Correlation Based Vms Placement Resource Provision

In Virtualized Data Center (VDC), a single Physical Machine (PM) is logically divided into one or more Virtual Machines (VMs) that share physical resources. Therefore, dynamically resource provisioning plays an important role in VDC. Moreover the resource provider would like to maximize resource utilization, which forms a large portion of their operational costs. To achieve this goal, several consolidations can be used to minimize the number of PMs required for hosting a set of VMs. However, consolidation is often undesirable for users, who are seeking maximum performance and reliability from their applications. These applications may indirectly increase costs due to under-provisioning of physical resources. Moreover frequent Service Level Agreement (SLA) violations result in lost business. To meet SLA requirements, over provisioning of resources is used. However, if the services do not use all the CPU they have been allocated; the provider will suffer from low resource utilization, since unused resources cannot be allocated to other services running in the provider. Therefore VM provisioning is the most vital step in dynamic resource provisioning. In this paper, correlation based VMs provisioning approach is proposed. Compared to individual-VM based provisioning, correlation based joint-VM provisioning could lead to much higher resource utilization. According to the experimental results, proposed approach can save nearly 50% CPU resource in terms of overall CPU utilization.