Research On Workload Character Based Energy Saving Algorithm In Cloud Datacenter

As the continuous expanding of the size of cloud datacenter, the energy consumption is growing worse. One of the soluation for this problem is virtual machine migration. But numous and frenquency virtual machine migrations have a negativ effect on the quality of service, and consume extra energy. In this paper, we use a algorithm which is based on Capacity Detections and Load Characters, and a modified Algorithm of Adaptive Heuristics for Dynamic Consolidation of VMs to reduce the migrations of virtual machines, so as to reduce the energy consumption of cloud datacenter.The main method of Virtual Machine Selection Algorithm based on Capacity Detections and Load Characters, is that: when a server is overload, get the capacity information of this server firstly, and depending on it, different virtual machine selection policies is used depending on whether it’s a high-capacity server or a low-capacity server. For a low-capacity server, a traditional Minimum Migration Time virtual machine selection policy is applied. For a high-capacity server, a independent designed Minimum Migration policy algorithm which based on history workload, is used to select the virtual machines need to migrate. In this algorithm, history workloads of a virtual machine are calculated with the mount of allocated CPU resource and the utilization of CPU. Then different selection policies are chose, depending on the mount of CPU resource used and the workload of the server currently. If there exits virtual machines, that by migrating one of which, the workload of a server will be at a certain level lower than the setting threshold. Then the history workload based minimum migration policy is used. Otherwise, the traditional Minimum migration time policy is used. Experiments results show that, the number of virtual machine migrations gets a significant reduction when using this algorithm, and the energy consumption decreased at the same time.In order to reduce number of the migrations of virtual machines, foucesd on the flaw of the Algorithm of Adaptive Heuristics for Dynamic Consolidation of VMs, a modified algorithm is proposed in this paper. The main idea of the modified Algorithm of Adaptive Heuristics for Dynamic Consolidation of VMs, is to reduce the number of migrations in the way that depart the determination of the overload of server from the judgment of whether migration should be taken. Detail is that : determinate where a server is overload in a automatically based on ananalysis of historical data; the decide whether migration is necessary, depending no whether the workload of the server is heavy than the setted migration threshold. If so, migrations take palace immidately, and no virtual machins can be created or migrate on this server; otherwise, just stop creating or migrating virtual machines on this server, no virtual machines should be migrate out. The results of the simulations show that with the modified algorithy applied, the migrations of virtual machines decreased, which lead to a reduction of energy consumption of datacenter.