Use of time-series based forecasting technique for balancing load and reducing consumption of energy in a cloud data center

The long-dreamed vision of “computing as a utility” has finally taken shape in the form of cloud computing. This paradigm shift is the biggest buzz in today's computer world. The pay-as-you-go model of cloud attracts more and more customers towards it. As a result the workload of the data center is increasing enormously. Due to uneven workload distribution some servers become overloaded while some other remains underloaded. For load balancing a few virtual machines are migrated from the over-utilized nodes. But unlike this, a load prediction algorithm is proposed in this paper which will take the decision of migration depending on the present as well as future workload of the system. So our proposed technique does not initiate immediate VM migration as soon as a node is declared overloaded. Our algorithm has been simulated in CloudSim and their performances are compared with the existing benchmark algorithms. Results prove that the proposed technique not only makes the data center more energy efficient but also balances workload more efficiently.