XML Access Modules: Towards Physical Data Independence in XML Databases

The advent of XML and the standardization of the XQuery language has
lead to the development of numerous XQuery processing engines. XQuery
processing, however, still poses significant performance challenges to
current engines, which often scale poorly with the database size.

This talk describes XAMs, a new approach for improving XQuery
processing performance based on XML materialized views. At the core of
this work is the XML Access Modules (or XAM for short) language,
describing a large set of storage structures, indices, and
materialized views for XML. XAMs are extended tree patterns, including
mandatory and optional nodes, structure and value predicates, node
identifiers, and special identifier properties which enlarge the space
of rewritings.
Given a set of materialized views described by XAMs, and an XQuery
query, our rewriting approach proceeds in three steps: algebraic
extraction of query tree patterns (which are also XAMs), rewriting of
the query XAMs based on the view XAMs, and finally re-assembling the
rewritings into logical algebra query plans. We have considered
rewriting under a special set of XML structural constraints,
encapsulated in a Dataguide. Together, the expressive power of XAMs,
and the Dataguide constraints, allow finding more (and more
interesting) rewritings than in previous works.

The XAM project is joint work with: Andrei Arion (INRIA), Veronique
Benzaken (LRI), and Yannis Papakonstantinou (UCSD).
More information on XAMs on can be found at:
http://gemo.futurs.inria.fr/projects/XAM/