Department of Computer Science & Engineering

Divisible load applications have received a lot of attention in recentscheduling literature. These applications consist of an amount of computation, or load, that can be divided arbitrarily into independent pieces. The problem of Divisible Load Scheduling (DLS) has been studied extensively, but mostly from the theoretical standpoint. In this paper we focus on practical issues and make the following contributions: we implement previously proposed DLS algorithms as part of a generic production Grid application execution environment, APST; we evaluate and compare these algorithms on a real-world two-cluster platform; and weuncover several issues that are critical for using DLS theory in practice effectively. To the best of our knowledge the software resulting from thiswork, APST-DV, is the first usable and generic tool for deploying divisible load applications on current distributed computing platforms.

Pre-2018 CSE ID: CS2004-0785