Multiple DAG Applications Scheduling on a Cluster of Processors

Abstract

Many computational solutions can be expressed as Directed Acyclic Graph (DAG), in which nodes represent tasks to be executed and edges represent precedence constraints among tasks. A Cluster of processors is a shared resource among several users and hence the need for a scheduler which deals with multi-user jobs presented as DAGs. The scheduler must find the number of processors to be allotted for each DAG and schedule tasks on allotted processors. In this work, a new method to find optimal and maximum number of processors that can be allotted for a DAG is proposed. Regression analysis is used to find the best possible way to share available processors, among suitable number of submitted DAGs. An instance of a scheduler for each DAG, schedules tasks on the allotted processors. Towards this end, a new framework to receive online submission of DAGs, allot processors to each DAG and schedule tasks, is proposed and experimented using a simulator. This space-sharing of processors among multiple DAGs shows better performance than the other methods found in literature. Because of spacesharing, an online scheduler can be used for each DAG within the allotted processors. The use of online scheduler overcomes the drawbacks of static scheduling which relies on inaccurate estimated computation and communication costs. Thus the proposed framework is a promising solution to perform online scheduling of tasks using static information of DAG, a kind of hybrid scheduling.