A multi-version cache replacement and prefetching policy for hybrid data delivery environments

  • Authors:

  • André Seifert;Marc H. Scholl

  • Affiliations:

  • University of Konstanz, Konstanz, Germany;University of Konstanz, Konstanz, Germany

  • Venue:
  • VLDB '02 Proceedings of the 28th international conference on Very Large Data Bases
  • Year:
  • 2002

Quantified Score

Hi-index 0.01

Visualization

Abstract

This paper introduces MICP, a novel multiversion integrated cache replacement and prefetching algorithm designed for efficient cache and transaction management in hybrid data delivery networks. MICP takes into account the dynamically and sporadically changing cost/benefit ratios of cached and/or disseminated object versions by making cache replacement and prefetching decisions sensitive to the objects' access probabilities, their position in the broadcast cycle, and their update frequency. Additionally, to eliminate the issue of a newly created or outdated, but re-cacheable, object version replacing a version that may not be reacquired from the server, MICP logically divides the client cache into two variable-sized partitions, namely the REC and the NON-REC partitions for maintaining re-cacheable and nonre-cacheable object versions, respectively. Besides judiciously selecting replacement victims, MICP selectively prefetches popular object versions from the broadcast channel in order to further improve transaction response time. A simulation study compares MICP with one offline and two online cache replacement and prefetching algorithms. Performance results for the workloads and system settings considered demonstrate that MICP improves transaction throughput rates by about 18.9% compared to the best performing online algorithm and it performs only 40.8% worse than an adapted version of the offline algorithm P.