Comparing shredded and native XML data management approaches in relational DBMSs

XML database functionality is becoming more mature over time, both in native XML database and relational database products. Exploration of new approaches for storing and managing XML information, as the major functionality of a database with XML-support, has become a serious and ongoing challenge to all database system vendors. Several benchmarks have been established to evaluate different aspects of XML database performance. However, there is still no benchmark to evaluate the performance of XML database functionality as compared to relational database functionality. This thesis will assess the current state of XML and XQuery support in commercial relational database systems. It will also present EXRT (Experimental XML Readiness Test), which is a new XML benchmark designed to methodically evaluate XML data management tradeoffs, such as the impact of query characteristics on the relative performance of shredded versus native XML. This benchmark has been implemented and comparisons have been made of two commercial relational database systems. Valuable results have been obtained which illustrate the relative performance of different approaches and methods of processing XML (even within the same system) as well as the importance of some of the factors that the performance depends on. The new benchmark, which differs from previous XML micro-benchmarks, tests basic insertion, deletion and update performance, in addition to read-only (query) performance.