Automatic configuration of virtual machine resources based on reinforcement learning

Automatic configuration of virtual machine resources based on reinforcement learning

Li Wenxian Peng Zhiping

Virtual machine technology allows multiple virtual machines to share resources on the same physical host. Resources allocated to virtual machines should be dynamically reconfigured to respond to changes in application requirements or to changes in the supply of resources. Therefore, this paper proposes an algorithm based on reinforcement learning to automatically process the configuration process, namely (Standard Reinforcement Learning auto-configuration). This paper emphasizes the problem of the stability and adaptability of the resource management system based on the model of the algorithm. This is done by implementing a representative server load experiment in a cloud test bed based on virtual machines in a cloud environment simulation software Cloudsim. The results proved to be effective. This method can find the optimal (or near optimal) configuration strategy in a small scale system, and it shows good stability and adaptability.

Automatic configuration of virtual machine resources based on reinforcement learning