A method for estimating the execution time of a parallel task on a grid node

Authors:
Panu Phinjaroenphan;Savitri Bevinakoppa;Panlop Zeephongsekul
Affiliations:
School of Computer Science and Information Technology, RMIT University, Melbourne, Australia;School of Computer Science and Information Technology, RMIT University, Melbourne, Australia;School of Mathematical and Geospatial Sciences, RMIT University, Melbourne, Australia
Venue:
EGC'05 Proceedings of the 2005 European conference on Advances in Grid Computing
Year:
2005

Citing 5
Cited 5

Locally Weighted Learning

Artificial Intelligence Review - Special issue on lazy learning
The grid: blueprint for a new computing infrastructure

The grid: blueprint for a new computing infrastructure
Statistical Prediction of Task Execution Times through Analytic Benchmarking for Scheduling in a Heterogeneous Environment

IEEE Transactions on Computers
A decoupled scheduling approach for Grid application development environments

Journal of Parallel and Distributed Computing - Special issue on computational grids
On the Mapping Problem

IEEE Transactions on Computers

RMOT: recursion in model order for task execution time estimation in a software pipeline

Proceedings of the Conference on Design, Automation and Test in Europe
DAGS: distribution agnostic sequential Monte Carlo scheme for task execution time estimation

Proceedings of the Conference on Design, Automation and Test in Europe
A multi-strategy collaborative prediction model for the runtime of online tasks in computing cluster/grid

Cluster Computing
A hybrid heuristic-genetic algorithm for task scheduling in heterogeneous processor networks

Journal of Parallel and Distributed Computing
A heuristic algorithm for mapping parallel applications on computational grids

EGC'05 Proceedings of the 2005 European conference on Advances in Grid Computing

Quantified Score

Hi-index	0.00

Visualization

Abstract

The mapping problem has been studied extensively and many algorithms have been proposed. However, unrealistic assumptions have made the practicality of those algorithms doubtful. One of these assumptions is the ability to precisely calculate the execution time of a task to be mapped on a node before the actual execution. Since the theoretical calculation of task execution time is impossible in real environments, an estimation methodology is needed. In this paper, a practical method to estimate the execution time of a parallel task to be mapped on a grid node is proposed. It is not necessary to know the internal design and algorithm of the application in order to apply this method. The estimation is based upon past observations of the task executions. The estimating technique is a k-nearest-neighbours algorithm (knn). A backward predictor elimination, leave-one-out cross validation, and a statistical technique are used to derive the relevant parameters to be used by knn. Experimental results show that on average the proposed method can produce 2.3 times the number of accurate estimated execution times (with errors less than 25%) greater than the existing method.