Resource Utilization Prediction Model for Efficient Dynamic Virtual Machine Consolidation in Cloud

Virtual Machine (VM) consolidation is an optimization approach for VM placement in cloud infrastructure, which is one of the effective ways to efficiently utilize cloud resources in order to optimize number of VM migrations, Service Level Agreement (SLA) violations and energy consumption problems. Static VM consolidation strategies have not been effective in handling variation of workloads in the cloud systems. Dynamic VM strategies have been proposed in the literatures. However, in the dynamic VM consolidation process, most existing algorithms consolidate active servers mainly based on the current resource requirements and forgo the demands of resources in the future during VM allocation. Thus, they result in needless VM migrations and cause high rate of SLA violations in data centers. This research proposed a prediction-based VM consolidation approach. The proposed method utilized a multi-resource utilization to predict the current and future CPU and memory utilization of active servers during allocation stage. The proposed method was compared with the existing one that did not consider future utilization of resources. CloudSim toolkit 3.0 simulator was used to implement and evaluate the performance of the algorithms. The proposed algorithm was found to have improved in terms of number of VM migrations and SLA violations. Using the number of hosts (400, 600, and 800), the proposed method was 28%, 32% and 38% decrease in number of VM migrations and 0.03243%, 0.03247% and 0.01474% decrease in SLA violations compared with the existing method, respectively. Recommendations and future directions were suggested for further improvements