XIST: An XML Index Selection Tool

XML indices are essential for efficiently processing XML queries which typically have predicates on both structures and values. Since the number of all possible structural and value indices is large even for a small XML document with a simple structure, XML DBMSs must carefully choose which indices to build. In this paper, we propose a tool, called XIST, that can be used by an XML DBMS as an index selection tool. XIST exploits XML structural information, data statistics, and query workload to select the most beneficial indices. XIST employs a technique that organizes paths that evaluate to the same result into structure equivalence groups and uses this concept to reduce the number of paths considered as candidates for indexing. XIST selects a set of candidate paths and evaluates the benefit of an index for each candidate path on the basis of performance gains for non-update queries and penalty for update queries. XIST also recognizes that an index on a path can influence the benefit of an index on another path and accounts for such index interactions. We present an experimental evaluation of XIST and current XML index selection techniques, and show that the indices selected by XIST result in greater overall improvements in query response times.