Scheduling of Parallel Tasks in Cloud Environment Using DAG MODEL

Abstract

The scheduling of tasks in a heterogeneous multiprocessor system in the cloud is still a demanding problem that is being explored by many researchers. Parallel computing which itself is a research area is now integrated with cloud computing. In this paper, we present a systematic study of a directed acyclic graph (DAG) model in a parallel multiprocessor system. Basic concepts of parallel computing along with issues like task scheduling have been discussed in detail. Computational solutions can be demonstrated as directed acyclic graphs (DAG) model having edges and nodes with weights. Each task or job in the DAG has its own execution time, which incorporates into various processors. This paper gives the basic idea of parallel processing using DAG. The main objective behind scheduling using the DAG model is to reduce the finish time or completion time of parallel applications through proper assignment of tasks among multiprocessors. Most of these schemes consider the precedence constraints among tasks. Up to now, Direct Acyclic Graph (DAG) is considered a prominent approach used for modeling the precedence constraints between the nodes or tasks. There are various scheduling algorithms that use DAG for scheduling jobs. The most popular heterogeneous earliest finish time (HEFT) is one of them. A literature review on various task scheduling schemes in combination with artificial intelligence concepts is also presented.