Communication Networks and Service Management in the Era of Artificial Intelligence and Machine Learning

1 Cover

2 IEEE Press IEEE Press 445 Hoes Lane Piscataway, NJ 08854 IEEE Press Editorial Board Ekram Hossain, Editor in Chief Jón Atli Benediktsson Xiaoou Li Jeffrey Reed Anjan Bose Lian Yong Diomidis Spinellis David Alan Grier Andreas Molisch Sarah Spurgeon Elya B. Joffe Saeid Nahavandi Ahmet Murat Tekalp

3 Title Page Communication Networks and Service Management in the Era of Artificial Intelligence and Machine Learning Edited by Nur Zincir–Heywood, Marco Mellia, and Yixin Diao

4 Copyright

5 Editor Biographies

6 List of Contributors

7 Preface

8 Acknowledgments

9 Acronyms

10 Part I: Introduction 1 Overview of Network and Service Management 1.1 Network and Service Management at Large 1.2 Data Collection and Monitoring Protocols 1.3 Network Configuration Protocol 1.4 Novel Solutions and Scenarios Bibliography 2 Overview of Artificial Intelligence and Machine Learning 2.1 Overview 2.2 Learning Algorithms 2.3 Learning for Network and Service Management Bibliography Note

11 Part II: Management Models and Frameworks 3 Managing Virtualized Networks and Services with Machine Learning 3.1 Introduction 3.2 Technology Overview 3.3 State‐of‐the‐Art 3.4 Conclusion and Future Direction Bibliography 4 Self‐Managed 5G Networks 1 4.1 Introduction 4.2 Technology Overview 4.3 5G Management State‐of‐the‐Art 4.4 Conclusions and Future Directions Bibliography Notes 5 AI in 5G Networks: Challenges and Use Cases 5.1 Introduction 5.2 Background 5.3 Case Studies 5.4 Conclusions and Future Directions Bibliography Note 6 Machine Learning for Resource Allocation in Mobile Broadband Networks 6.1 Introduction 6.2 ML in Wireless Networks 6.3 ML‐Enabled Resource Allocation 6.4 Conclusion and Future Directions Bibliography Note 7 Reinforcement Learning for Service Function Chain Allocation in Fog Computing 7.1 Introduction 7.2 Technology Overview 7.3 State‐of‐the‐Art 7.4 A RL Approach for SFC Allocation in Fog Computing 7.5 Evaluation Setup 7.6 Results 7.7 Conclusion and Future Direction Bibliography Note

12 Part III: Management Functions and Applications 8 Designing Algorithms for Data‐Driven Network Management and Control: State‐of‐the‐Art and Challenges 1 8.1 Introduction 8.2 Technology Overview 8.3 Data‐Driven Algorithm Design: State‐of‐the Art 8.4 Future Direction 8.5 Summary Acknowledgments Bibliography Note 9 AI‐Driven Performance Management in Data‐Intensive Applications 9.1 Introduction 9.2 Data‐Processing Frameworks 9.3 State‐of‐the‐Art 9.4 Conclusion and Future Direction Bibliography Notes 10 Datacenter Traffic Optimization with Deep Reinforcement Learning 10.1 Introduction 10.2 Technology Overview 10.3 State‐of‐the‐Art: AuTO Design 10.4 Implementation 10.5 Experimental Results 10.6 Conclusion and Future Directions Bibliography Notes 11 The New Abnormal: Network Anomalies in the AI Era 11.1 Introduction 11.2 Definitions and Classic Approaches 11.3 AI and Anomaly Detection 11.4 Technology Overview 11.5 Conclusions and Future Directions Bibliography Notes 12 Automated Orchestration of Security Chains Driven by Process Learning* 12.1 Introduction 12.2 Related Work 12.3 Background 12.4 Orchestration of Security Chains 12.5 Learning Network Interactions 12.6 Synthesizing Security Chains 12.7 Verifying Correctness of Chains 12.8 Optimizing Security Chains 12.9 Performance Evaluation 12.10 Conclusions Bibliography Notes 13 Architectures for Blockchain‐IoT Integration 1 13.1 Introduction 13.2 Blockchain‐IoT Integration (BIoT) 13.3 BIoT Architectures 13.4 Summary and Considerations Bibliography Note

13 Index

14 End User License Agreement

1 Chapter 3 Table 3.1 Summary of the state‐of‐the‐art for virtual network embedding. Table 3.2 Summary of the state‐of‐the‐art for ML‐based placement in NFV. Table 3.3 Summary of the state‐of‐the‐art for ML‐based scaling in NFV. Table 3.4 Summary of the state‐of‐the‐art for ML‐based admission control app... Table 3.5 Summary of the state‐of‐the‐art for ML‐based resource allocation a...

2 Chapter 5Table 5.1 Categorization of covered use cases.Table 5.2 Impact of including different monitoring types on QoE estimation a...

3 Chapter 6Table 6.1 Summary of ML techniques.Table 6.2 Machine learning‐based power control.Table 6.3 blackMachine‐learning based scheduling.Table 6.4 Machine learning‐based user association.Table 6.5 Machine learning‐based spectrum allocation.

4 Chapter 7Table 7.1 Variables used for the minimization of the overall system cost.Table 7.2 A sample fraction of the observation space of the gym‐fog environm...Table 7.3 A sample fraction of the action space of the gym‐fog environment.Table 7.4 The reduced observation space of the gym‐fog environment.Table 7.5 The hardware configuration of each node.Table 7.6 The gym‐fog environment configuration.Table 7.7 The MILP model execution time.

5 Chapter 8Table 8.1 Node and graph features used for problem representation and learni...Table 8.2 provides overview of research work, general optimization problems ...Table 8.3 overviewing networking research work, general optimization problem...

6 Chapter 9Table 9.1 Summary of the key characteristics of data‐processing platform.Table 9.2 Summary of the AI techniques considered for performance management...

7 Chapter 11Table 11.1 Summary of reviewed papers.Table 11.2 Summary of the reviewed tools.

8 Chapter 12Table 12.1 The set of Android applications considered for evaluation.Table 12.2 Number of rules for combined chains.Table 12.3 Accuracy of chains generated for protecting applications.

9 Chapter 13Table 13.1 Performance of selected cryptographic functions on different IoT ...

1 Chapter 1 Figure 1.1 Network and service management at large. Figure 1.2 Example of monitoring architecture. Figure 1.3 The SDN architecture. Figure 1.4 Network functions virtualization architecture.

2 Chapter 2 Figure 2.1 Supervised learning model. Figure 2.2 Unsupervised learning model. Figure 2.3 Reinforcement learning model.

3 Chapter 3 Figure 3.1 Technologies for virtualizing network functions with examples. Figure 3.2 Link virtualization technologies with examples. Figure 3.3 Examples of network slices.

4 Chapter 4Figure 4.1 Slicing and prioritization of 5G network traffic. An eHealth netw...Figure 4.2 O‐RAN high‐level architecture.Figure 4.3 vrAIn's control loop.Figure 4.4 5G‐PPP network architecture.

5 Chapter 5Figure 5.1 Overview of AI/ML use cases in a service‐aware 5G integrated netw...Figure 5.2 Possible framework for ML integration in 5G.Figure 5.3 Overview of the prediction process.Figure 5.4 Mean per class accuracy on the testing set when using different a...Figure 5.5 Components of the QoE management framework.

6 Chapter 6Figure 6.1 Brief overview of ML techniques.

7 Chapter 7Figure 7.1 High‐level view of a fog computing environment [6].Figure 7.2 An example of a service function chaining deployment [14].Figure 7.3 The representation schema of most RL scenarios.Figure 7.4 The fog–cloud infrastructure for the gym‐fog environment evaluati...Figure 7.5 The OpenAi gym environment structure.Figure 7.6 The accumulated reward and the cost difference for the static use...Figure 7.7 The percentage of accepted requests for the static scenario.Figure 7.8 The execution time of each episode run.Figure 7.9 The accumulated reward and the cost difference for the dynamic ca...Figure 7.10 The percentage of accepted requests for the dynamic scenario.

8 Chapter 8Figure 8.1 Comparison between (a) traditional and (b) ML/AI‐based approaches...Figure 8.2 Problem instances are first represented as graph. The graph is th...