Load Balancing on Virtual Machines Using Levy Flight Combined with Gray Wolf Optimization Algorithm

Abstract

In a cloud computing environment, virtual machines must be allocated wisely among queued resources waiting for a specific service. In this case, the optimal load balancing on those virtual machines must be done using optimization methods. Because the distribution of processes on the VMs is considered one of the NP-Hard optimization problems, it is necessary to use optimization methods. This paper proposes a hybridized Gray Wolf Optimizer (GWO) and Levy Flight (LF) search approach. The proposed method is called Levy Flight-GWO-Virtual Machine Load Balancing (LF-GWO-VM-LB). Despite the success of the standard GWO alone as a load-balancing method, it needs to improve its explorative capability and premature convergence. Levy flight is an ideal way to increase the exploration capability of the GWO as it introduces randomness and unpredictability, enhancing the diversity of the search process. This increased diversity can prevent the algorithm from getting stuck in suboptimal solutions. The algorithm's performance was measured using the makespan and throughput. However, the last is only used as one of the objectives used in the optimization process. Both makespan and throughput results showed the outperformance of the LF-GWO-VM-LB compared to the other state-of-the-art methods, i.e., QMPSO, MPSO, and the Q-learning methods.