Index Selection for Efficient XML Path Expression Processing

One approach to building an efficient XML query processor is to use RDBMSs to store and query XML documents. XML queries contain a number of features that are either hard to translate into SQLs or for which the resulting SQL is complex and inefficient. Among them, path expressions pose a new challenge for efficient XML query processing in RDBMSs. Building index structures for path expressions is necessary. Meanwhile, indexes occupy much disk space. There is a tradeoff between the consumption of disk space and the efficiency of query evaluation. In this paper, we present a cost model for the space consumption of indexes and their benefit to XML queries. Making use of the statistics of XML data and the characteristics of the target application, we adopt greedy algorithm to select some map indexes to be built. Our experimental study demonstrates that query performance get comparatively significant improvement over the case without indexes while only consuming disk space of modest size.