements. This thesis is an extension to the current work on providing QoS enhancements in the 802.11 M A C layer studied by IEEE 802.11 Task Group E. In particular, we focus on the improvement of the 802.1 l e E D C F protocol by proposing our A D B retransmission algorithm. 64 6.1 Summary of the Work The primary contribution of this thesis is our proposal, analysis, and evaluation of the ADB retransmission algorithm that can alleviate delay, jitter, and drop rate for real- time traffic with time-bounded requirements without reducing the throughput of best- effort data traffic. The ADB algorithm is used together with the BEB algorithm in 802.11 wireless LANs. The ADB algorithm is for high priority real-time traffic, while the BEB algorithm is for best-effort data traffic. The ADB retransmission mechanism is easy to implement. It requires minor modifications in the computation of CW thereby minimizing the migration effort from the new 802.1 le EDCF protocol. Our ADB retransmission strategy provides backward compatibility to the 802.11 DCF protocol. The implementation of the ADB algorithm is relatively simple and is practical with current software and hardware technologies. OPNET simulation models are built to study the performance of the new 802.11 EDCF protocol with the proposed ADB retransmission algorithm in two typical environments which are modeled by two simulation scenarios, namely an ah-hoc network and a hotspot scenarios. The results from both scenarios indicate that using ADB in EDCF provides low delay, jitter and drop rate for real-time traffic. The delay asymmetry which exists without ADB is reduced significantly in the hotspot environment when ADB is employed. The improvements are more noticeable and more pronounced as the data traffic load increases. 65 In conclusion, A D B is a useful retransmission algorithm for the new 802.1 l e E D C F protocol. Since A D B can be implemented by changing the software without alternating the existing hardware, the cost of implementation is considered to be relatively low. 6.2 Future Work To further extend the work of this thesis, the following possible directions for future research are suggested. 1. In our work, it was assumed that all packets are transmitted at the same rate. In reality, some stations may transmit faster than others depending on the wireless link quality. Simulation of the proposed A D B algorithm in a noisy environment is desirable for extending validation of the results presented here. 2. Admission control is a key aspect for the real4ime mechanism to work wel l , since such control limits the amount of real-time traffic admitted to 802.11 wireless L A N s . A n exact design of an admission control scheme for the new 802.11 E D C F protocol with the proposed A D B algorithm requires further investigation. 3. The proposed A D B retransmission algorithm can be applied to the C S M A / C D protocol to support QoS differentiation in 802.3 Ethernet. It would be interesting to study the performance of the 802.3 C S M A / C D protocol with A D B . 66 References [1] William Stalling, "IEEE 802.11: Moving closer to practical wireless LANs", IT Professional, vol. 3, issue 3, pp. 17-23, May-June 2001. [2] Jim Geier, Wireless LANs, Second Edition, Indianapolis, Indiana: Sams Publishing, 2002. [3] IEEE Std. 802.11-1997, Information technology - Telecommunications and information exchange between systems - Local and metropolitan area networks - Specific requirements - Part 11: Wireless L A N Medium Access Control (MAC) and Physical Layer (PHY) Specifications, 1997. [4] IEEE Std. 802.1 la-1999, Information technology - Telecommunications and information exchange between systems - Local and metropolitan area networks - Specific requirements - Part 11: Wireless L A N Medium Access Control (MAC) and Physical Layer (PHY) specifications - Amendment 1: High-speed Physical Layer in 5 GHz band, 1999. [5] IEEE Std. 802.1 lb-1999, Information technology - Telecommunications and information exchange between systems - Local and metropolitan area networks - specific requirements - Part 11: Wireless L A N Medium Access Control (MAC) and Physical Layer (PHY) specifications - Amendment 2: High-speed Physical Layer (PHY) extension in the 2.4 GHz band - Corrigendum 1, 1999 [6] IEEE Std. 802.1 lg-2003, Information technology - Telecommunications and information exchange between systems - Local and metropolitan area networks - Specific requirements - Part 11: Wireless L A N Medium Access Control (MAC) and 67 Physical Layer (PHY) specifications - Amendment 4: Further Higher-Speed Physical Layer Extension in the 2.4 GHz Band. "High on Wi-Fi," Laptop, pp. 36- 47, March 2003. H. L. Truong and G. Vannuccini, "The IEEE 802.1 le M A C for quality of service in wireless LANs," IBM Zurich Research Laboratory, 2002. IEEE 802.11 TGe, "EDCF Proposed Draft Text," TF-01/131rl, March 2001. Antonio Grilo and Mario Nunes, "Performance evaluation of IEEE 802.1 le," Proc. IEEE PIMRC '02, Lisbon, Portugal, pp. 511-517, Sept 2002. George W. Wong and Robert W. Donaldson, "Improving the QoS Performance of EDCF in IEEE 802.1 le Wireless LANs," Proc. IEEE PACRIM '03, pp. 392-396, Aug. 2003. Gil Held, Deploying Wireless LANs, New York: McGraw-Hill, 2002. Alberto Leon-Garcia and Indra Widjaja, Communication Networks, Singapore: McGraw-Hill, 2000. P. Crow, I. Widjaja, J. G. Kim, and P. T. Sakai, "IEEE 802.11 wireless local area networks," IEEE Commun. Mag., pp 116-126, Sept 1997. C. Heide, "CODIAC protocol - Centrialized or distributed integrated access control," IEEE P802.il/93-54, May 1993. M . A. Visser and M . E. Zarki, "Voice and data transmission over an 802.11 wireless network," Proc. IEEE PIMRC '95, Toronto, Canada, Sept. 1995. Wen-Tsuen Chen, Bo-Bin Jian, and Shou-Chih Lo, "An adaptive retransmission scheme with QoS support for the IEEE 802.11 M A C enhancement," Proc. IEEE V T C Spring 2002, Birmingham, USA, pp. 70-74, May 2002. 68 [18] Stefan Mangold, Sunghyun Choi, Peter May and Guido Hietz, "IEEE 802.1 le - Fair resource sharing between overlapping basic service sets," Proc. IEEE PIMRC 2002, Lisbon, Portugal, pp. 166-171, Sept 2002. [19] Stenfan Mangold, Sunghyun Choi, Peter May, Ole Klein, Guido Hiertz, and Lothar Stibor, "IEEE 802.1 le wireless L A N for quality of service," Proc. European Wireless '02, Florence, Italy, Feb. 2002. [20] Lamia Romdhani, Qiang Ni, and Thierry Turletti, "Adaptive EDCF: Enhanced Service Differentiation for IEEE 802.11 Wireless Ad-Hoc Networks," see http://citeseer.nj.nec.com/romdhani02aedcf.html. [21] Mathilde Benveniste, "Tiered Contention Multiple Access (TCMA), a QoS-based distributed M A C protocol," Proc. IEEE PIMRC '02, Lisbon, Portugal, pp. 166-171, Sept. 2002. [22] OPNET Wireless LANs Model Description, OPNET Technologies Inc., 2000. [23] Raffaele Bruno, Marco Conti, and Enrico Gregori, "IEEE 802.11 Optimal performances: RTS/CTS mechanism vs. basic access," Proc. PIMRC '02, Lisbon, Portugal, pp. 166-171, Sept. 2002. [24] Thomas J. Kostas, Michael S. Borellla, Ikhlaq Sidhu, Guido M . Schuster, Jacek Gragiec, and Jerry Mahler, "Real-Time voice over packet-switching networks," IEEE Network, pp. 18-27, Jan. 1998. [25] OPNET TCP Model Description, OPNET Technologies Inc., 2000. [26] Larry L. Peterson and Bruce S. Davie, Computer Networks: A System Approach, Second Edition, Morgan Kaufmann, 1999. 69 [27] P. Karn and C. Patridge, "Improving round-trip time estimates in reliable transport protocols," Proc. A C M SIGCOMM '87, Aug. 1987. [28] V. Jacobson, "Congestion Avoidance and Control," Proc. A C M SIGCOMM, Aug. 1988. [29] P. Brady, "A model for generating on-off speech patterns in two-way conversation," Bell Syst. Tech. Journal, vol. 48, no. 7, pp. 2245-2272, Sept. 1969. [30] Richard V. Cox, "Three new speech coders from the ITU cover a range of applications," IEEE Commun. Mag., pp. 40-47, Sept. 1997. 70

Cite

Citation Scheme:

    

Usage Statistics

Country Views Downloads
Germany 3 7
China 3 21
United States 1 0
City Views Downloads
Unknown 3 9
Beijing 3 0
Ashburn 1 0

{[{ mDataHeader[type] }]} {[{ month[type] }]} {[{ tData[type] }]}

Share

Share to:

Comment

Related Items