Research Of The Deployment And Dynamic Migration Of Virtual Cluster On Cloud

Cloud computing is an innovative method, which integrates resources in computer like hardware and software. And it can serve their customer any time through the internet in a convenient way. The entity of the service can be infrastructure, platform or software. The customer can get what they need due to the variety of the service, and can enjoy the convenience of technology.As the great demand for cloud service, the energy waste of the cloud service provider in physical cluster data center followed. There is a survey shows that70%of the pay in cloud data center comes from the infrastructure and the power consumption. Saving the energy and increasing the Utilization is extremely important on condition o that service level agreement is guaranteed.This article mainly discuss on strategy of dynamic migration of virtual machine cluster considering energy saving and utilization increasing. The article is based on the research of predecessor, and makes some optimization from the viewpoint of energy saving, discussing the strategy respectively from the creation of the virtual machine, the migration of overloaded virtual machine and the migration of under loaded machine. We defined the Dynamic overload Threshold and Dynamic under-load Threshold which is history load aware. We also proposed Most Fit Selection (MFS) which consider the history load of virtual machine, as well as Weight Placement (WHP) which consider the history load of the host. And we also set up an infrastructure cloud using the open source project CloudStack, to prove the feasibility of the dynamic strategy proposed.Finally, we evaluate the dynamic migration strategy in detail using the CloudSim, which is a brilliant tool for the testing of cloud algorithm. The result shows that the MFS strategy can decrease the times of migration and reduce energy consumption; the DTH can weaken the degradation of energy consumption of dynamic migration; WHP propose a flexible strategy to fit different environment through adjust the weight of different factor.