Research on a cloud resource scheduling strategy based on asynchronous reinforcement learning

Abstract

The effective management of cloud-based IT infrastructure resources plays an important role in the development of grid business and the reduction of operation and maintenance costs. For cloud resource scheduling, there are many factors that affect its performance, and it is difficult to use general methods to effectively solve the problem of cloud resource scheduling. In order to achieve efficient resource scheduling, this paper proposes a cloud resource scheduling strategy based on reinforcement learning. At the same time, in order to deal with the problem of slow convergence speed and low accuracy when the exploration and update of a single agent, By introducing a heterogeneous model to construct a cloud resource scheduling mechanism, which uses multithread to explore the environment at the same time to improve the convergence speed. Experiments show that the scheduling strategy of this method has better performance than the random scheduling strategy.