Multi-Resource Fairness: Objectives, Algorithms and Performance

Abstract : Designing efficient and fair algorithms for sharing multiple resources between heterogeneous demands is becoming increasingly important. Applications include compute clusters shared by multi-task jobs and routers equipped with middle-boxes shared by flows of different types. We show that the currently preferred objective of Dominant Resource Fairness (DRF) has a significantly less favorable efficiency-fairness tradeoff than alternatives like Proportional Fairness and our proposal, Bottleneck Max Fairness. We propose practical algorithms to realize these sharing objectives and evaluate their performance under a stochastic demand model. It is shown, in particular, that the strategyproofness property that motivated the choice of DRF for an assumed fixed set of jobs or flows, is largely irrelevant when demand is dynamic .
Liste complète des métadonnées

Cited literature [22 references]  Display  Hide  Download

https://hal.inria.fr/hal-01243985
Contributor : Thomas Bonald <>
Submitted on : Tuesday, December 15, 2015 - 1:40:41 PM
Last modification on : Wednesday, February 20, 2019 - 2:40:00 PM
Document(s) archivé(s) le : Saturday, April 29, 2017 - 3:07:26 PM

File

paper.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01243985, version 1

Citation

Thomas Bonald, James Roberts. Multi-Resource Fairness: Objectives, Algorithms and Performance. ACM Sigmetrics, 2015, Portland, United States. ⟨hal-01243985⟩

Share

Metrics

Record views

490

Files downloads

240