Enhanced cluster computing performance through proportional fairness - Archive ouverte HAL Access content directly
Journal Articles Performance Evaluation Year : 2014

Enhanced cluster computing performance through proportional fairness

(1, 2) , (1, 3)
1
2
3

Abstract

The performance of cluster computing depends on how concurrent jobs share multiple data center resource types like CPU, RAM and disk storage. Recent research has discussed efficiency and fairness requirements and identified a number of desirable scheduling objectives including so-called dominant resource fairness (DRF). We argue here that proportional fairness (PF), long recognized as a desirable objective in sharing network bandwidth between ongoing flows, is preferable to DRF. The superiority of PF is manifest under the realistic modelling assumption that the population of jobs in progress is a stochastic process. In random traffic the strategy-proof property of DRF proves unimportant while PF is shown by analysis and simulation to offer a significantly better efficiency-fairness tradeoff.
Fichier principal
Vignette du fichier
1404.2266.pdf (216.09 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-01112964 , version 1 (04-02-2015)

Identifiers

Cite

Thomas Bonald, James Roberts. Enhanced cluster computing performance through proportional fairness. Performance Evaluation, 2014, 79, pp.134 - 145. ⟨10.1016/j.peva.2014.07.009⟩. ⟨hal-01112964⟩
236 View
265 Download

Altmetric

Share

Gmail Facebook Twitter LinkedIn More