Congestion prices as feedback signals: an approach to QoS management

  • Authors:
  • Rolf Neugebauer;Derek McAuley

  • Affiliations:
  • University of Glasgow, Glasgow, G12 8QQ, Scotland, U.K.;Microsoft Research Limited, St. George House, Cambridge CB2 3NH, U.K.

  • Venue:
  • EW 9 Proceedings of the 9th workshop on ACM SIGOPS European workshop: beyond the PC: new challenges for the operating system
  • Year:
  • 2000

Quantified Score

Hi-index 0.00

Visualization

Abstract

Recently there has been a renewed interest in the application of economic models to the management of computational resources. Most of this interest is focused on pricing models for the Internet; in particular, on congestion or shadow prices, that address the phenomenon of what economists call external costs --- users are exposed to the costs they impose on other users when causing congestion of a resource.This paper describes how congestion prices can be applied to resource management in operating systems. Shadow prices are interpreted as feedback signals to applications which can adjust their resource requirements according to an application-specific strategy. This leads to a decentralised approach of resource management where applications are enabled and encouraged to perform resource and quality tradeoffs themselves. We have implemented a simulation environment and a number of strategies to evaluate the usefulness of congestion prices as a feedback signal and demonstrate that this approach can offer different service levels to different tasks. We also discuss how the simulation results can be applied in a real operating system and how this work can be extended to form a generic resource management framework.