Deep Reinforcement Learning Based Routing Scheduling Scheme for Joint Optimization of Energy Consumption and Network Throughput

With traffic demands growing exponentially, a great number of new network applications emerging, traffic load balancing and resource utilization have become the key issues that severely affect network performance of data center networks (DCNs). The efficient routing scheduling scheme is considered as the key factors to affect the network throughput and resource consumption. To maximize the network throughput and reduce the resource consumption, this paper investigates the efficient routing scheduling scheme by joint optimizing the network throughput and energy consumption. We first formulated the problem as a mixed integer nonlinear program (MINLP) problem, and then introduce the deep reinforcement learning based routing scheduling problem to solve it. Results show that the proposed scheme can significantly improve the network performance.