Research On Key Techniques Of Virtual Server Consolidation

Virtual server consolidation is to use virtual machines to encapsulate applcations running on multiple physical servers in the cluster and then integrate them into a small number of servers. Nowadays, with the expanding of enterprise-class data centers, virtual server consolidation can reduce large number of servers in order to help the enterprises reduce hardware and operational costs significantly and improve server utilization greatly. How to solve the three problems: virtual cluster monitor, virtual machine packing and virtual machine migration, is the key to enhance the performance of virtual server consolidation.For Xen virtual cluster, this paper designs and implements VSCM, a virtual server consolidation manager. The whole work can be devided into four parts. 1) We design the general framework of VSCM and divide it into three functional modules. 2) We study the technique of virtual cluster monitor, propose the information model of monitor, design and implement a virtual cluster monitor system Prospector, which can dynamically collect all real-time information on physical servers and their virtual machines placed on. 3) We study the technique of virtual machine packing, and propose a strategy of global optimal virtual machine packing, GOVMP, which can significantly reduce the number of servers from the perspective of global optimization, solving the problem of virtual machine packing. 4) We study the technique of virtual machine migration, and propose a strategy of feedback based virtual machine migration, FVMM, which uses feedback mechanism to greatly reduce the migration overhead and guarantee the feasibility of migration commands generated through removing migration constraints, solving the problem of virtual machine migration.Experiment results in Xen virtual cluster show that, the virtual server consolidation manager which is designed and implemented can solve the three problem of virtual server consolidation, and greatly reduce the migration overhead while significantly reducing the number of servers.