基于增强学习算法的云资源动态弹性伸缩

Dynamic elastic scaling of cloud resources based on reinforcement learning algorithm

为解决云资源节点负载不均衡导致利用效率低、资源负载均衡性较差等问题,提出以增强学习算法改善云资源动态负载均衡弹性伸缩的方法。采用云资源动态特性基准分析云资源动态的最初基准,对云资源动态特性进行节点采样,设置云资源节点负载的上限和下限,实现云资源动态负载均衡控制;采用非均匀离散傅里叶变换方法,对其节点负载状态均匀调整,引入增强学习算法中Q学习方法,不断自适应执行弹性收缩策略,实现云资源动态负载均衡的智能弹性伸缩。实验结果表明,采用所提方法后,云资源利用效率高达约97%,资源负载均衡性能得到提升。

In order to solve the problems of low utilization efficiency and poor resource load balance caused by unbalanced load on cloud resource nodes,a study is proposed to improve the elastic scaling of cloud resource dynamic load balance to enhance learning algorithm.The dynamic characteristics of cloud resources is adopted to analyze the dynamic characteristics of cloud resources.Node sampling is conducted on the dynamic characteristics of cloud resources,and upper and lower limits of cloud resource node load are set to realize the dynamic load balancing control of cloud resources.The non-uniform discrete Fourier transform method is adopted to adjust the load state of the nodes uniformly,the Q learning method in the reinforcement learning algorithm is introduced,and the elastic shrinkage strategy is continuously adaptive to realize the intelligent elastic shrinkage of cloud resources with dynamic load balancing.The experimental results show that the cloud resource utilization efficiency is up to about 97%after adopting the proposed method,and the resource load balancing performance is improved.