Self-tuning caching: the Universal Caching algorithm: Experiences with Auto-adaptive and Reconfigurable Systems

  • Authors:

  • Ganesh Santhanakrishnan;Ahmed Amer;Panos K. Chrysanthis

  • Affiliations:

  • Department of Computer Science, University of Pittsburgh, Pittsburgh, PA 15260, U.S.A.;Department of Computer Science, University of Pittsburgh, Pittsburgh, PA 15260, U.S.A.;Department of Computer Science, University of Pittsburgh, Pittsburgh, PA 15260, U.S.A.

  • Venue:
  • Software—Practice & Experience
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

A cache replacement policy is normally suited to a particular class of applications, or limited to a set of fixed criteria for evaluating the cache-worthiness of an object. We present Universal Caching as a mechanism to capture the generality of the most adaptive algorithms, while depending on a very limited set of basic criteria for cache-replacement decisions. Our testing was limited to Web workloads, where we tested the Universal Caching policy using real-world traces from both the server-side and client-side proxies. Using a self-tuning mechanism, combined with a generalization of the criteria employed in GD-* Web caching, Universal Caching was able to consistently outperform any other fixed choice of algorithm we tested. Copyright © 2006 John Wiley & Sons, Ltd.