Reseach On Placement And Migration Algorithms Of Virtual Machines In Green Data Centers

With the expanding of Data Centers(DCs) all over the world, the growing environmental pollution of Data Centers due to the high energy comsumption is becoming a main problem. “Green Data Centers” refers to architecture design, protocols, devices, infrastructure, and the breakthrough algorithms, which could enhence energy efficiency and minimize energy consumption. Most of the modern DCs are provisioned for peak load, leading that the servers are ilde most of the time, which are consuming considerable amount of energy. There are two means to save energy based on resources consolidation: virtual machines(VM) consolidating placement and virtual machines consolidating migration. VM consolidating placement aims to put VMs together as much as possible which leads less server usage and energy saving. But excessive consolidation will decrease the reliability when server failure occured or VM demands changed. VM consolidating Migration means remap partial of the VMs and use least servers to hold, which actually is the change of VMs embedding blueprint. Previous migration schemes only take resource saving into consideration without network cost generated by inter-VMs bandwidth demands.Centering on resource consolidation, the article mainly studied the following three peoblems.(1) Tenant-Aware virtual machine placement problem: When a failure occurs on a server, all the VMs on this server belonging to some tenants will stop working. To solve this problem, we need to locate all the VMs of each tenant dispersedly on as many servers as possible. By add an up threshold of VMs number on server for each tenant, the model could decrease the effect rate of each tenant. Then the author gave an algorithm named Grouped First Fit Decrease placement algorithm to solve the problem.(2) Dynamic Traffic-Aware virtual machine placement problem: Dynamic traffic can be represented as stochastic variable. As experiment shows that most VMs in date centers keep a relatively stable traffic rate during a long time, resource allocation following the maximum demand of stochastic range for each VM will make a great waste of resource. In the article, a solution that using a factor ? to control the number of demands with stochastic feature, is proposed. It saves resource usage efficiently and could satisfy the specific SLAs at the same time. Then we propose a robust optimization algorithm and the ? factor heuristic algorithm to solve the two dimension resource demands placement problem.(3)Traffic-Aware virtual machine migration problem: VM traffic that crossing racks and networks may cause the waste of network resources. When migrating VMs to consolidating, we could also take network cost saving into consideration. Differing from any other researches, there are three optimized objectives in this migration model, including minimizing active servers’ number, minimizing traffic cost and minimizing VMs’ migration cost. In this part, an algorithm named Minimum Traffic Cost Best Fit Migration is proposed, which sacrifice a little of consolidation performance and migration cost to get theh better traffic cost.