Task Scheduling Performance Evaluation of Unreliable Virtual Machines and Cloudlets

Abstract

For load balancing in cloud computing environment, scheduling algorithm are used. This paper analyses the research on scheduling algorithms. Further, using cloudsim simulation tools, evaluation of performance of the three algorithms i.e. First Come First Serve (FCFS), Generalized Priority (GP) and Shortest Job First (SJF) was carried out. Three algorithms were compared by varying both virtual machines (VM) and cloudlets, keeping constant VM and increasing Cloudlets, varying the nature of tasks. For all the three algorithms, as the number of cloudlets increases, execution time increases. Generalized Priority algorithm (GP) has lower execution time than both First Come First Serve (FCFS) and Shortest Job First (SJF) algorithms. For homogeneous type of tasks, execution time of FCFS is lower and for heterogeneous type of tasks, execution time of GP is lower. Homogeneous tasks required less execution time than Heterogeneous tasks.