Comparing Native XML Data Management Approaches in Relational and Native XML Databases.

摘要

As XML has become widely used, XML storage and query support has been added to a number of commercial database systems. This has taken two forms, relational shredded storage – when XML data is mapped to rows of tables - and native XML storage. Relational shredded storage has been supported for a number of years, while native XML storage is just starting to be prevalent in the commercial database area. For many use cases, native XML storage is now recommended by database vendors. In this thesis, a new benchmark called EXRT (Experimental XML Readiness Test) is used to examine the performance of XML data management features for native XML storage in three commercial database systems, including two relational database systems and one XML database system. Also, details of the implementation of EXRT, which is a simple micro-benchmark that methodically evaluates the impact of XML query characteristics on the tradeoffs between various XML storage and query processing schemes, are described. Preliminary results from running benchmarks using EXRT are presented and the lessons learned so far are summarized.