Research On Computing Resources Hotspots In Cloud Clusters

Cloud computing has aroused widespread concern because of its convenient physical resource usage pattern in the form of on-demand services.Cloud computing cluster provides resources for various applications through virtualization technology.But the larger cloud computing cluster size and the diversity of applications had brought enormous challenges for the management of the resources in cloud computing.One of the most important goals of Clouds is to provide users with stable physical resource support.But the issue of server resource hotspots in cloud clusters is a serious impediment to this goal.Resource hotspot problem refers to the usage of certain resources on the servers exceeds the limit and thus interfere with the virtual machines which are on the normal operation.In order to ensure the stability of the applications,cloud computing platform must be able to detect hotspot in the cluster,and try to get it back to normal.Therefore,this paper studies the problem,focusing on the hotspot detection and hotspot solution.The research contents are as following:1.We designs and implements a cloud platform monitoring system.The system can collect the resource usage of the servers and the virtual machines and notify the user when the resource hotspot occurs.2.Based on the minimization of migration time,this paper proposes MCMS(Min-Cost-Min-Server)algorithm to develop the best virtual machine migration scheme.The algorithm solves hotspots by migrating virtual machines which are on hotspot servers.And,after completing the hotspot elimination,the algorithm tries to migrate all virtual machines on the physical machine with less resource usage to reduce the number of servers in the cluster,and then reduce energy consumption.3.We designs and implements a hotspot elimination system in this paper.Through the monitoring systems and MCMS algorithm,the system automatically find hotspots and eliminate resource hotspots.4.This paper verifies the feasibility of the hotspot elimination system by deploying it in the real environment.In addition,this paper tests the performance of MCMS algorithm by simulation experiment.Compared with other algorithms,MCMS algorithm can solve the resource hotspot problem with less migration time.And the MCMS algorithm can significantly reduce the number of servers in a cloud cluster.