Energy Efficient Virtual Machine Consolidation for Cloud Environment

Abstract

Cloud computing offers an on-demand computing model which has transformed the working environment of IT companies and thus, the trend of cloud computing has increased with the passage of time. Out of many different services of the cloud environment, IaaS Infrastructure as a service is one of the leading services which has given the birth to pay per use model and on the basis of this model, cloud service providers offer these services to the users. To satisfy the requirements of cloud users, it requires a large number of physical devices and, it requires a massive amount of electricity to power and cools down the several electronic components present inside these devices. It has been observed that the electricity consumption of data centers is 1.1% to 1.5% of the total electricity consumed all around the world and it is rising by 12% every year. This high consumption of electricity results in the emissions of high carbon dioxide and therefore, it is very necessary to reduce the consumption of energy for the betterment of environmental sustainability. Energy consumption of data center can be reduced by using minimum number of resources and improvement of their utilization. Moreover, the consumption of energy can be minimized by deactivating and reactivating the physical hosts present inside data center. newlineThis thesis presents a new model, algorithms and heuristics for all different steps of dynamic virtual machine consolidation in Infrastructure as a service architecture that meets the requirements of SLA agreement and deals with energy-performance trade-off. The process of virtual machine (VM) consolidation has been selected for the improvement of energy consumption along with the use of live migration of virtual machines to minimize the number of active physical hosts. Thus, for the process of energy efficient VM consolidation, we have proposed a median based threshold approach (MEDTH) for the selection of underutilized and overutilized host present inside the datacenter.