Paging with connections: FIFO strikes again

Leah Epstein, Yanir Kleiman, Jiří Sgall, Rob van Stee

Research output: Contribution to journalArticlepeer-review

Abstract

We continue the study of the integrated document and connection caching problem. We focus on the case where the connection cache has a size of one and show that this problem is not equivalent to standard paging, even if there are only two locations for the pages, by giving the first lower bound that is strictly higher than k for this problem. We then give the first upper bound below the trivial value of 2 k for this problem. Our upper bound is k + 4 ℓ where ℓ is the number of locations where the requested pages in a phase come from. This algorithm groups pages by location. In each phase, it evicts all pages from one location before moving on to the next location. In contrast, we show that the lru algorithm is not better than 2 k-competitive. We also examine the resource augmented model and show that the plain fifo algorithm is optimal for the case h = 2 and all k ≥ 2, where h is the size of the offline document cache. We show that also in this case fifo is better than lru, although this is not true for standard paging.

Original languageEnglish
Pages (from-to)55-64
Number of pages10
JournalTheoretical Computer Science
Volume377
Issue number1-3
DOIs
StatePublished - 31 May 2007

Bibliographical note

Funding Information:
Jiˇí Sgall was partially supported by research project MSM0021620838 of MSˇ MT Cˇ R. Rob van Stee’s research was supported by the Alexander von Humboldt Foundation.

Keywords

  • Connection caching
  • Online algorithms
  • Paging

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science

Fingerprint

Dive into the research topics of 'Paging with connections: FIFO strikes again'. Together they form a unique fingerprint.

Cite this