Capacity quantification of virtual machines in cloud

Abstract

Virtual machines are the computing resources in cloud computing architectures. Job scheduler assigns users' requests into these computing nodes. This assignment principle is governed by the load balancing strategy. Therefore, adopting a suitable load balancing principle plays a key role in highperformance tuning. Equal load distribution across the computing resources is a desirable objective of any job scheduling algorithm. In this paper, we introduce a methodology that can be incorporated in the equal load balancing principle. This methodology quantifies each computing node's capacity in terms of percentage. Each virtual machine is configured with different parameters. We used load capacity as the parameter for assessing the capacity of the computing node. This novel approach uses the z-score statistical method to perform the quantification process. Based on the quantified value, the total workload is proportioned and assigned to each node. We also presented the equal load balancing algorithm that uses the z-score. Experimental results prove that the proposed principle yields better performance when compared to the round robin and throttled load balancing algorithms.