On-Line Parallelizable Task Scheduling on Parallel Processors

Abstract

In a parallelizable task model, a task can be parallelized and the component subtasks can be executed concurrently on multiple processors. The problem of on-line scheduling a stream of independent parallelizable task on a set of parallel identical processors is considered. The processors service the subtasks independently. An analytical model approach based on continuous-time Markov chain is proposed. We study finite capacity queuing model by obtaining numerical expression for the sojourn time. The goal is to choose a technique for evaluation of a resource allocation policy.