Intelligent Cloud Resource Management with Deep Reinforcement Learning

Abstract

The cloud provides low-cost and flexible IT resources (hardware and software) across the Internet. As more cloud providers seek to drive greater business outcomes and the environments of the cloud become more complicated, it is evident that the era of the intelligent cloud has arrived. The intelligent cloud faces several challenges, including optimizing the economic cloud service configuration and adaptively allocating resources. In particular, there is a growing trend toward using machine learning to improve the intelligence of cloud management. This article discusses an architecture of intelligent cloud resource management with deep reinforcement learning. The deep reinforcement learning makes clouds automatically and efficiently negotiate the most appropriate configuration, directly from complicated cloud environments. Finally, we give an example to evaluate and conclude the remarkable ability of the intelligent cloud with deep reinforcement learning.