Self-configurable scheduling algorithm for heterogeneous computing systems

Abstract

Research in real time task scheduling algorithm is a challenging problem for high performance computing system; especially achieving mapping of tasks to processors is a key design issue in the heterogeneous computing environment. There are many existing scheduling algorithms covered in the literature but none of them mapped the specific task to the corresponding processor. In this paper, we build a new scheduler model that is suitable to provide timing requirement constraints, (specific processor, specific task) pair and load balancing taken into account. This paper also addresses percentage of tasks missing their deadline due to the running out of resources of the system. When scheduler makes use of cloud of resources, the missing tasks can be scheduled properly. The new scheduler combines all types of real time and non real-time tasks and schedules them. Hereafter, we call this algorithm as Self-Configurable Scheduling (SCS) Algorithm. This algorithm automatically adjusts the jobs among processing elements and there will be a feedback loop from each processing element to the load balancer. We will build a simulation that can estimate shortage of resources in order to add resources and also allocate missed tasks to the cloud of resources. © 2011 Springer-Verlag.