Skip to main content

2022 | Book

Bringing Machine Learning to Software-Defined Networks


About this book

Emerging machine learning techniques bring new opportunities to flexible network control and management. This book focuses on using state-of-the-art machine learning-based approaches to improve the performance of Software-Defined Networking (SDN). It will apply several innovative machine learning methods (e.g., Deep Reinforcement Learning, Multi-Agent Reinforcement Learning, and Graph Neural Network) to traffic engineering and controller load balancing in software-defined wide area networks, as well as flow scheduling, coflow scheduling, and flow migration for network function virtualization in software-defined data center networks. It helps readers reflect on several practical problems of deploying SDN and learn how to solve the problems by taking advantage of existing machine learning techniques. The book elaborates on the formulation of each problem, explains design details for each scheme, and provides solutions by running mathematical optimization processes, conducting simulated experiments, and analyzing the experimental results.

Table of Contents

Chapter 1. Machine Learning for Software-Defined Networking
In this chapter, we introduce software-defined networking, and its two typical application scenarios: wide area networks and data center networks. We also briefly introduce emerging machine learning techniques to improve network performance that are used in the rest of this book.
Zehua Guo
Chapter 2. Deep Reinforcement Learning-Based Traffic Engineering in SD-WANs
In this chapter, we introduce ScaleDRL, which combines the control theory and DRL to achieve an efficient network control scheme for Traffic Engineering (TE). ScaleDRL employs the pinning control to select a subset of links in the network as critical links and uses a DRL algorithm to dynamically adjust link weights of the critical links. Thus, the dynamic link weight adjustment coupled with the weighted shortest path algorithm enables dynamic adjust most of the forwarding paths of flows.
Zehua Guo
Chapter 3. Multi-Agent Reinforcement Learning-Based Controller Load Balancing in SD-WANs
In this chapter, we introduce a dynamic controller workload balancing scheme named MARVEL using emerging MARL for switch migration. We design a DRL framework for each agent in the MARL model. The DRL-based solution takes the workload pattern in the control plane as input and generates the migration decision as the output. When training is done, the DRL agent can quickly and accurately decide how to migrate switches among the controllers.
Zehua Guo
Chapter 4. Deep Reinforcement Learning-Based Flow Scheduling for Power-Efficient Data Center Networks
In this chapter, we introduce SmartFCT, which employs DRL coupled with SDN to improve the power efficiency of DCNs and guarantee FCT. SmartFCT dynamically collects traffic distribution from switches to train its DRL model. The well-trained DRL agent of SmartFCT can quickly analyze the complicated traffic characteristics and adaptively generate an action for scheduling flows and deliberately configuring margins for different links. Following the generated action, flows are consolidated into a few of active links and switches to save power, and fine-grained margin configuration for active links avoids FCT violation of unexpected flow bursts.
Zehua Guo
Chapter 5. Graph Neural Network-Based Coflow Scheduling in Data Center Networks
In this chapter, we introduce DeepWeave, a DRL framework to generate coflow scheduling policies. DeepWeave works for both the intra-coflow scheduling and inter-coflow scheduling. To improve the inter-coflow scheduling ability in the job, DeepWeave employs a GNN to process directed-acyclic graph information. DeepWeave learns from the historic workload trace to train the neural networks of the DRL agent and encodes the scheduling policy in the neural networks, which make coflow scheduling decisions without expert knowledge or a pre-assumed model.
Zehua Guo
Chapter 6. Conclusion and Future Work
In this book, we introduce two critical application scenarios for SDN: SD-WANs and SD-DCNs. To improve network performance of SD-WANs and SD-DCNs, we leverage emerging ML techniques (i.e., DRL, MARL, and GNN) to maintain the load balance, increase the power efficiency, and improve the QoS. This book exhibits the effectiveness of ML for solving networking problems and paves the way for the future research on the usage of DRL, MARL, and GNN for computer networks.
Zehua Guo
Bringing Machine Learning to Software-Defined Networks
Zehua Guo
Copyright Year
Springer Nature Singapore
Electronic ISBN
Print ISBN

Premium Partner