Radio link control and transport layer protocol design issues in wireless IP networks

This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer.

The quality of this reproduction is dependent upon the quality of the

copy submitted. Broken or indistinct print, colored or poor quality illustrations

and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction.

In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, th ese will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.

Oversize materials (e.g., m aps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand corner and continuing from left to right in equal sections with small overlaps.

Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order.

ProQuest Information and Leaming

300 North Zeeb Road, Ann Arbor, Ml 48106-1346 USA 800-521-0600

This reproduction is the best copy available.

b\-Abu Zafar M oham m ad Ekram Hossain

M.Sc.(Eng.). Bangladesh University of Engineering and Technology. 1997

.A. Thesis S ubm itted in Partial Fulfillment of the Retjuirenients for the Degree of

D O C T O R OF P H I L O S O P H Y

in the D epartm ent of Electrical and C om p u ter Engineering

We accept this thesis as conforming to the required sta n d a rd

Prof. \ '. K. B h a rgava. Supervisor, b e p t . of Electrical A' C o m p u ter Engineering

---Prof. E .^ l- G u ib a ly . Member. Djzpt. of Electrical A C o m p u ter Engineering

Prof. K. E. Li. Member. Dept, of Electrical A C o m p u ter Engineering

Prof. S. Dost. O utside Me of Mechanical Engineering

Prof. S. Rov. E xternal Exam iner

© .Abu Zafar M oham m ad Ekram Hossain. 2000 L'niversitv of X'ictoria

All rights reserved. This thesis mag not be reproduced in whole or in part by photocopy or other means, without the permission of the author.

Supervisor:

A B ST R A C T

Packet-switched wireless d a ta networks built upon IP (Internet Protocol)-hased infrastru c tu re are being envisioned to provide ubiquitous Internet access to mobile users. S u pporting packet-data services along with the cellular voice services in an integrated networking framework gives rise to new network infrastructure and pro­ tocol design issues th a t are to be resolved to facilitate the introduction of th e next generation wireless IP networks.

T his thesis addresses several protocol design issues in the area of wireless packet d a t a networking, namely, retransmission control design for multichannel protocols, radio link level protocol for dynam ic rate and error control, inter-layer protocol de­ pendency. radio link-layer and transport-layer protocol fairness and radio link-level d ynam ic bandw idth allocation. retransmission control policy for a multichannel S- .A.LOHA (slotted .-VLOH.A.) protocol in a high speed wireless d a t a network is proposed and analyzed. .A. sub-optimal dynam ic rate a d a p ta tio n procedure is proposed for up­ link d a t a transmission in WCDM.A (W ideband Code Division Multiple .Access)-based wireless IP networks. The performance of this scheme is analyzed using a novel "mean- sen.se' approach for interference calculation in cellular WCD.M.A environment. The im pact of macrodiversity packet combining on tran sport-protocol th roughput perfor­ mance is analyzed under different link-level retransm ission control policies. .A unified T D M A (Time Division Multiple .Access)-based centralized bandw idth m anagem ent mechanism is proposed as a link-level solution for providing service fairness am ong com p etin g users for uplink d a t a transm ission in a wireless IP network. T C P (Trans­ mission Control Protocol) performance is evaluated under different transport-level packet scheduling policies in a correlated fading environm ent and a tim e frame-based scheduling policy is proposed to provide sendee fairness am ong mobile users in the case of downlink transmission. A set of centralized burst-level bandw idth alloca­ tion policies are investigated as a means of service integration with QoS (Q uality of Service) provisioning in the wireless IP air interface.

Exam iners:

Prof. \*. K. Bhargava. S u p o rv i^ r. Dept, of Electrical &: C o m p u ter Eu^ineerlng

Prof. F. E ^ ^ u ih a ly . Memjier. Dept, of Electrical & C o m p u ter Eu^iiu'erim:

Prof. K. F. Li. Member. Dept, of Electrical & C om puter Engineering

Prof. S. D o s t . ^ i t s i d e M e rn b e ^ D e p tt'T if'^ a 'lia n ic a l En^,ineerinu;

. External Exam iner Prof. S. Rov

Table of Contents

A bstract

ii

Table o f C on ten ts

iv

List o f Figures

ix

List o f Tables

xv

A cknow ledgem ent

xvii

D ed ication

xviii

1 Introd u ction

1

1.1 Wireless Communication.s and Networking T o c h r i o l o s y ... 1

1.2 IP (Internet Protocol)-Based Wireless D a ta Networks ... ô 1.3 Motivation and Scope of the T h e s i s ... 7

1.4 Thesis O u t l i n e ... 9

2 B inary Feedback-Based R etransm ission C ontrol for a M ultichannel

S-A LO H A P r o to co l in Fading Channels

11

2.2 System Model a n d D e s c r i p t i o n ... 14

2.3 Performance Evaluation for Error-Free Channels by Markov Analysis IG 2.3.1 Capacity A n a l y s i s ... 16

2.3.2 T h ro u g h p u t Analysis ... 23

2.4 Performance Evaluation in Fading C h a n n e l s ... 28

2.4.1 Fading C hannel M o d e l ... 28

2.4.2 Performance R e s u l t s ... 31

Integrated R ate and Error Control in W C D M A -B ased Cellular W ire­

less IP N etw orks

37

3.1 Introduction ... 37

3.2 D ynamic R ate and E rror Control in W CDM A S y s t e m s ... 39

3.2.1 D ynamic R ate .Allocation P r o b l e m ... 39

3.2.2 Realization of \'ariab le Rate T r a n s m i s s i o n ... 41

3.2.3 Error Control .Alternatives ... 41

3.3 Channel M o d e l s ... 43

3.3.1 Single P a th Channel with .\o F a d i n g ... 43

3.3.2 M ultip ath Rayleigh Fading Channel with E([ual P a th G ain . 44 3.3.3 M ultip ath Rayleigh Fading Channel with Cnequal P ath Gain 43 3.4 .A Two-Step R ate Selection P r o c e d u r e ... 43

3.4.1 Description of the P r o c e d u r e ... 43

3.4.2 Performance R e s u l t s ... 47

3.4.3 Performance Comparison with the O ptim al Rate Selection . . 30

3.4.4 Performance Comparison with L'ncontrolled Random Rare Se­ lection 32

3.3 .A Markov A nalytical Model for RLC/.M.AC-Layer Performance Eval­ uation ... 33

3.6 -A BS-.Assisted Distributed RLC/.M AC Protocol with Integrated Rate and Error C o n t r o l ... 37

3.7 C h a p te r S u m m a r y ... 60

H igher Layer P r o to co l Perform ance in C D M A S-A LO H A N etw orks

w ith Packet C om bining in R ayleigh Fading C hannels

63

4.2 Background and M o t i v a t i o n ... 63

4.3 System Model and .Assumptions ... 66

4.3.1 Traffic Source M o d e l ... 66

4.3.2 C hannel and Receiver M o d e l ... 67

4.3.4 T P (Transport Protocol) Models for Two Levels of E rror Re­

covery ... 71

4.3.4.1 T P with EB (Exponential Backoff)-Based R T O Control 72 4.3.4.2 T P with EM’M.A. ( Exponential Weighted Moving .\veraging)-Based RTO C ontrol ... 73

4.3.5 L’nreliablo T P T h ro u g h p u t P e r f o r m a n c e ... 73

4.3.6 Performance M e t r i c s ... 74

4.3.7 Simulation .Architecture and P a r a m e t e r s ... 75

4.4 Simulation Results and D is c u s sio n s ... 76

4.4.1 Effect of Diversity Combining on ... 78

4.4.2 Effect of Diversity Com bining on T n ... 79

4.4.3 Effect of R T O Control M e c h a n i s m ... 80

4.4.4 Sensitivity of T n to c ... 82

4.4.5 Sensitivity of T n to \'a r ia tio n in M S S ... 84

4.4.6 Effect of EEC Coding on T n ... 85

4.4.7 Effects of Channel Dispersiveness and the Number of M ultip ath s 86 4.4.8 L'nreliable T P P e r f o r m a n c e ... 86

4.5 C h a p te r S u m m a r y ... 87

Fair B andw idth A llocation T hrough T D M A -B ased C entralized M A C

P ro to co ls in W ireless IP N etw orks

90

5.2 Background and M o t i v a t i o n ... 91

5.3 System Model and D e s c r i p t i o n ... 93

5.4 Traffic Conditioning and Scheduling for Fair Bandw idth .Allocation . 95 5.4.1 .Admission Control .A lg o rith m ... 95

5.4.2 Traffic Conditioning at Mobile Terminals ... 96

5.4.3 The Bandw idth .Allocation . A l g o r i t h m ... 97

5.5 Simulation Model and .Assumptions ... 100

5.5.1 Channel M o d e l ... 100

5.5.2 Flow M o d e l s ... 101

5.5.2.1 Poisson Flow M o d e l ... 101

5.5.2.3 LRD Flow Model ... 102

5.5.3 Measure of Energ}' E f f i c i e n c y ... 102

5.5.4 Simulation E n v i r o n m e n t ... 103

5.6 Simulation R e s u l t s ... 105

5.6.1 Homogeneous Poisson Flows ... 105

5.6.2 Homogeneous GB.A.R F l o w s ... 108

5.6.3 Homogeneous LRD F l o w s ... 109

5.6.4 Heterogeneous F l o w s ... I l l 5.7 C h a p te r S u m m a r y ... 114

C hannel U tilization and T C P T hroughput Fairness in W ireless IP

N etw orks

117

6.2 Background and M o t i v a t i o n ... 119

6.3 Related W o r k s ... 121

6.4 System Model and .Assumptions ... 123

6.4.1 T C P .M odel... 123 6.4.2 Link-Layer Model ... 124 6.4.3 Channel M o d e l ... 125 6.4.4 Channel Probing M o d e l ... 126 6.4.5 Scheduling S c h e m e s ... 126 6.4.6 Performance M e t r i c s ... 128

6.4.7 Simulation P aram eters and Simulation E n v i r o n m e n t ... 128

6.5 Simulation Results and D is c u s s io n s ... 130

6.5.1 Effects of X'ariation in i r ... 130

6.5.2 Effects of X'ariation in T D ... 132

6.5.3 Effects of X'ariation in . V c ... 133

6.5.4 Effects of \ ariation in f , i T ... 134

6.5.5 Effects of \ ariation in r i m a x ... 135

6.5.6 Effects of \ ariation in A’ ... 136

6.6 Integration of T C P Level Packet Scheduling into the G P R S Transm is­ sion Protocol S t a c k ... 138

7 Link-Level Traffic Scheduling for P roviding P red ictive QoS in W ire­

less M u ltim ed ia N etw orks

144

7.1 I n t r o d u c t i o n ... 144

7.2 Link-Layer Traffic Scheduling in Wireless M ultimedia Networks . . . 147

7.3 Burst-Level Cell Scheduling S c h e m e s ... 131

7.3.1 TDM.A./TDD Frame S t r u c t u r e ... 151

7.3.2 Description of the P r o t o c o l s ... 152

7.3.3 Scheduling . \ l g o r i t h m s ... 153

7.3.3.1 Param eters and N o t a t i o n ... 153

7.3.3.2 . \ l g o r i t h m s ... 154

7.3.3.3 C om putational C o m p l e x i t y ... 159

7.4 Sim ulation of the Scheduling S c h e m e s ... 159

7.4.1 Traffic Models and Performance M e a s u r e s ... 159

7.4.2 Simulation E n v i r o n m e n t ... 161

7.4.3 Simulation Results and D is c u s sio n s ... 162

7. 5 Effect of Correlated Channel Errors and C hannel-S tate .A.ware Schedul­ ing 168

7.6 C h a p te r S u m m a r y ... 175

8

C onclusions

178

8.2 Future Work and Work in P r o g r e s s ... 181

B ibliography

184

A p p en d ix A E valuation o f

192

A p p en d ix B E valuation o f

in (3.8)

193

A p p en d ix C Im p lem entation o f th e secon d -step o f th e rate search pro­

cedure

194

List of Figures

Figure 1.1 T he transmission protocol stack in a wireless IP network. . . . 6

Figure 2.1 Markov chain showing the system states under full load condition. 17 Figure 2.2 MBFS-1: \ ariation of normalized system capacity with and

X (for .V = 10Ü. .U = 10. r = 9 ) ... 19

Figure 2.3 .\[BFS-2: \ ariation of normalized system capacity with <i and

X (for .V = 100. ,\[ = 10. r = 9 )... 20

Figure 2.4 .\[BFS- k \'ariation of normalized system capacity with q and

X (for .V = 100. M = 10. r = 9 )... 21 Figure 2.Ô M B F S - l . M B F S - 2 and MBFS-d: Effect of r on normalized sys­

tem capacity (for .V = 400. .1/ = 10) 22

Figure 2.6 MBFS- 1. .MBFS-2 and .MBFS-2: \ ariation of normalized sys­

tem th ro u g h p u t with .V (for .M = 5. r = 9. = 0.1)... 27 Figure 2.7 . MBFS-l. .MBFS-2 and .MBFS-2: \ a r i a t i o n of C with .V (for .M

= 10. L = 1. r = 9. Pe = 0.0. 0.1. 0.2. 0.4)... 28 Figure 2.8 .MBFS-l. .MBFS-2 and .MBFS-2: \ aria tio n of C with .V under

correlated fading (for .M = 10. L = 1. r = 9. P/r = 0.1. 0.2 and f , tT = 0.0084. 0.0252. 1.5003)... 32 Figure 2.9 . MBFS-l. .MBFS-2 and .MBFS-2: \ aria tio n of C with .V under

correlated fading for different values of p, (for .M = 10. I = 2. r = 9. Pe = 0.1 a n d / r f P = 1.5003)... 33 Figure 2.10 .MBFS-l. .MBFS-2 and .MBFS-2: \ aria tio n of C with .V in

correlated fading channel under local a d a p ta tio n (for M = 10. L = 1.

r = 9. Pe = 0.0. 0.1. 0.2. f ^ T = 0.0084. 0.0252. 0.5004. 1.5003). . . . 34

Figure 3.3 F irst-step ra te selection m„ vs. num ber of active users n (for G 5m -based error control)... 50 Figure 3.4 Effects of r/ and t on variations in .i' with n (for channel model-C). 51 Figure 3.5 Effects of G and ^ on variations in 3 ' with n (for channel

m odel-C )... 52 Figure 3.6 Com parison between optim al and sub-optim al rare selection

perform ance (with 5/î-based error control)... 53 Figure 3.7 C om parison between optim al and sub-optim al rate selection

performance (with CBm-based error control)... 54 Figure 3.8 \ aria tion in 3 with G with two-step rate selection procedure. . 57 Figure 3.9 T h ro u g h p u t performance of the proposed M.AC scheme (with

5/?-based error control)... 60 Figure 3.10 T h ro u g h p u t performance of the proposed MAC scheme (with

GBm-based error control)... 61

Figure 4.1 T ra n s m itte r and receiver m odel... 68 Figure 4.2 D a ta flow between RLC/M.4.C layer and physical layer... 68 Figure 4.3 T n for the different access control schemes (for f = 2: .\’C =

no combining. \VC = with com bining)... 78 Figure 4.4 T n for the different link-layer access control schemes (for t =

2. M S S = 24 and E B -b ased R T O control: .\'C = no combining. WC = with com bining)... 81 Figure 4.5 T n for the different link-layer access control schemes (for t =

2. M S S = 24 and E B -b ased R T O control: G.A. = global a d a p a ta tio n . L .\ = local a d a p t a ti o n ) ... 82 Figure 4.6 T n \vith EB-based and EW M A -based RTO control (for f = 2.

c = 0.2. .1/EE = 24)... 83 Figure 4.7 Sensitivity of T n co c for EW M A -based RTO control (for f =

5. A/55 = 2 4)... S3 Figure 4.8 Effect of variation of M S S on T n (for t = 5. c = 0.1)... 84 Figure 4.9 Effect of E E C coding on T n (for c = 0.2. M S S = 2 4 ) ... 85

Figure 4.10 Effect of L on T a (for t = 2. M S S = 24)... 87 Figure 4.11 \ ariation of T a a n d Pdrop with .V for the different access control

schemes (for f = 5. .1/55 = 24)... 88

Figure 5.1 TDM.A. frame stru c tu re for (a) burst-level scheduling, a n d for (b) packet-level scheduling... 93 Figure 5.2 Long-term th ro u g h p u t for Poisson flows under hnrst-lerelschedul­

ing (with /?, = 32. 64. 100 Khp.s for / = 0 - 14. 15 — 24 and 25 — 29.

respectively, and Pg = 0.1. v = 5 k n i / h r . f,{T = 0.00833)... IOC Figure 5.3 Long-term th ro u g h p u t for Poisson flows under pucAe/-/e/ef schedul­

ing (with /?, = 32. 64. 100 Kb ps for ; = 0 - 14. 15 — 24 and 25 — 29.

respectively, and = 0 . 1 . v = 5 k m / h r . f j T — 0.00833)... 107 Figure 5.4 Comparison am ong the throughput, delay and loss performances

under burst-level and packet-level scheduling (for Poisson flows with

R, = 32. 64. 100 Kbps for ( = 0 - 14. 15 - 24 and 25 — 29. respectively.

and Pe = 0.1. v = 5 k m / h r . f , i T = 0.00833)... 108 Figure 5.5 The average transmi tter usage time and the a\e ra g e rereirer

usage time under packet-level and burst-level scheduling (for Poisson

flows with R, — 32. 64. 100 Kb ps for / = 0 — 14. 15 — 24 and 25 - 29. respectively, and Pe = 0.1. v = 5 k r n / h r . f, i T = 0.00833)... 109 Figure 5.6 The effects of channel-error correlation on the average tr a n s m it­

ter and receiver usage times under packet-level and burst-level schedul­ ing (for Poisson flows with R, = 32. 64. 100 Kbps for / = 0 - 14. 15 — 24 and 25 - 29. respectively, and -V, = 10. / - 0.25 -s. Z?„,nx = I >)• ■ . . HO Figure 5.7 The effects of channel-error correlation on the thro u g h p u t, aver­

age tra n s m itte r and receiver usage times under packet-level and burst- level scheduling (for G B .\ R flows with R, = 200. 250. 300 K b p s for /■ = 0 — 9. 10 — 19 and 20 — 29. respectively, and . \ \ = 10. / = 0.1 .s.

Dmax = 1 s)... I l l

Figure 5.8 Long-term th ro u g h p u t for LRD flows under packet-level and burst-level scheduling (for /?, = 20. 40. 100 Kbps for ( = 0 — 9. 10 — 19 and 20 — 29. respectively, and Pe = 0.01. v = 5 k m / h r . f T T = 0.00833). 112

Figure 5.9 \ ariation in long-term th ro u g h p u t and average tra n s m itte r and receiver usage times under packet-level and burst-level scheduling in a heterogeneous traffic scenario (for /?o- 9 = 64 Kbps. Poisson: Ri o-vj =

100 Kbps. G BAR: Roo-’ i = 20 Kbp s. LRD: Pe = 0.1. v = 5 k r n / h r . /rfF = 0.00833)... 113 Figure 5.10 Effect of variation in a on the average tra n s m itte r and receiver

usage times under packet-level and burst-level scheduling in a hetero­

geneous traffic scenario (for = 64 Kbps. Poisson: = 100

Kbps. G B .\R : R>q-2i ~ 20 Kbps. LRD: I — 0.25 s. 0.25 > and 0.1 .s

for Poisson. GB.A.R and LRD flows, respectively)... 114

Figure 6.1 Simulation topology (F T = fixed term inal. MT = mobile ter­ minal. BS = base s ta tio n )... 129 Figure 6.2 \ ariation in channel utilization for different values of IF under

different scheduling schemes (for K = 3. u,„„x = 20. .V,. = 4. Pe = 0.1.

V = Ô k i n / h r ) ... 132

Figure 6.3 \ ariation in throughput fairness for different values of IF under different scheduling schemes (for K = 3. n,nax = 20. .V,. = 4. T D = 10

m s . Pe = 0.1)... 133

Figure 6.4 \ ariation in channel utilization for different values of T D under different scheduling schemes (for K = 3. nmax = 20. .\\. = 4. Pe = 0.1.

V - 5 k i n / h r ) ... 134 Figure 6.5 \ ariation in throughput fairness for different values of T D under

different scheduling schemes (for I\ = 3. u,nnx = 20. .V^ = 4. Pe = 0.1.

V = 5 k m / h r ) ... 135 Figure 6.6 \ ariation in channel utilization for different values of under

different scheduling schemes (for K = 3. n^ax = 20. IF = 10. Pe = 0.1. V = 5 k m / h r ) ... 137 Figure 6.7 \ a r i a t i o n in throughput fairness for different values of .V^ (for

K = 3. n^ax = 20. IF = 10. Pe = 0.1. v = 5 k m / h r ) ... 138

Figure 6.8 \ ariation in T F P for the different scheduling policies (for K =

Figure 6.9 \ ariation in U and l / F for different values of n,nnr (for !\ = 3.

T D = 10 m s. W = Ô. Xr = A. Pe = O.l)... 142

Figure 6.10 G P R S transmission plane protocol a rc h ite c tu re... 143

Figure 7.1 T D M . \ / T D D frame fo rm a t... 132 Figure 7.2 Traffic trace generated from (a) a voice source with fast speech

activity detector, (h) a G B .\ R source and (c) an LRD model (Pareto distribution with mean = 10. a = 1.1)... 160 Figure 7.3 \ a r i a t i o n of U. C T D r t - V B R and C L P r t - v s R hi a single traffic

( r t - \ 'B R only) environm ent... 163 Figure 7.4 FC F S-F R : \ ariation of U . C T Dc b r/ C T D rt-\ hr and C L P r R R / C L P r t - \ nR

in a multitraffic (GBR and r t- \ 'B R ) environm ent (for Xrt~\ br = 5). . 164

Figure 7.5 F C F S-F R -r: \ ariation of L'. C T D r B R / C T D r t - \ B R ' ^ n d C L P c f ) f { / C L P r t - \ br

in a multitraffic (GBR and r t- \ 'B R ) environm ent (for . \ 'n-\ BR = 3.

f ^ r t - V B R = 0

)...

Figure 7.6 ED F -FR : \ ariation of U. C T Dc r r/ C T D ri-\ b r and C L Pc b r/ C L P r,-.\-br

in a multitraffic (GBR and r t - \ 'B R ) environm ent (for Xr t- vB R = 3.

PcBR = 6 )... 166

Figure 7.7 E D F -F R -f: \ ariation of i ' . C T Dc b r/ C T O rt-vBR and C L Pc b rI C L P ri-\ br

in a multitraffic (GBR and r t - \ ’BR) environm ent (for .\\-i- \ br = 3. Rcflff = 6 ) ... 167 Figure 7.8 M TD R; \ ariation of C. C T Dc b r/ C T Drt - vBR and C LPc b r/ C LPrt- \ br

in a multitraffic (GBR and r t- \ 'B R ) environm ent (for X r t - v s R = 3.

PcBR — 6. Rrt-\ BR = 3)... 168

Figure 7.9 E D F -F R -r and M TDR: \ ariation of C L P c b r - C L P c - v b r " 'h h

Xnrt-VBR In a multitraffic (GBR. r t - \ 'B R and LRD n r t- \ 'B R ) environ­

ment (for X r t - v B R = 3. XcB R ~ 30. for EDF-FR-r: R c - v b r = 4.

R n r t - V B R = 2. for M TDR: R c - V B R = 3. R n r t ~ \ BR = 0 )... 169

Figure 7.10 E D F-FR -r and MTDR: \ ariation of C L P c b r - C L P c - v b r " ‘ith

Xnrt-VBR In a multitraffic (GBR. r t - \ 'B R and non-LRD n r t- \'B R ) en­

vironm ent (for X c B R = 30. Xrt-VBR = 3. Rrt-VBR = 4. RcBR =

Rnrt-VBR = 0 ) ... GO

Figure 7.12 E D F -F R -f. M TD R: Effect of channel-error correlations on U and CL Pr t~ vBR in a single traffic ( r t - \ 'B R only) scenario (for Pe = 0.01) 17.3 Figure 7.13 E D F -F R -f. M TD R: Effect of channel-error correlations on C

and C L P r t - v B R in a multitraffic (C B R a n d r t-\'B R ) scenario (for Pe

= 0.01. . \ ’r t - \ . B R = Ô. R r t - \ B R = -I- P c B R = 0)... 174

Figure 7.14 Performance of E D F -F R -f in an integrated traffic scenario for LRD and non-LRD n r t - \ 'B R (shown as EX P) traffic (for Pe = 0.01. c

List of Tables

Table 2.1 The values of q and x for which C is m axim um u n d er different

X / M (for MBFS -1 with A/ = 20. 1 = 1)... 23

Table 2.2 The values of q and x for which C is m axim um under different

X / M (for M B F S - 2 with .W = 20. 1 = 1)... 23

Table 2.3 The values of q and x for which C is m axim um under different

X / M (for M B F S - U v i t h M = 20. L ^ i ) ... 24

Table 4.1 Simulation p a ra m e te rs... 77 Table 4.2 T a with F EC w ith /w ith o u t diversity packet com bining (for .V

= 60: N'C = no combining. WC = with com bining)... 80

Table -5.1 Selected sim ulation p aram eters ... 104

Table 6.1 Simulation p a ra m e te rs ... 130 Table 6.2 \ ariation in i ’ and l / F for different values of IF u nder different

scheduling schemes (for A" = 3. .V,- = 4. = 20. P f = O.l. r = .j

k r n / h r ) ... 131

Table 6.3 \ ariation in F and 1 / F for different values of F D u n d er different scheduling schemes (for K = 3. u,nax = -0- F e = 0.1. r = .3 k r n / h r ) . 136 Table 6.4 \ ariation in U and l / F for different values of .V,. u n d er different

scheduling schemes (for K = 3. n,nax = -6. IF = 10. F D = 10 ni.s. Pp = 0.05. c = 5 k m / h r ) ... 139 Table 6.5 \ a r i a t i o n in U and l / F for different values of f j T u n d er different

scheduling schemes (for K = 3. n,nai = 20. IF = 10. F D = 10 m.i. Xc = 4)... 141 Table 6.6 \ aria tion in F and l / F for different values of K u n d e r different

scheduling schemes (for TD = 10 m.s. rimai = 20. F e = 0.1. c = 5

Table 7.1 Different service classes for wireless m ultim edia networks. . . . 145 Table 7.2 Degree of Q oS... 147 Table 7.3 Simulation p a ram eters... 162 Table 7.4 Typical performance results for ED F-FR -r in a G B A R r t - \ ’BR

only traffic scenario ( Pe = 0.01)... 175

Table 7.5 Typical performance results for EDF-FR--- in a C B R and G B .\R r t - \ ’BR traffic scenario (/■ = 1 k m / h r . Rr t- \ br = 4. Rc b r = 0). . . . 176

Acknowledgement

I would like to express my profound g ratitu d e and appreciation to Professor \'ija y

K. Bhargava for his continuous sup p o rt, inspiration and understanding. I have been

very fortunate to work in such a superb intellectual environm ent under Professor Bhargava whose supervision and mentoring have always been creating a 'm orality of aspiratio n ' inside me.

I am very grateful to Professor Payez El-Guibaly. Professor Kin F. Li and Professor Sadik Dost for serving on my exam ination com mittee.

Special th a n k s goes to Professor Sum it Roy for agreeing to be the external exam ­ iner in my Ph.D . oral examination.

Last but not the least. I would like to extend my sincere thanks t(j all of my colleagues and friends here in X'ictoria for their friendship a n d cooperation.

Dedication

To my grand m a late Mrs. Zobeda Khaturi and to the members of my family whose blessing, love and affection have been the greatest possession in my life

" .\n d when old words die out on the tongue, new melodies break forth from the heart: and where the old tracks are lost, new country is revealed with its wonders. "

Introduction

1.1

W ireless C om m unications and Networking Tech­

nology

Today, wiroloss com m unications and networking is the (astest growing segment of the telecom munications industry. Traditional wireless com m utdcations and networking systems, which include cordless and cellular phones, paging systems, mobile d a ta networks and mobile satellite systems, have evolved from the enormous research and development efforts in b o th academ ia and industry.

Wireless com m unications and networking systems can be categorized as follows Ti:

• High Power Wide Area Systems (or Cellular S ys t em si th a t have been designed to support mobile users roaming over wide geographic area

• Low Power Local Area Systeins. for example, cordless telephone systems th a t are im plem ented w ith relatively simpler

technology-• Low Speed Wide Area Sys tems th a t have been designed for mobile d a ta ser­ vices with relatively low d a t a rates (e.g.. C D P D (Cellular Digital Packet D ata) systems)

• High Speed Local Area Sys tems th a t have been designed for high speed and local com m unications (e.g.. W LAX (Wireless LAX) systems).

T he first two categories are voice-oriented systems while the rem aining two are data-oriented systems. Developments in voice-oriented systems can be described by- referring to three generations of these systems as follows:

(Xordic Mobile Telephone). .\M P S (.Advanced Mobile Phone Service) and T.\CS (Total .\ccess C om m unications Systems) are the three prim ary analog cellular radio systems standards. C T (N orth .\m erican Cordless Telephone). CTO and C T 1 / C T 1 + are the first generation analog cordless standards.

• Second Generation (2G) Systems refer to the digital cellular and cordless sys­ tem s which employ digital m odulation and advanced call processing capabilities. GSM (Global System for Mobile C o m m u n ic a tio n )-th e pan-E uropean digital cel­ lular radio system. [S-54/IS-136 -the TDM.A. (Tim e Division Multiple .Access)- based North .\m erican digital cellular systems, and IS-95-the C D M A (Code Division Multiple .\ccess)-based digital cellular system and the .Japanese PD C (Personal Digital Cellular Radio) represent the second generation of wireless cellular systems. C T 2 /C T 2 4 -. D E C T (Digital European Cordless Telecom mu­ nications). PHS (Personal Handy Phone System) and the cordless in the ISM' bands are am ong the m ajor second generation digital cordless standards.

• Third Generation (SG) Sys tems are envisioned to support multidimensional (multi-inform ation media, multi-transmission media and multilayered networks) high-speed wireless com m unications [2]. High speed packet d a ta services up to 2 Mbps for wireless Internet access is expected to be the main a t tr ib u t e of :3G systems. Studies and s tan d ard izatio n efforts on 3G systems are being carried out under the names IMT-2000 (International Mobile Telecom munications in the year 2000). L'MTS (Universal Mobile Telecommunication Service) and MBS (Mobile B roadband System). M ajority of candidate RTT (Radio Transmission Technology) proposals for IMT-2000 choose D S /C D M A (Direct Secjuence Code Division Multiple Access) as the leading multiple access technique. Activities on the harm onization of the technical specifications for the different R T T can­ didates are being carried out with an aim to achieve a single converged 3G s tan d ard .

'I t refers to the 902-928 MHz. th e 2400-2483.5 .\IHz and the 5725-5850 .MHz bands w hich have b een allow ed for unlicensed use o f sp read-spectruni (direct-sequence or frequency-hopping) radios up to 1 ir.

Some of the contributing factors are ([3]-[6|): low speed wireless access links, poor link performance, fragm entation of wireless access stan d ard s, protocol a n d software design issues, lack of core packet network infrastructure, lack of synergy- w ith Internet, expensive user devices and b a tte ry technology’.

But with the explosive dem and for Internet access and the development of a huge num ber of innovative Internet-based applications and services, it is expected that there will be widespread adoption of wireless d a t a services in the near future. Similar to the evolution of wireline Internet, wireless Internet services will evolve to high ban d w id th services such as full m ultim edia email and web browsing, s ta r tin g with the low b a n d w id th text-based services such as wireless e-mail and web inform ation retrieval. Evolving wireless packet networking technology would make it possible to provide seamless wide-area Internet service to mobile users.

C urrent low -bandw idth wireless d a t a systems include

• A R D I S (Advanced Radio Data Information Service): the .\RD1S mobile d a ta network formed by IBM and M otorola is providing coverage in over 400 m etropoli­ ta n areas in the USA. It uses 25 KHz channels in F D M .\ mode in the 800 MHz band, and provides 4.8 A'èp.s transm ission rates in most areas [Tj.

• R A M Mobile Data: the R.\.M Mobile D a ta Network, based on the MOBITE.X protocol developed by Ericsson and Swedish Telecomm, uses 12.5 A'//c channels in the 896-901 MHz and 935-940 MHz bands and provides 8 Kbps transm ission rates. It uses S-.A.LOH.A. as the distrib u ted m edium access control mechanism.

• C D P D (Cellular Digital Packet Data): the C D P D system provides packet d a t a service as a non-interfering overlay to the .A.MPS analog cellular system using paired 30 KH z channels. It provides a transm ission rate of 19.2 Kbps w ith the m axim um user thro u g h p u t a b o u t half this rate. C D P D uses the DSM.A. (Digital Sense M ultiple Access) technique. Higher layers of the C D P D protocol stack are based on s ta n d a rd ISO and Internet protocols.

.Apart from the above systems. 2G cellular systems also s u p p o rt circuit- and packet-switched d ata. For example. GSM s u p p o rts circuit-switched d a t a from 2.4-14.4 Kbps. G P R S (General Packet Radio Service), which is a sep arate packet d a ta

Kbps. IS-95. the COM.A.-based X orth .American digital cellular standard, supports

circuit-switched d a ta up to 14.4 Kbps. IS-95B can su pport a peak d a t a rate of 76.8

Kbps. .Although the cordless systems, such as PHS. D E C T and P.ACS. were designed

prim arily for voice service, with larger bandw idth available in these systems, d a t a services in the 32-500 Kbps range could be supported.

3 0 systems aim to provide d a t a services with rates of 144 Kbps in wide-area vehicular environm ents and up to 2 .Mbps in indoor and picocellular environments. E D G E (Enhanced D a ta for GSM Evolution), which is an enhancement to GSM. is expected to increase d a ta rates to over 384 Kbps. Peak d a ta rate achievable in systems based on cdrna2000. the 3G Xorth .American CDM.A stan d ard , would be 307.2 ATp.s in the 1.25 .MHz bandw idth or 614.4 Kbps in the 5 .MHz bandw idth.

The WL.AXs. because of their limited transmission range and wider bandw idths (com pared to cellular systems), can provide d a t a rates of 1-16 .\[bps. Three dis­ tinct technologies are currently in use for WL.A.X systems: IR (Infrared WL.AXs). microwave WL.AXs operating at 18-19 GHz And spread-spectrum WL.AXs o perating in the ISM bands. IR WL.AXs are of two types DFIR (Diffused IR) WL.A.X and DBIR (Directed-bearn IR) WL.A.X [1], IR systems are relatively simple, are flexible to install, operate at very low power levels and are less likely to interfere with each other. IR signals do not penetrate walls. D a ta rates of currently available D FIR and D BIR WL.AXs are 1 .Mbps and 10 .Mbps, respectively. Microwave WL.A.X products from M otorola (e.g.. .Altair and .Altair Plus systems) provide a d a ta rate as high as 15 .Mbps. S pread-spectrum WL.AXs provide larger coverage than IR and microwave WL.AXs a n d its achievable d a ta rates are 2-6 .Mbps. Lucent s WaveL.AX supports d a t a rate up to 2 .Mbps.

Xew WL.AX standards, such as HIPERL.A.X (High-Performance L.A.X). aim to deliver d a t a rates of up to 20 .Mbps.

Networks

Packet-switched wireless d a t a networks built upon IP (Internet Protocol)-based in­ frastructure can leverage the rapidly evolving Internet technology in the wireless dom ain so th a t user expectations can be met in a cost-effective m anner. IP-centric solutions for high-speed wide area wireless networking are expected to incur lower network infrastructure cost, provide radio technology independent core network in­ frastructure. provide flexible service architecture based on client-server paradigm, simple service creation and application development environment and ultim ately will pave the way to build a global mobile Internet.

In a wireless IP networking scenario, each PCS (Personal C om m unications Ser­ vices) network will consist of one or more IP-based d a t a sub-networks. Each mobile d a t a user th a t has a unique IP address may home to a sui)-network within the PCS network. D ata sub-networks within a PCS network will interwork with the Inter­ net through some IW F (Interworking Function) or gateway (e.g.. GGSN (Gateway G P R S Support Xode) in G P R S networks). D ata traffic am ong users in different PCS networks are tra n s p o rte d using T C P / I P protocols and. in an 'all-IP ' scenario, d a ta packets within the PCS networks will be also routed by the IP-aw are' BS (Base S ta­ tion )/B S C (Base S tation Switching C enter)/M S C (.Mobile Switching Center), based on the IP address of the destination .MT (Mobile Terminal). Mobility m anagem ent can be performed using the MIP (Mobile IP) protocol. MIP hides the movement of the mobile host to u p p e r layer protocols and applications through the use of home and foreign agents th a t handle the routing of packets to the mobile host. In a wire­ less IP network, core network functionalities need to be enhanced to handle mobile m ultim edia networking functionalities (e.g.. call control signaling using H.3‘23‘ ).

Some infrastructure design alternatives for wireless IP networks were discussed in [8j. W ith an aim of developing an all-IP' core network architecture to deliver 3G ser\ices, efforts are currently under way at 3 G P P (3G P artn ersh ip Project) [9].

■ R ecom m endation H.3'23 developed by the T elecom m unications sector of the International T elecom m unications Union describes the procedures for p oint-to-point and p oin t-to-m u ltip oin t audio and \ddeo conferencing over p acket-sw itched networks.

TE LN ET , FTP, SM TP, DNS,

H TTP

TC P. UDP

IP

W

C

D '

M

A

C

'

E

D

!

G

_ M . _ ^ - P

A- ;

_R

2000 :

S

6

A pplication Layer

T ransport Layer

N etw ork Layer

RLC/M A C Layer

Physical Layer

F i g u r e 1.1. The t r a n . s T n i . ' i s i o n p r o t o c o l s t a c k i n a i c i r e l e s s IP n e t w o r k .

Evolving wireless IP networks will have several air-interface technologies (e.g.. \ VC DM.A (W ideband Code Division Multiple .Vccess). TDM.V-based E D G E ). T h e all-IP" core network will provide radio independent networking facility. Robust RLC (Radio Link C o n tr o l)/M .\C (M edium .\ccess Control) protocol would provide effi­ cient tra n s p o rt of real-tim e and non-real-time d a t a traffic both in the uplink and the downlink.

The IP layer in the wireless IP protocol stack deals with the routing of IP packets. Extensions of M IP or some altern ate solution may be required in this layer to handle vehicular mobility of d a t a users in PCS networks. Functions such as T P C (Transm is­ sion Power Control) are also to be performed in this layer for CDM.V-based physical layer.

In the wireless IP transm ission protocol stack, the transport-layer protocol func­ tions include end-to-end transm ission control and error recovery (in the case of two- level error recovery), some convergence functions and some optim ization functions

1.3

M otivation and Scope of the Thesis

T h e m otivation for this research flows from one main fact: the potential of a large market th a t stems from the increasing dem an d for wireless d a ta services through the Internet. T he emergence of private and public micro-cellular wireless networks are envisioned to provide a variety of broad b an d mobile services, including high­ speed Internet access and au dio/video delivery. Supporting multim edia applications along with the legacy wireless applications (e.g.. cellular voice) within an integrated networking framework gives rise to new network infrastructure and protocol design issues th a t are to be resolved to facilitate m igration to IMT-2000. the third generation wireless scenario and beyond, .\chieving IP-centric solutions for wireless m ultim edia networks to provide reliability, scalability and QoS is a grand technical challenge.

This thesis encompasses several protocol design issues in the area of wireless packet networking technology- for m ultim edia wireless communications. The m ajor areas of this research th a t have been pursued are R L C /M A C and transport-layer protocols a n d the interrelations am ong mechanisms in different protocol layers. The aim of this research has been to accelerate the development of wireless mobile multiservice packet networks and to integrate them efficiently with the present Internet.

M ultichannel protocols with advanced m odulation and coding techni(|ues are en­ visioned to provide high th ro u g h p u t wireless d a ta services. Link-layer retransmission control (or access control) is an im p o rta n t issue in designing efficient R L C / M .\C pro­ tocols (for uplink transm ission control) in multichannel systems. Efficient retrans­ mission control schemes th a t can provide high system capacity over a wide range of system load are to be designed for multichannel random access protocols taking the tim e varying channel im pairm ents into consideration. This issue is addressed in this work. A retransmission control policy is proposed and analyzed for a multichannel S-ALOH A (Slotted ALOHA) protocol in a high speed wireless d a ta network. W ith properly chosen protocol param eters, the proposed scheme can provide high system capacity for a wide range of system load.

is non-trivial in the case of a packet-switched cellular CDM.A. network with hetero­ geneous traffic load in different cells. In this thesis, a sub-optim al two-step dynam ic ra te selection procedure is proposed for uplink packet d a t a transm ission in cellular \ VC DM.A networks. .A novel mean-sense' approach for inter-cell interference cal­ culation is employed assuming homogeneous traffic load in the different cells. Two different error control alternatives for this variable rate packet transmission environ­ ment are presented and their performances are analyzed for three different channel models.

U nderstanding the impact of different physical layer mechanisms (e.g.. m acrodi­ versity packet combining) on higher layer protocol performance would be recpiired for wireless IP transmission protocol stack performance optim ization. T he relative per­ formance of different link-layer access control schemes, in term s of RLC/M.AC-layer and transport-layer th roughput under retransmission diversity packet combining at the physical layer and different tim er control mechanisms at the tran sp o rt layer, is investigated for a CDM.A S-.ALOH.A system. .All of these enable us to b e tte r under­ s ta n d the issues involved in the design of higher layer protocols and to identify the suitable protocol mechanisms for CDM.A-based wireless IP networks.

Fairness would be an im po rtan t issue in the evolving wireless IP networks since wireless subscribers with similar levels of subscription would expect similar service from their W ISPs (Wireless Internet Service Providers). For W ISPs. this may not be simple due to tim e and location dependent channel errors in a wireless environ­ ment. This issue is addressed in this work. .A TDM.A-based centralized bandw idth m anagem ent mechanism is proposed as a RLC/M.AC-layer solution for providing fair b an d w id th allocation am ong com peting uplink traffic flows in a wireless IP network. It consists of an admission control policy, a traffic conditioning policy and a bandw idth allocation policy. T he proposed scheme can be used in an adaptive QoS framework for dynam ically adjusting the QoS of flows in order to accom m odate wireless channel errors and user mobility.

Improved fairness am ong mobile subscribers in the case of downlink transm is­ sion (e.g.. mobile stations downloading web documents) can be provided through

provide improved channel utilization while m aintaining th ro u g h p u t fairness in the case of multiple T C P flows traversing the downlink. This issue is addressed through evaluation of the performances of different transport-layer packet scheduling schemes in a correlated fading environment with the revelation th a t a fair packet scheduling scheme, along with the dynam ic a d ap tatio n of T C P param eters (e.g.. nuixirnurn tran s­ mission window size) based on network delay and error characteristics, can enhance both channel utilization and throughput fairness in wide area wireless IP networks.

Dynamic bandw idth allocation at the R L C /M .\C level for tra n s p o rtin g multiser­ vice traffic is one of the prim ary research issues for the wireless IP network designers. This issue is addressed here by evaluation of the performances of several central­ ized burst-level packet scheduling schemes for the transm ission of voice, video and d a t a traffic over T D M .\ (Time Division Multiple .\c c e s s )/T D D (T im e Division Du­ plex) channels. These schemes are based on the .soft-reservation' of bandw idth for the different traffic types in each TD.M.A./TDD frame along with their instantaneous band w id th requirements. .Mthough the simulation experim ents are based on ,\TM (.\synchronous Transfer .Mode)-based transmission, the concepts here readily apply to wireless IP networks. Smaller RLC/M.AC-layer frame size (e.g.. .ATM cell) provides fine-grain multiplexing, which leads to improved queueing delays and delay jitte r; this is desirable for service integration in the wireless IP air-interface.

T he literature review pertinent to each of the problems addressed in this disser­ ta tio n will be presented in the relevant chapters separately.

1.4

Thesis Outline

Subsequent chapters of this dissertation are organized as follows:

• In C h a p te r 2. a retransmission control scheme for a multichannel S-ALOHA protocol in a micro-cellular wireless environment is proposed and analyzed. T h e d elay-throughput performance is evaluated using exact M a r k w analyses for error-free channel and i.i.d. (independent and identically distributed) Rayleigh fading channel. For correlated Rayleigh fading channels, co m p u te r simulations

are used to evaluate performance.

• In C h a p te r 3. a sub-optim al dynam ic link-level rate a d a p ta tio n procedure is proposed for tra n s m ittin g uplink packet d a t a in \'S F (\'ariable S preading Fac­ tor) W CDM A-based cellular wireless IP networks. Two different error control alternatives for this variable rate packet transmission environment are presented and their performances are analyzed for three different channel models.

• In C h a p te r 4. the performance im plications of retransmission diversity packer combining on link-layer and transport-layer protocol performance are investi­ gated for three different heuristic-based link-layer access control schemes and two transport-layer tim er control mechanisms in a CDM.A. S-.ALOH.A network under frequency selective Rayleigh fading.

• In C h ap ter 5. a unified TDM.A-based centralized wireless access scheme is pro­ posed for performing the statistical multiplexing of bursty d a t a sources in a wireless packet d a t a network. This scheme combines dynam ic b a n d w id th allo­ cation with admission control and packet conditioning (at the mobile stations) to provide fair bandw idth distrib u tio n am ong bursty d a t a flows with different profile rates (or subscription levels) in an error-prone environment. T h e perfor­ mance of the scheme is evaluated using com puter simulations for different total subscription levels, for different compositions of flows with different profile rates and for different channel quality with different channel-error correlation p attern .

• In C h a p te r 6 . performance evaluation of a number of T C P (Transmission C on­ trol Protocol) level packet scheduling policies is carried out in te rm s of channel utilization and th ro u g h p u t fairness (among similar com peting T C P connections) in a wide area wireless IP network.

• In C h a p te r 7. a set of centralized burst-level cell scheduling schemes, namely. F C F S -F R (First Come First Served with Frame Reservation). FCFR-FR-f-. E D F -F R (Earliest Deadline First with Frame Reservation). E D F - F R -r and M T D R (Multitraffic D ynam ic Reservation), are investigated for transm ission of multiservice traffic over T D M .A /T D D channels.

• C h a p te r 8 summarizes the contributions of the thesis and discusses a few direc­ tions for future research.

Chapter 2

Binary Feedback-Based

Retransmission Control for a

Multichannel S-ALOHA Protocol

in Fading Channels

2.1

Introduction

Third-generation wide-area wireless d a t a sendees (e.g.. EDGE) are envisioned to pro­ vide a d a t a trahie th roughput of at least several hundred Kbps |4j. M ultiehannel networks with advanced modulation and coding techniques ap pear to be possible solutions for providing wireless d a ta services with such a high th roughput.

.\LOH.A-based single-channel wireless packet networks, such as C D P D . offer a th ro u g h p u t of less than 20 Kbps due to some intrinsic limitation in carrying high d a t a rate traffic. Since the duration of each packet decreases as the d a t a ra te increases, it is necessary to increase the signal power to m aintain a constant Ef, (Energy- per Bit) [10] so th a t the same BER (Bit Error Rate) can be achieved. M ultiple .\L O H A channels in parallel can be used to meet the high d a ta rate requirements. To provide a high d a t a rate. G PR S, which has been specified as part of GSM Phase 2-r. enables a mobile station to use more th a n one timeslot from one or more physical channels Î5].

In addition to providing a high d a t a rate, a multichannel wireless network can function in a graceful degrading mode in case one or more channels fail due to harsh

channel conditions. In addition, in a distributed environm ent, prioritized access to different mobile stations can be provisioned by restricting th e ir access to a differ­ ent num ber of available parallel channels. For example, depending on the service requirement, a set of mobile stations can be prioritized to have access to a larger num ber of channels tha n another set of mobile stations. Network designers and service providers have to evaluate the network performance under different system p aram eters for proper dimensioning and allocation of system resources.

S-.\LOH.-\ has been widely adopted as a media access protocol and as a component in many reservation protocols in wireless networks. Due to the simplicity of the protocol, it is still of great interest to wireless network designers. Unless modified, some of the protocols, such as CSM.A.. C S M .\/C D and their multichannel extensions ( [ l l | - [ I2 ]). will not work in a wireless network due to the 'hidden t e n n m a l problem'^.

Like any oth e r random access protocol, the satisfactory perform ance of a multi­ channel S -.\L O H A protocol depends on its ability to adjust th e protocol param eters, especially when the offered load to the network is near or above the capacity of the channels. Retransm ission probability is the most a p p ro p ria te p aram eter th a t can be varied if the network load changes. T he selection of the retransm ission probability is crucial for the satisfactory performance of the system. T h e use of an appropri­ ate backoff scheme (local a n d /o r global) can result in an improved delay-throughput performance com pared to the performance of non-adaptive system s "UJj. The choice of the retransmission probability as a function of the num ber of busy (i.e.. either successful or suffering collision)/idle channels can lead to stable tnultichannel access control protocols.

A simple retransm ission backoff scheme for single channel S-.\LOH.A networks was proposed in [l-lj and it was extended for single-hop W D M (Wavelength Division Multiplexing)* star-coupler networks [15]. Literature on retransm ission probability selection in a multichannel finite population S -.\L O H A system is very limited. .\ multichannel infinite population S-.\LO H A protocol w ith co n stant probability re­ transm ission control was studied in [16] and an optim izing criteria for stability was

'T h is refers to a situ ation where two m obile term inals, each o f w hich is w ithin the range o f som e intended third e n tity (e.g.. B S ). are out of range o f each other.

-It is an approach for ver>- high speed networking w hich divides th e o p tical spectrum into m any different w avelength s assigned to each channel.

formulated.

In a S-.ALOH.A, based wireless network, frame^ loss may occur due to the simul­ taneous transm ission of more than one frame (unless any one of the simultaneously tra n s m itte d frames can capture' the receiver) and due to the channel fading. There­ fore. in the case of high system load, transmission a tte m p ts from the mobile stations need to be controlled by delaying them appropriately in order to improve wireless channel utilization. Retransmission control policies can also exploit the channel fad­ ing inform ation so th a t there are more frequent frame transm issions when the channel s ta te is good and there are fewer transmission a tte m p ts when there are errors, which are typically bursty in nature. Delaying retransmissions ra th e r tha n persisting on re­ transm issions under correlated channel fading is desirable from a power conservation viewpoint. T h e performance of a multichannel ISM.A (Inhibir-Sense Multiple .Access) protocol (where the mobile stations a tte m p t transm issions in th e uplink channels based on the busy/idle sta tu s of the channels in each slot broadcasted by the BS in the downlink) in the presence of bursty frame losses was studied in ,17; for a simple ■persist-until-snccess' retransmission strategy.

In this chapter, we provide Markov analyses for the ste a d y -s ta te performance evaluation of a m ultichannel finite population S-.ALOH.A protocol with a general­ ized retransm ission backoff scheme under i . i . d . Rayleigh fading. T h e retransmission control scheme is based on the binary feedback inform ation ab o u t the most recent transm ission a t te m p t s in the uplink channels which is broadcasted by the BS in the downlink channel(s). T h e incorporation of a cap tu re model in the analyses enables us to get insight on the effects of large-scale p a th loss (and hence near-far effect) on the system performance. More interestingly, it makes analyses of a multichannel S-.ALOH.A system, w ithout capture, simpler. T he performance results for the single channel systems with constant probability retransm ission control can be found from the general model as special cases. The stead y -sta te perform ance of the proposed retransm ission control scheme in correlated fading channels is evaluated with the aid of c o m p u ter simulation. The performance results for a more power conservative ver­ sion of the retransm ission control policy based on local a d a p ta tio n are also derived

refers to a unit d ata block at the RLC/M.A.C level. We prefer to use f r a m e rather than packet since the la tter is more appropriate for referring to higher layer d a ta units.

through simulation. In addition, we discuss some other possible application areas of the backoff scheme and possible extensions of this work.

T h e organization of the rest of the chapter is as follows. T h e system model is described in Section 2.2. In Section 2.3. we present Markov analyses for the proposed retransm ission control scheme. T h e performance results, obtained th ro u g h com puter simulation for the proposed scheme in fading channels, are presented in Section 2.4. In section 2.5. we analyze a variant of the proposed scheme based on local ad a p ta tio n . Finally, in Section 2.6. the main points are summarized.

2.2

System M odel and D escription

.A. multichannel S-.ALOHA system w ith .V mobile stations and M parallel channels (i.e.. carriers) is considered where each mobile station has access to all the channels (i.e.. a mobile s ta tio n can tra n s m it to and receive from any of the channels). The time axis is considered to be divided into ecpial-lengtli slots. T he slots in the different channels operate in parallel. The mobile stations transm it or receive constant length frames and each frame fits into one timeslot. Each mobile s tatio n is assum ed to have a tran sm it buffer of one frame size and the station remains blocked until the frame (if any) in the buffer is successfully tran sm itted through one of the channels in a timeslot.

Due to the capture effect in a cellular radio environment, the BS can decode a frame with normal accuracy even in the presence of collisions and channel fading if the signal power corresponding to the frame is sufficiently stronger th a n th a t of its contenders. Due to spatial a tte n u a tio n , m u ltip a th reflections, diffractions and different spatial distributions in a mobile radio channel, the mobile stations are d y nam ically and random ly divided into different classes of access power. R a th e r th a n explicitly taking all these factors into consideration, we consider only the received power levels of the tra n s m itte d frames a t the receiver.

Here, we assume t h a t frames received at the BS can be classified into L classes (class 1 through class L. which are linearly spaced) based on their power levels where class 1 corresponds to the class w ith the lowest power and class L corresponds to the one with the highest power. T he cap tu re model considered here is as follows: during

a timeslot, the BS can decode a frame correctly if the corresponding power level is g reater th a n th a t of each of the contending frames in th a t slot in the same physical channel. T h a t is. a frame from mobile i. i G {1.2. •••..V } is captured if f , > Tj.

j = 1. 2. • ■ •. -V. j # L where .V is the num ber of frames simultaneously tra n s m itte d

in the same timeslot in the same physical channel in which mobile i is tr a n s m ittin g and f t . k = 1.2 . • • • ..V is the received power for the frame tra n s m itte d by mobile

k. This is a more optim istic capture model th a n one where c ap tu re is assum ed to

take place if cum ulative powers of the contending frames is less th a n the desired user power, namely, f , > = [18).

We consider here three variations of the retransmission control policy based on the binary feedback about the channel status: M B F S - l (Multichannel Binary Feedback- 11. M B F S - 2 (M ultichannel Binary Feedback-2) and M B F S - S (M ultichannel Binary Feedback-3). M B F S - l . .\lBFS-2 'And MBFS -. l nse the .s«crc.s.s/non-.snrcR.s>-. Uile/nati-

idle and colli'^ion/non-collmon feedback, respectively. Non-collision feedback cor­

responds to both success and idle. Each blocked mobile station u pdates the value of the retransm ission probability according to the feedback received in the previous timeslot. . \ mobile s tatio n newly generating frames can synchronize to oth e r mobile stations if we assume th a t the BS broadcasts the current value of the retransm ission probability at the end of each timeslot.

T he round trip propagation delay is assumed to be less th a n the slot d u ra tio n so t h a t a mobile station can learn the sta tu s of a channel (i.e.. w hether the transm ission is successful, a collision has occurred, or the channel has been idle) in a slot before the beginning of the next slot. This is a reasonable assum ption for micro-cellular wireless networks. We also assume th a t the feedback inform ation from the BS is tra n s m itte d error-free in the downlink channel(s). This is also reasonable since the small n um ber of bits conveying the feedback information can be adequately protected through channel coding. T h e performance results obtained with the assumptions of instantaneous and error-free feedback serve as an upper bound for the system performance.

Let p{t) be the retransm ission probability in the fth timeslot. For the three vari­ ations of the retransm ission control scheme. p{t) is u p d ated as follows:

(i) M B F S .t

= \ " " "

' -''J

(ii) M B F S . i : p(t +

max(pm, n- pit) X g ) . i f A/,,,, < [x x M \

(iii) .V/SFS-.J: p(, + 1) =

I

""I " ” ■

■"«"

-[ m/n(p,nax- p { t ) / q ) . if < [j: x A/J

where 0

<

>

The protocol param eters x and q determ ine the rate of a d a p ta tio n in th e value of the retransm ission probability. In a particular system configuration, for a p articular value of X . higher values of q cause slow a d a p ta tio n whereas lower values of q result

in high rate of adaptatio n .

D P T (Deferred First Transmission) mode is considered here. i.e.. after the generation of a new frame, a mobile station im m ediately goes into blocked mode. We assume p^ax to be ecpial to 1.

2.3

Performance Evaluation for Error-Free Chan­

nels by Markov Analysis

2.3.1 C apacity A nalysis

The normalized system capacity. C. is defined as the num ber of successfully tr a n s m it­ ted frames per slot per channel when all the mobile stations have frames to tran sm it during each timeslot. In other words. C is the th ro u g h p u t per channel u n d er full load condition. T he system states under full load condition can be described by the discrete-tim e Markov chain shown in Figure 2.1 where the system s ta te A', is chosen such th a t the value of the retransm ission probability pr is (/' in th a t state.

T h e s ta te transition probabilities are given by (i) For M B F S - l .

1 -p '

■'fr.r-F i g u r e 2 .1 . Markov chain .•ihowing the system states under full load condition.

* A / j I A ; } =

^

I

, ) ■ ( I

whore p,urr(n is given as follows:

Psiirc[i) (2.2)

In (2 .1 ). r*; is the probability that a frame is successfully decoded at the BS when k frames are tra n s m itte d simultaneously in a slot in the same physical channel. This is given by (2.3)

27" h > 0 (2.3)

/= I

where is an indication function defined as follows:

I if L = I and k ■ I

0 otherwise.

(2.4)

(ii) For MBFS- 2.

where Pidun) is given as follows:

( .1/ \

^ I _/' ) ^ ) "(2.5)