Effects of correlation-based VM allocation criteria to cloud data centers

Abstract

Virtualization technology has been widely adopted in Cloud data centers for adaptive resource provisioning. With virtualization, multiple virtual machines (VMs) can be collocated on a single physical host to yield maximum efficiency. However, VMs which show high CPU utilization correlations to other co-located peers are more likely to trigger overloading incidents. This work provides an analysis on effects of correlation-based VM allocation criteria to Cloud data centers. The correlations among VMs' CPU utilizations are considered as parameters for decision making in VM allocation processes. Three different expressions of correlation-based criteria are introduced and evaluated in this work. According to our simulation results obtained from CloudSim with real-world workload traces, Cloud data centers with correlation-based allocation criteria can perform better in terms of reducing energy consumption and avoid committing Service Level Agreements violations than those with power-based criteria.