Live Migration Decision based on Integrated Server-Storage Virtualization

Abstract

Virtualization can provide significant benefits in data centers by eliminating overload nodes. An increasing workload can be handled by allocating the resources to a virtual server and storage, if idle resources are available on the physical servers, or by simply migration of the virtual machine (VM) to a less loaded physical server. In case of migration, which VM migrate to which server is important for resource utilization. There are many considerations; application-level statistics (such as request rate, service time and response time) and physical resources in host (such as memory, CPU, I/O bandwidth, network bandwidth) and storage node (such as space capacity and I/O capacity). This paper proposes a VM migration decision applying G/G/1 queuing theory. It reduces the overhead in changing data center across multiple resource layers-server and storage nodes. Finally, we will measure how much efficient on open source Xen virtualized infrastructure as an ongoing work.