Real-Time Distributed Scheduling of Precedence Graphs on Arbitrary Wide Networks

Previous work on scheduling dynamic competitive jobs is focused on multiprocessors configurations. This paper presents a new distributed dynamic scheduling scheme for sporadic real-time jobs with arbitrary precedence relations on arbitrary wide networks. A job is modeled by a directed acyclic graph (DAG). Jobs arrive on any site at any time and compete for the computational resources of the network. The scheduling algorithm developed in this paper is based upon a new concept of computing spheres in order to determine a good neighborhood of sites that may cooperate for the execution of a job if it cannot be guaranteed locally. The salient feature of this new concept is that it allows the algorithm to be performed on arbitrary wide networks since it uses a limited number of sites and communication links.