Power-aware virtual machine scheduling technique for cloud

Abstract

Cloud Computing offers services to end-users rather than a product, by sharing resources, software and other information under a usage based payment model. It enables hosting of various kinds of applications such as business, scientific, social network, etc. as it has key characteristics like multi-tenancy, scalability, performance, security, etc. Economic benefits are the main driver for the Cloud, since it promises the reduction of Capital Expenditure (CapEx) and Operational Expenditure (OpEx). While Cloud Computing provides many benefits, it still has some issues as well. In recent years, Cloud Computing is facing many challenges like Data Security, Energy Consumption, Server Consolidation, Virtual Machine Migration, etc. This research work focuses on the study of energy-aware management of VMs in heterogeneous cloud environment. Energy-aware management of VMs in Cloud data centers point towards systematic and correct scheduling of server resources such that energy consumption and SLA violations are minimum, resulting in reduced operational costs which benefits end-users with decreased prices for resource usage. In this thesis, existing virtual machine management approaches have been surveyed and a power-aware virtual machine scheduling policy is designed for Cloud systems. In particular, an optimized technique for efficient server allocation is discussed. Objective of this technique is to place VMs on host while keeping total utilization of CPU below defined threshold and then optimizing the VM allocation by constrained consolidation of VMs and switching idle nodes to sleep mode to minimize power consumption. As excessive consolidation of VMs can lead to significant SLA violation, in proposed approach not only overall number of used servers are minimized, but also number of migrations are minimized, resulting in minimum energy consumption and least SLA violations. A simulated environment provided by CloudSim Toolkit has been used to validate experimental results, which demonstrates that the proposed approach can provide substantial energy savings, reduced SLA violations and fewer VM migrations as compared to existing techniques, while preventing frequent power cycling of servers, thus validating the proposed policy.