The Extensible Markup Language (XML) has attracted much attention as a data model for data exchange, data integration and rich data representation. A challenging question is how to manage native XML data in distributed databases. This leads to the problem of how to obtain a suitable distribution design for XML documents.

Fragmentation, replication and allocation are database distribution design techniques that aim at improving the system performance.

Fragmentation can either be horizontal or vertical. Among the two fragmentation techniques, vertical fragmentation is often considered more complicated than horizontal fragmentation because of the huge number of alternatives. Vertical fragmentation has been studied in the relational data model and the object oriented data model. Existing vertical fragmentation approaches are mainly affinity-based.

In this seminar, I will first show that affinity-based vertical fragmentation approach cannot be adapted to XML because of their deficiencies. Then I will present a design approach for vertical fragmentation which is based on a cost model that takes the complex structure of queries on XML data into account. It will be shown that system performance can be improved after vertical fragmentation using our approach, which is based on user access patterns.