Home // PESARO 2015, The Fifth International Conference on Performance, Safety and Robustness in Complex Systems and Applications // View article

Estimation of Job Execution Time in MapReduce Framework over GPU clusters

Authors:
Yang Hung
Sheng-Tzong Cheng
Chia-Mei Chen

Keywords: MapReduce; Stochastic Petri Net; Estimation of execution time

Abstract:
The development of Graphic Processing Unit (GPU) makes it possible to put hundreds of cores in one processor. It starts a new direction of high-speed computing. In addition, in a Cloud computing environment, MapReduce over GPU clusters can execute graphics processing or general-purpose applications even much faster. It is crucial to manage the resources and parameters on GPU devices. In this paper, we study the execution time of MapReduce tasks over GPU clusters. We use Stochastic Petri Net to analyze the influence of GPU computing and develop SPN-GC model. The model defines formulas of every stage’s execution time and estimates the execution time under different input data size. Our experimental result presents the comparison between the estimated execution time and actual values under different input data size. The error range is found out to be within 10%. This paper can be a useful reference when a developer is tuning the program.

Pages: 15 to 20

Copyright: Copyright (c) IARIA, 2015

Publication date: April 19, 2015

Published in: conference

ISSN: 2308-3700

ISBN: 978-1-61208-401-5

Location: Barcelona, Spain

Dates: from April 19, 2015 to April 24, 2015