A fine-grained replacement strategy for XML query cache

  • Authors:

  • Li Chen;Song Wang;Elizabeth Cash;Burke Ryder;Ian Hobbs;Elke A. Rundensteiner

  • Affiliations:

  • Worcester Polytechnic Institute, Worcester, MA;Worcester Polytechnic Institute, Worcester, MA;Worcester Polytechnic Institute, Worcester, MA;Worcester Polytechnic Institute, Worcester, MA;Worcester Polytechnic Institute, Worcester, MA;Worcester Polytechnic Institute, Worcester, MA

  • Venue:
  • Proceedings of the 4th international workshop on Web information and data management
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

Caching popular queries and reusing results of previously computed queries is one important query optimization technique, especially in modern distributed environments such as the WWW. Based on the recent proliferation of XML data and the emergence of the XQuery language, we are thus developing a query- based caching system for XQuery queries, called ACE-XQ. ACE-XQ applies innovative query containment and rewriting strategies to answer incoming user queries based on the cached XQueries, whenever possible, instead of accessing remote XML data sources.To manage the space of the cache, a straightforward application of traditional replacement strategies would correspond to removing a complete cached query and its derived XML document as a whole when space needs to be freed. This coarse granularity however does not match well with the typical access pattern of web searches where new queries often partially overlap with cached queries.In this paper, we propose a novel replacement strategy appropriate for such query-based XML caching systems. In particular, we collect user access statistics at the granularity of the XML path structure instead of the complete XML query regions. We then apply a more fine-grained replacement strategy that purges XML fragments off a cached region instead of the whole XML document and accordingly adjusts the query descriptor. This may better capture the user access patterns since more frequently used XML document fragments are likely to remain in the cache while other less beneficial parts are purged. This approach has been implemented in our ACE-XQ System. Preliminary experiment results illustrate the performance improvement achievable by our fine-grained replacement strategy over the one which replaces a whole XML view at a time when the cache size is relatively large.