Performance Evaluation for Network Services, Systems and Protocols (eBook)

Stênio F. L. Fernandes is a Professor of Computer Science at the Centro de Informática (CIn) of the Universidade Federal de Pernambuco (UFPE), in Recife (Brazil). He is Brazilian Research Chair and Senior Member of the IEEE. He received his B.S. and M.Sc. degrees in Electronic Engineering from the Universidade Federal da Paraiba (Campina Grande, Brazil) and his Ph.D. degree in Computer Science from the Universidade Federal de Pernambuco in collaboration with the University of Ottawa, Canada, where he has also held a Post-Doctoral fellowship position. He has been serving for a number of scientific journals and conferences. His research interests include Performance Evaluation of Network and Communications Systems; Internet Traffic Measurement, Modeling, and Analysis; and Virtualization of Networks and Services.

Foreword 5
Acknowledgments 7
Contents 9
Chapter 1: Principles of Performance Evaluation of Computer Networks 12
1.1 Motivation: Why Do We Need to Assess the Performance of Computer Networks? 12
1.2 Classical and Modern Scenarios: Examples from Research Papers 13
1.2.1 Performance Evaluation in Classical Scenarios 14
1.2.1.1 Application Layer 14
1.2.1.2 Transport Layer 18
1.2.1.3 Network Layer 22
1.2.1.4 Link Layer 25
1.2.2 Performance Evaluation in Modern Scenarios 27
1.2.2.1 Virtualization and Cloud Computing 27
1.2.2.2 Software-Defined Networking 33
1.2.2.3 Network Functions Virtualization 38
1.3 The Pillars of Performance Evaluation of Networking and Communication Systems 41
1.3.1 Experimentation/Prototyping, Simulation/Emulation, and Modeling 41
1.3.1.1 Network Experimentation and Prototyping 42
1.3.1.2 Network Simulation and Emulation 43
1.3.1.3 Analytical Modeling 45
Video Game Characterization 46
TCP Throughput 46
Aggregate Background Traffic 48
Error Models for Communication Channels 49
1.3.2 Supporting Strategies: Measurements 50
References 51
Chapter 2: Methods and Techniques for Measurements in the Internet 55
2.1 Passive vs. Active vs. Hybrid Measurements 57
2.2 Traffic Measurements: Packets, Flow Records, and Aggregated Data 61
2.3 Sampling Techniques for Network Management 64
2.4 Internet Topology: Measurements, Modeling, and Analysis 67
2.4.1 Guidelines for creation, selection, and registration of an Autonomous System (AS) 69
2.4.2 Internet Topology Discovery: Tools, Techniques, and Datasets 71
2.4.2.1 Topology Discovery Tools 71
2.4.2.2 Topology Generator Tools 72
2.4.2.3 Topology Datasets 72
2.5 Challenges for Traffic Measurements and Analyses in Virtual Environments 74
2.5.1 Cloud Computing Environments 74
2.5.2 Virtualization at Network Level 77
2.6 Bandwidth Estimation Methods 79
References 80
Chapter 3: A Primer on Applied Statistics in Computer Networking 84
3.1 Statistics and Computational Statistics 84
3.2 I’m All About That Data 87
3.3 Essential Concepts and Terminology 89
3.4 Descriptive Statistics 91
3.4.1 I Mean It (Or Measures of Centrality) 92
3.4.2 This Is Dull (Or Measures of Dispersion) 92
3.4.3 Is It Paranormally Distributed? (Or Measures of Asymmetry and Tailedness) (Fig. 3.2) 95
3.5 Inferential Statistics 98
3.5.1 Parameter Estimation: Point vs. Interval 99
3.5.1.1 Confidence Intervals 100
Confidence Interval for the Mean 101
3.5.2 Estimators and Estimation Methods 104
3.5.2.1 Essential Properties for Estimators 105
3.5.2.2 Likelihood Function and Its Estimation 105
3.6 The Heavy-Tailed Phenomenon 108
3.6.1 Outlier Detection 109
3.6.2 Heavy-Tailed Distributions and Its Variations (Subclasses) 111
3.6.2.1 What Is This That Stands Before Me? 112
3.6.2.2 A Less Informal Look at the Tail 114
Right Tail Properties of Distributions 114
3.6.3 Evidence of Heavy-Tailedness in Computer Networks 116
References 120
Chapter 4: Internet Traffic Profiling 122
4.1 Traffic Analysis 123
4.1.1 Identification and Classification 123
4.1.2 Techniques, Tools, and Systems for Traffic Profiling 127
4.1.2.1 Port-Based Traffic Profiling 127
4.1.2.2 Shallow and Deep Packet Inspection 128
DPI in Practice 131
Specialized Architectures and Hardware Accelerators 131
Specialized Libraries and Operating System Support 134
4.1.2.3 Flow-Based Traffic Profiling 136
Applied Machine Learning for Internet Traffic Profiling 138
4.2 Industrial Approach for Traffic Profiling: Products and Services 139
4.3 Traffic Models in Practice 141
4.3.1 Workload Generators 142
4.4 Simulation and Emulation 146
4.4.1 Discrete-Event Simulation and Network Simulation Environments 149
4.4.2 Practical Use of Network Simulators and Traffic Profiles 156
References 157
Chapter 5: Designing and Executing Experimental Plans 162
5.1 Designing Performance Evaluation Plans: Fundamentals 162
5.2 Design of Experiments (DoE) 165
5.2.1 The DoE Jargon 168
5.2.2 To Replicate or to Slice? 170
5.3 DOE Options: Choosing a Proper Design 173
5.3.1 Classification of DOE Methods 174
5.3.2 Notation 175
5.4 Experimental Designs 175
5.4.1 2k Factorial Designs (a.k.a. Coarse Grids) 175
5.4.2 2k ? p Fractional Factorial Designs 176
5.4.3 mk Factorial Designs (a.k.a. Finer Grids) 176
5.5 Test, Validation, Analysis, and Interpretation of DOE Results 176
5.6 DOEs: Some Pitfalls and Caveats 177
5.7 DOE in Computer Networking Problems 178
5.7.1 General Guidelines 178
5.7.1.1 Advice for Conducting Experiments in Computer Networking 182
References 183
Erratum to: Performance Evaluation for Network Services, Systems and Protocols 185