Abstract

In this paper, we propose resource provisioning as the means to reduce the completion time of scientific workflows in a Grid environment. We propose task clustering as a form of workflow optimization that can be used along with provisioning in order to achieve this reduction in completion time. Provisioning can be done statically using advance reservations (ARs) or using dynamic provisioning mechanisms. A simulation is done using the Maui simulator, a workload trace collected from the NCSA Teragrid cluster and 13 workflows to study the effect of provisioning on the completion time of the scientific workflows. The results show in general a reduction of about 50% in the workflow completion time using provisioning for the First In First Out and fair share scheduling policies. In this paper, we also examine the cost of resource provisioning and propose a utilization‐based metric that can be used to guide the …

Date

2011

Authors

Gurmeet Singh, Ewa Deelman

Journal

Concurrency and Computation: Practice and Experience

Volume

23

Issue

16

Pages

1969-1989

Publisher

John Wiley & Sons, Ltd.