PathGuide: an efficient clustering based indexing method for XML path expressions

Abstract

This paper focuses on the performance improvement for long-path XML query processing. It is motivated by the fact that the existing inverted index and join algorithms are efficient for short path XML queries, but are inefficient for long path XML queries since the response time of the existing approaches is exponential to the length of paths. We propose a clustering based indexing method, called PathGuide, in this paper, which enhances the XML inverted index with the clustering technique. The element nodes are clustered based on their path patterns and the summary for such path information is kept in a suffix tree as the index of these element nodes. In addition, new operations are proposed to fully utilize PathGuide. With the assistance of PathGuide, unlike the path expansion approach used in Lore, the set of a relative location path can be found via one-step index lookup. Compared to the existing structural join method, PathGuide significantly reduces both join overhead and disk I/O cost. The extensive experimental studies are conducted and our results show that PathGuide outperforms the structural joins at least four times in most cases.