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.