Deadline-driven task scheduling in clustered heterogeneous distributed systems

摘要

A distributed system can be viewed as a collection of autonomous computing nodes and communication resources, shared by active users. Because of a large number of job arrivals in some nodes, they may be overloaded while other nodes may remain idle as the response time, storage space, and processing speed differ. Real-time scheduler must consider deadlines of tasks while scheduling. Meeting the deadlines of tasks in a distributed system depends crucially on allocating tasks to idle or lightly loaded and high-speed processors. In this paper, a two-level dynamic scheduling model is proposed by considering deadline of each task. The proposed model is semi-distributed as each cluster is represented by a cluster master, which controls scheduling decisions of worker nodes.