Analytical Approach to Optical Burst Switched Networks (eBook)

Preface 6
Contents 8
Acronyms 11
1 Introduction to Optical Burst Switching 15
1.1 Evolution of Optical Networks 15
1.2 Optical Circuit Switching 16
1.3 Optical Packet Switching 17
1.4 Optical Burst Switching 18
1.4.1 Node Architectures 21
1.4.2 Burst Assembly 23
1.4.3 Signaling Schemes 25
1.4.4 Routing and Wavelength Assignment 25
1.4.5 Channel Scheduling 26
1.4.6 Contention Resolution 28
1.4.7 Service Differentiation 30
1.4.8 TCP over OBS 31
1.4.9 Other Issues 32
1.5 Variants of OBS 33
1.5.1 Labeled OBS 33
1.5.2 Wavelength-Routed OBS 34
1.5.3 Dual-Header OBS 35
1.5.4 Time-Slotted OBS 36
1.5.5 Hybrid Architectures 38
1.6 Applications of OBS 40
1.6.1 Grid Computing 40
1.6.2 Access and Metro Networks 43
1.7 Testbed Implementations 44
1.8 Need for Modeling and Analysis of OBS Networks 48
1.9 Organization of the Book 50
References 51
2 Traffic Characteristics in OBS Networks 56
2.1 Introduction 56
2.2 Modeling Traffic at an Ingress Node 57
2.2.1 Models for Specific Input Traffic Characteristics 57
2.2.1.1 Poisson Arrival Process for Input Traffic 59
2.2.1.2 Bernoulli Arrival Process for Input Traffic 62
2.2.1.3 Exact Model for Poisson Traffic 67
2.2.1.4 Self-Similar Input Traffic 71
2.2.2 Models Independent of Input Traffic Characteristics 75
2.2.2.1 Bursty Input Traffic 79
2.3 Queueing Network Model for an Ingress Node 85
2.3.1 Burst Arrival Process 86
2.3.2 Modeling the Ingress Node with Infinite Buffer 89
2.3.2.1 Node Without Wavelength Converters 89
2.3.2.2 Node with Wavelength Converters 95
2.3.2.3 Node Supporting Multiple Classes of Traffic 97
2.3.3 Modeling the Ingress Node with Finite Buffer 99
2.3.3.1 Approximate Solution 103
2.3.4 Modeling Waiting Time at the Ingress Node 106
2.3.4.1 Modeling Burst Assembler Queue 106
2.3.4.2 Modeling Burst Transmission Queue 109
References 111
3 Blocking Probability in OBS Networks 113
3.1 Introduction 113
3.2 Blocking Probability in JET/JIT-Based Reservation 114
3.3 Markov Chain Model for Blocking Probability 115
3.4 Blocking Probability with Variable Offset Time 118
3.5 Blocking Probability with and Without Wavelength Conversion 124
3.5.1 Model for Node Without Wavelength Conversion 125
3.5.1.1 General Model for No-Wavelength Conversion 128
3.5.2 Model for Node with Wavelength Conversion 132
3.6 Blocking Probability with Partial Wavelength Conversion 135
3.7 Blocking Probability with Limited Wavelength Conversion 139
3.7.1 Overflow Fixed Point Method 139
3.7.2 Queueing Model 144
3.8 Blocking Probability: Erlang Fixed Point Method 145
3.9 Blocking Probability for Multi-class OBS Networks 147
3.9.1 Generic Model Without Class Isolation 152
References 156
4 Contention Resolution in OBS Networks 158
4.1 Introduction 158
4.2 Burst Segmentation 158
4.2.1 Blocking Analysis of a Node with Burst Segmentation 159
4.2.1.1 Analysis for Small Burst Sizes 162
4.2.2 Benefit of Burst Segmentation 164
4.2.2.1 Probabilistic Model for Burst Segmentation 165
4.2.3 Prioritized Burst Segmentation 166
4.2.4 Probabilistic Preemptive Burst Segmentation 171
4.2.5 Preemption Probability 175
4.3 Wavelength Preemption 178
4.4 Deflection Routing 183
4.4.1 Modeling Deflection Routing at a Single Port 185
4.4.2 Probabilistic Deflection Routing 189
4.4.2.1 Approximate Model with Wavelength Conversion 189
4.4.2.2 Approximate Model Without Wavelength Conversion 190
4.4.2.3 Accurate Model with Wavelength Conversion 191
4.4.2.4 Accurate Model Without Wavelength Conversion 192
4.4.3 Stabilizing Deflection Routing 193
4.4.3.1 Wavelength Reservation 195
4.4.3.2 Preemptive Priority 196
4.4.3.3 Analysis for Large Networks 197
4.5 Analysis of Fiber Delay Lines 200
4.5.1 Exact Markovian Model 201
4.5.2 Asymptotic Approximation Models 204
4.5.2.1 Balking Model for Short FDLs 205
4.5.2.2 Queueing Model for Long FDLs 206
References 207
5 TCP over OBS Networks 210
5.1 Introduction 210
5.2 A Brief Introduction to TCP 211
5.2.1 Phases in the Congestion Control 212
5.2.1.1 Slow Start 212
5.2.1.2 Congestion Avoidance 212
5.2.1.3 Fast Retransmit and Fast Recovery 213
5.2.2 Popular Variants of TCP 213
5.2.2.1 TCP Reno 214
5.2.2.2 TCP NewReno 214
5.2.2.3 TCP SACK 215
5.3 Factors Affecting TCP Performance over OBS 215
5.3.1 Impact of the Burst Assembly Algorithms 216
5.3.2 Impact of Burst Loss 216
5.4 Analytical Models for TCP over OBS 218
5.4.1 Markov Regenerative Process Theory-Based Models 218
5.4.1.1 Slow Class TCP Sources 220
5.4.1.2 Fast Class TCP Sources 225
5.4.2 Renewal Theory-Based Model for Loss-Based Variants 228
5.4.2.1 TCP SACK 229
5.4.2.2 TCP NewReno 232
5.4.2.3 TCP Reno 232
5.4.3 Markov Chain Theory-Based Models 235
5.4.3.1 Overview of the Model 236
5.4.3.2 Modeling the MBMAP Assembly Algorithm 236
5.4.3.3 Modeling the TCP Congestion Window 238
5.4.3.4 State Transition Probabilities of Markov Chain 239
5.4.3.5 Throughput of a TCP Source 242
5.4.3.6 Modeling the OBS Network 242
5.4.3.7 Fixed Point Analysis of TCP over OBS 243
5.4.4 Modeling High-Speed TCP 244
5.4.5 TCP Vegas 246
5.5 TCP Implementations in OBS 249
5.5.1 Burst TCP 250
5.5.2 Generalized AIMD 253
5.5.2.1 BAIMD 254
5.5.2.2 TCP-ENG 254
5.5.2.3 SAIMD 256
5.6 Improving the Performance of TCP 257
5.6.1 Adaptive Burst Assembly Algorithm 257
5.6.2 Using Burst Acknowledgments and Retransmission 259
5.6.3 Drop Policy 262
References 264
Index 267