A Dynamic List Scheduling Algorithm for Scheduling HPC Application in a Cloud Environment

Cloud computing has seen tremendous growth over the last few years. The pay-as-you go model has enabled a large number of users to host their applications in cloud. High Performance Computing (HPC) applications are used by engineers and scientists to solve complex engineering and scientific problems. These applications require high bandwidth networking, and very high compute capabilities. HPC applications are bandwidth intensive, and are cost vs. performance sensitive. These are also power performance. The scientists have to wait in large queues to get access to hardware sysytems to run their applications. For these reasons, lately, many scientists are moving to the cloud to run their applications. This paper proposes a dynamic scheduling algorithm that is a modified version of the list scheduling algorithm which schedules a HPC application onto cloud resources so as to minimize the makespan. It is based on the concept of Critical Path and is dynamic because the priority of tasks to be scheduled at each step is re-calculated based on the already scheduled tasks.