Metaheuristic-Based Virtual Machine Task Migration Technique for Load Balancing in the Cloud

Abstract

Cloud service centers require an efficient load balancing strategy to reduce the excessive workload on some virtual machines (VM). The VM migration technique, which achieves migration of overloaded VM from one physical machine to another, is quite popular. This technique consumes excessive time and monetary cost. Instead of migrating actual VM, migrating the extra tasks of overloaded VM has been found to be more beneficial with respect to time and cost (Ramezani et al. in International Journal of Parallel Programming 42:739–754, 2014, [1]). The task migration technique does not pause the overloaded VM, and the VM pre-copy process is not involved. This technique also provides other advantages such as elimination of VM downtime, no loss of customers’ recorded activities, and better quality of service (QoS) to the customer. The VM task migration technique presented in (Ramezani et al. International Journal of Parallel Programming 42:739–754, 2014, [1]), utilizes an ineffective discriminant function to identify overloaded VM. This discriminant function may falsely identify non-overloaded VM as overloaded VM. In this chapter, a new discriminant function is designed to identify actual overloaded VM. Cost functions are designed to model the actual cost of performing task migration. A particle swarm optimization (PSO) technique is proposed to search for efficient task migration strategies. The proposed technique PSOVM is simulated in MATLAB, and the results are compared with a contemporary technique (Ramezani et al. in International Journal of Parallel Programming 42:739–754, 2014, [1]). The simulated results exhibit greater effectiveness of the proposed technique in identifying actual overloaded VM for task migration.