Deep reinforcement learning for multi-resource cloud job scheduling

Abstract

The resource scheduling problem in the cloud environment has always been a difficult and hot research field of cloud computing. The difficult problem of online decision-making tasks for resource management in a complex cloud environment can be solved by combining the excellent decision-making ability of reinforcement learning and the strong environmental awareness ability of deep learning. This paper proposes a multi-resource cloud job scheduling strategy in cloud environment based on Deep Q-network algorithm to minimize the average job completion time and average job slowdown. The experimental results show that the scheduling strategy is better than the scheduling strategy based on the standard policy gradient algorithm, and accelerate the convergence speed.