Statistical analysis of cloud based scheduling heuristics

Abstract

Scheduling of cloudlets (tasks) on virtual machines in cloud has always been of prime concern. Various heuristics have already been proposed in this area of research and are well documented. In this work, authors have proposed a unique method of statistically evaluating the results of simulation of these heuristics for cloud-based model. The results are evaluated for a standard set of performance metrics. The statistical method applied proves the reliability of simulation results obtained and can be applied to evaluation of all heuristics. In addition to this a recent and more advanced CloudSim Plus simulation tool is used as there is paucity of work that demonstrates using this tool for this research problem. The simulations use a standard model of task and machine heterogeneity that is pertinent to cloud computing. To make the simulation environment more realistic, Poisson distribution is used for the arrival of cloudlets, and exponential distribution for length (size) of cloudlets (tasks).