Design, analysis, and performance evaluation of an efficient resource unaware scheduling strategy for processing divisible loads on distributed linear Daisy chain networks

Authors:
Bharadwaj Veeravalli;Jingxi Jia
Affiliations:
IEEE, IEEE-CS;CNDS Lab, Department of Electrical and Computer Engineering, The National University of Singapore, Singapore
Venue:
HiPC'08 Proceedings of the 15th international conference on High performance computing
Year:
2008

Citing 8
Cited 0

Divisible task scheduling — concept and verification

Parallel Computing - Special issue on task scheduling problems for parallel and distributed systems
Scheduling Divisible Loads in Parallel and Distributed Systems

Scheduling Divisible Loads in Parallel and Distributed Systems
Ten Reasons to Use Divisible Load Theory

Computer
Performance limits of divisible load processing in systems with limited communication buffers

Journal of Parallel and Distributed Computing
Optimized Distributed Delivery of Continuous-Media Documents over Unreliable Communication Links

IEEE Transactions on Parallel and Distributed Systems
Adaptive Divisible Load Scheduling Strategies for Workstation Clusters with Unknown Network Resources

IEEE Transactions on Parallel and Distributed Systems
Multiround Algorithms for Scheduling Divisible Loads

IEEE Transactions on Parallel and Distributed Systems
Computing BLAS level-2 operations on workstation clusters using the divisible load paradigm

Mathematical and Computer Modelling: An International Journal

Quantified Score

Hi-index	0.01

Visualization

Abstract

In this paper, we present a generalized and novel load distributionstrategy for scheduling divisible loads on linear networks, whenspeeds of the computing nodes and the communication links are unknowna priori. This strategy, which is referred to as Wait and ComputeStrategy (WCS), uses a portion of the total load to estimate speedparameters and then use processors in an iterative manner in phasesto minimize the overall processing time. We present an analysis of thisstrategy with respect to time performance and compare its performancewith the previous works both analytically and through simulation studiesunder several influencing parameters. From our findings, we will showthat the proposed generalized strategy achieves a better performance inmost cases.