Latency Aware Scheduling Policy for Tasks in IaaS Cloud

In Infrastructure as a Service cloud computational resources of different clouds within its federation can be used in the form of leases. This may result in tasks getting completed in other clouds that have more computational power or currently have free resources to service the request. In this paper a new task scheduling approach has been considered that not only considers the effective finish time of clouds as parameters similar to list scheduling, but also takes into account the latency of networks involved in the federation. This consideration is due to the fact that the estimated finish time may differ from actual finish time of tasks when there is a communication delay between clouds in the federation. When dependency among tasks are considered it becomes necessary that a dependent task can start execution only if its predecessors have completed. So if the predecessor task has been scheduled in a different cloud the calculation of estimated finish time with communication delay will help in finding the proper cloud to schedule the task. This would not be possible if the communication delay is not considered as a parameter in estimating the finish time of task. By this approach makespan time of an application which is a set of tasks can be considerably reduced. The experimental results show that the proposed method outperforms the existing list scheduling algorithm.