International Journal of Computer Science Issues

More than a traditional journal...

Cloud datacentres are large datacentres with thousands of servers that consume excessive energy and have significant carbon footprints. The increasing cloud users, on the other hand, are demanding more services with better response time. Hence, resources allocation, power management as well as better service delivery to users are challenging tasks for cloud providers. The problem has been formulated as a Bin Packing problem and many algorithms have been proposed with the aim of attaining maximum throughput and minimum computation time in order to achieve an energy efficient datacentre. In this study, a new algorithm called Neighbour-Fit was proposed to address the aforementioned problem. A model based on the algorithm, for Virtual Machine (VM) allocation, was designed. A web-based simulator was also developed using HTML, CSS and PHP to simulate the proposed model. The proposed algorithm was benchmarked with five existing allocation algorithms using throughput and computation time. The Neighbour-Fit algorithm is about 90 percent faster than the Almost Worst-Fit, Best-Fit, First-Fit and Worst-Fit algorithms. Although the Next-Fit algorithm is about 20 percent faster than the Neighbour-Fit algorithm, Neighbour-Fit algorithm utilizes 4 percent less number of servers than the Next-Fit algorithm. This performance infers that the Neighbour-Fit algorithm with a moderate computational time and a high throughput optimizes server utilization. This in turn reduces the power consumed by the servers in cloud data centers.