Virtual Machine Dynamic Scheduling Technology Based On Load Forecast Research And Implementation

With the development of cloud computing and virtualization technology,the appliaction of cloud computing more and more widespread,such as Amazon’s AWS,IBM’s Bluemix,Ali cloud feitian platform,open-source system OpenStack.In the cloud computing system there are a large number of virtualized resources species,so that cloud computing system energy saving,efficient use of resources,load balancing and other issues attract more and more attention,and through the virtual machine scheduling can effectively solve these issues.From the direction of the load balancing and improve resource utilization in cloud computing system,we combine load forecasting and dynamic scheduling of virtual machines,and implemented load forecasting based virtual machine dynamic scheduling system in OpenStack system,including load forecasting function and virtual machine dynamic scheduling function.(1)Virtual machine or physical machine load value is a time series,which changes over time,so choose the method of time series forecasting load herein,ARMA(4,3)as the load forecasting model,the load value is divided into different time series,and reduce the forecasting step to improve the accuracy of the prediction.(2)By improving the process of simulated annealing algorithm to generate a new solution that we design a virtual machine dynamic scheduling algorithm,it is divided into virtual machines selection and virtual machine allocation.In the virtual machine selection stage by comparing the load variation of physical and virtual machines select a virtual to migrate.In the virtual machine allocation stage,when a virtual machine that is high load time but low load physical machine as the destination physical machine.In the system we put different virtual machines on the same compute node can achieve efficient use of resources.Through Load forecasting we can know the load variation of physical machine in the next period of time by early,we can avoid system physical machine high load,and makes the system run more stable.