• DocumentCode
    1656600
  • Title

    An Improved Algorithm to Enhance the Utilization of Shortest Path Caches

  • Author

    Xiaohua Li ; Shimeng Wang ; Bin Wang ; Xiaochun Yang ; Ge Yu

  • Author_Institution
    Coll. of Inf. Sci. & Eng., Northeastern Univ., Shenyang, China
  • fYear
    2013
  • Firstpage
    419
  • Lastpage
    424
  • Abstract
    The caching has been developed for shortest path queries. At present, existing methods for the shortest path caching include the dynamic cache method Least-Recently-Used (LRU), the static cache method Highest-Query-Frequency (HQF), as well as a recently proposed method Shortest "Path" Cache (SPC). It is commonly known that LRU and HQF are not efficient enough for the shortest path caching. To the best of our knowledge, SPC is the state-of-art technique in the shortest path caching. However, in their method, they fully rely on the information from the historical log, making their method over fit the historical log. In this regard, this paper proposes an improved cache loading method, in which we try to cover as many paths as possible. At the same time, the cache utilization is maximized. Experiments on benchmark datasets demonstrate that the proposed approach has a better perfrmance than that of SPC, improve the cache utilization and decrease the query process time.
  • Keywords
    cache storage; network theory (graphs); query processing; HQF method; LRU; SPC; benchmark datasets; dynamic cache method; highest-query-frequency method; historical log; improved cache loading method; least-recently-used method; query process time; shortest path cache utilization; shortest path caching; shortest path queries; static cache method; Benchmark testing; Computational modeling; Loading; Roads; Servers; Time-frequency analysis; Training; cache utilization; over fitting avoiding; service level balance; shortest path caching; utilization measurement model;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Web Information System and Application Conference (WISA), 2013 10th
  • Conference_Location
    Yangzhou
  • Print_ISBN
    978-1-4799-3218-4
  • Type

    conf

  • DOI
    10.1109/WISA.2013.85
  • Filename
    6778676