Abstract
Cloud computing has emerged as a new cost-efficient technology that provides on-demand resources over the Internet for users who pay only for their actual use. Load balancing plays an important role in cloud computing; it schedules the tasks on the virtual machines effectively to ensure cost-efficient execution of users tasks and optimal utilization of cloud resources. Because load balancing is a NP-hard optimization problem, much effort has been directed toward proposing fast algorithms that approximate the optimal solution. This paper deals with this problem and proposes new load balancing algorithms that meet requirements of cloud users and providers by reducing the makespan and improving resource utilization. For this, we modeled load balancing as a bin-stretching problem. By adopting the Worst-Fit heuristic to the bin-stretching problem, we propose a new load balancing algorithm called Worst-Fit-Based Load balancing algorithm (WFBLBA). Furthermore, by investigating the Decreasing Worst-Fit heuristic, we propose a decreasing variant of load balancing algorithm (WFDBLBA). Experimental evaluation using CloudSim simulator show that our algorithms not only outperform compared heuristics in terms of makespan, resource utilization and waiting time, but also cope better with high machine heterogeneity than compared ones.