A Trace Driven Simulation Study of VM Consolidation Algorithms in Cloud Computing

Abstract

Virtual machine consolidation is a prominent method for decreasing consumption of energy in cloud computing. Tremendous amount of work is found in literature for development of techniques for hosts underload, overload detection, virtual machine selection and placement to perform consolidation. The major data set used for testing the performance of these algorithms is publicly accessible Planet lab workload trace. It is essential to know the behavior of standard algorithms for virtual machine consolidation on different workload traces. This paper evaluates the performance of these algorithms on Bitbrains and Google cluster trace. The simulations are planned and performed on CloudSim platform. Results of execution of algorithms on Bitbrains trace are compared with Google workload trace. Results indicate that energy consumption, number of virtual machine migrations, number of hosts and service level agreement violation time per active host of best algorithms differ by 50.2%, 13.6%, 73.9% and 57.1% respectively. Results of objectively comparing these algorithms can contribute in formation of future research strategy in VM consolidation.