Scalable distributed query and update service implementations for XML document elements

Materialization and indexing of XML views is an important issue, and a number of 'XML-to-RDBMS' mappings have been proposed. To build a scalable system for XML document storage where those views are up-to-date, our approach is to provide efficient implementations of basic services for query processing and updates. As a first contribution we define such services, and show that they support many 'XML-to-RDBMS' mappings. As a second contribution, we describe implementations of such services in a database cluster. We evaluate two implementations, one based on a two-level logging and isolation mechanism and one using Windows 2000 COM+ transactions. The main results are as follows: the first approach scales linearly for updates, and query response times are interactive. For example, with high workloads of 100 concurrent clients and 8 cluster components, query response times are 3 seconds on average. The alternative approach is weaker both regarding query response times (10 seconds on average for the above scenario) and scalability.