Energy Efficient Traffic-Aware Virtual Machine Migration in Green Cloud Data Centers

Abstract

Cloud computing has overtaken the traditional computing technologies by providing virtualized resources on demand. Cloud data centers consume a huge amount of power and emit much carbon dioxide, resulting in two challenging problems: high energy consumption and global warming. Our prior work has derived an enhanced energy model that considered energy consumed in computing, migration and host reactivating, and proposed three highly efficient virtual machine (VM) migration schemes. In this work, the three schemes are further enhanced, through considering the traffic factor in VM migration and adopting VM clustering. The resulting schemes have significantly reduced the number of VM migration and their migration costs. The time complexities of these schemes have been analyzed, their performance evaluated through simulation. Results showed that, comparing with the ones without VM clustering, the migration energy usage is dropped to 32%, resulting in the total energy saving of 23%, with a small increase in SLA violation. The proposed schemes would have significant impact when applying to virtualization of wireless mobile networks, in which the communication cost, including bandwidth and delay factors, is significant.