Bringing Machine Learning to Software-Defined Networks

1 Machine Learning for Software-Defined Networking

1.1 Introduction of Software-Defined Networking

1.1.1 Software-Defined Wide Area Network

1.1.2 Software-Defined Data Center Networks

1.2 Introduction of Machine Learning Techniques

1.2.2 Multi-Agent Reinforcement Learning

1.2.3 Graph Neural Network

2 Deep Reinforcement Learning-based Traffic Engineering in SD-WANs

2.1 Introduction of Traffic Engineering

2.2.1 Problems of Existing Solutions

2.2.2 Opportunity

2.3 Overview of ScaleDRL

2.4 Design Details of ScaleDRL

2.4.1 Pinning Control in the Offline Phase

2.4.1.1 Pinning Control

2.4.1.2 Link Selection Algorithm

2.4.2 DRL Implementation of the Online Phase

2.4.2.2 Customization of Neural Networks and Interfaces

2.5 Performance Evaluation

2.5.1 Simulation Setup

2.5.2 Comparison Scheme

2.5.3 Simulation Results

2.6 Conclusion

3.1 Introduction of Controller Load Balancing

3.2 Motivation

3.2.1 Problems of Existing Solutions

3.2.2 Opportunity

3.3 Controller Load Balancing Problem Formulation

2.3.1 Control Plane Resource Utilization Modeling

2.3.2 Control Plane Load Balancing Problem Formulation

2.3.3 Problem Complexity Analysis

3.5 Design Details of MARVEL

3.5.1 Training Phase

3.5.2 Working Phase

3.5.3 MARVEL Model Implementation

3.6 Performance Evaluation

3.6.1 Simulation Setup

3.6.3 Simulation Results

3.7 Conclusion

4 Deep Reinforcement Learning-based Flow Scheduling for Power Efficiency in Data Center Networks

4.1 Introduction of Data Center Networks

4.1.2 Traffic Dynamic Analysis

4.2 Motivation

4.2.1 Problems of Existing Solutions

4.2.2 Opportunity

4.3 Problem formulation

4.3.1 Design Consi