On the Soft Real-Time Scheduling of Parallel Tasks on Multiprocessors

Recently, parallel task models have received more attentions as the prevalence of multiprocessors. The general purpose by using such parallel programming models is to guarantee bounded response time with minimal resource. Unfortunately, most previous work focus on hard real time problem while could not providing such performance guarantees. In this paper, we address the soft real-time scheduling problem under a general DAG (Directed Acyclic Graph) task model and present conditions where each parallel application with arbitrary deadline can achieve a bounded response time by using federated scheduling algorithm. To the best of our knowledge, this is the first time to consider the soft real-time scheduling under general DAG task model on multiprocessors.