Virtual Machine Allocation Policy Optimization Based On Workload Prediction

Research have shown that servers operate at 10%-50% of their full capacity.Dynamic consolidation of virtual machines using live migration and switching idle nodes to the sleep mode allows Cloud providers to optimize resource usage and reduce energy consumption.However,the obligation of providing high quality of service to service to customers leads to the necessity in dealing with the energy-performance trade-off,as aggressive consolidation may lead to performance degradation.We adapt the Random Forest algorithm to predict the workload so that we can get a better usage of resources in the servers and reduce the migration of VMs,while meeting performance requirement.The proposed algorithm is based on learning the VMs' historic data of resource usage,by establishing thousands of CARTs to predicting the future resource demands of workload.We set an adaptive utilization threshold to determining whether the host is overload because of the diversity and complexity in Cloud environment;we adapt the Maximum Correlation policy to ensuring the efficiency of resource utilization in servers by selecting those VMs to be migrated that the highest correlation of the resource utilization with other VMs;we take the VM placement as a bin packing problem,and we apply the Best Fit Decreasing algorithm.Tests have proved that our method can improve the prediction precision by 8% at most compared to the traditional ways.We combine the predict algorithm and the VMs' allocation policy,and apply it to the CloudSim.Tests have proved that our method can reduce 22.7% energy consumption of servers,44.1% of VMs' migrations,and 49.8% of SLAV time than traditional ways.