Exploiting Native XML Indexing Techniques for XML Retrieval in Relational Database Systems

In XML retrieval, two distinct approaches have been established and pursued without much cross-fertilization taking place so far. On the one hand, native XML databases tailored to the semistructured data model have received considerable attention, and a wealth of index structures, join algorithms, tree encodings and query rewriting techniques for XML have been proposed. On the other hand, the question how to make XML fit the relational data model has been studied in great detail, giving rise to a multitude of storage schemes for XML in relational database systems (RDBSs). In this paper we examine how native XML indexing techniques can boost the retrieval of XML stored in an RDBS. We present the Relational CADG (RCADG), an adaptation of several native indexing approaches to the relational model, and show how it supports the evaluation of a clean formal language of conjunctive XML queries. Unlike relational storage schemes for XML, the RCADG largely preserves the underlying tree structure of the data in the RDBS, thus addressing several open problems known from the literature. Experiments show that the RCADG accelerates retrieval by up to two or even three orders of magnitude compared to both native and relational approaches.