Skip to main navigation Skip to search Skip to main content

Cost performance analysis in parallel computing networks with divisible load scheduling

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

A monetary network cost issue in a parallel computing network with a homogeneous single-level tree topology is discussed. The monetary network cost, which is linearly dependent on the amount of divisible workload, is composed of a communication cost and a computing cost. Through mathematical analysis for the monetary network cost in two different load distribution strategies (sequential distribution and simultaneous distribution), the issue of the relationship between the monetary network cost and ratio of network speed parameters is studied comprehensively. By introducing a new parameter, cost efficiency, a numerical network model suited for cost efficient parallel processing is examined. Simulation results yield insights for trends of network performance against network cost. The numerical analysis and simulation works here are worth being highlighted for high performance parallel computing networks under limited network cost.

Original languageEnglish
Title of host publicationProceedings of the 21st IASTED International Conference on Parallel and Distributed Computing and Systems, PDCS 2009
Pages329-336
Number of pages8
StatePublished - 2009
Event21st IASTED International Conference on Parallel and Distributed Computing and Systems, PDCS 2009 - Cambridge, MA, United States
Duration: Nov 2 2009Nov 4 2009

Publication series

NameProceedings of the IASTED International Conference on Parallel and Distributed Computing and Systems

Conference

Conference21st IASTED International Conference on Parallel and Distributed Computing and Systems, PDCS 2009
Country/TerritoryUnited States
CityCambridge, MA
Period11/2/0911/4/09

Keywords

  • Cost efficiency
  • Divisible load theory
  • Network cost

Fingerprint

Dive into the research topics of 'Cost performance analysis in parallel computing networks with divisible load scheduling'. Together they form a unique fingerprint.

Cite this