X-WACoDa - Data Warehousing Design and Advanced Engineering Applications - page 41

Unfortunately, in this context, though XML
and XQuery are normalized, XML DBMSs and
data warehouse architectures are not. The XML
warehouse models and approaches proposed in
the literature share a lot of concepts (originating
from classical data warehousing), but they are
nonetheless all different. In this paper, we aim at
addressing this issue. We first quite exhaustively
present and discuss related work regarding XML
warehousing and OLAP. Then, we motivate and
recall our XML warehousing methodology, where
XML is used for integrating and warehousing
complex data for analysis. Our main contribu-
tion is a unified, reference XML data warehouse
architecture that synthesizes and enhances existing
models. We also present an XML warehousing
software platform that is architectured around
our reference model and illustrate its usage with
a case study. Finally, we conclude this paper and
provide future research directions.

descriptive dimensions and a set of usually numeric
and additive measures (such as sale amounts, for
instance). Dimensions contain textual descrip-
tors ( members ) of the studied business (Kimball,
2002). Star-like schema including hierarchies
in dimensions (e.g., town, region and country
granularity levels in a geographical dimension)
are termed snowflake schemas . Star-modeled data
warehouses with several fact tables are termed
constellation schemas .
Finally, On-Line Analytical Processing or
OLAP (Codd et al. , 1994) is an approach for ef-
ficiently processing decision-support, analytical
queries that are multidimensional by nature. Data
are stored in multidimensional arrays called data
cubes that are typically extracted from data ware-
houses. Data cubes are then manipulated with the
help of OLAP operators.

xML Data Warehousing

Several studies address the issue of designing and
building XML data warehouses. They propose
to use XML documents to manage or represent
facts and dimensions. The main objective of these
approaches is to enable a native storage of the
warehouse and its easy interrogation with XML
query languages.
Research in this area may be subdivided into
three families. The first family particularly focuses
on Web data integration for decision-support pur-
poses. However, actual XML warehouse models
are not very elaborate. The second family of XML
warehousing approaches is explicitly based on
classical warehouse logical models (star-like sche-
mas). They are used when dimensions are dynamic
and they allow the support of end-user analytical
tools. Finally, the third family we identify relates to
document warehousing. It is based on user-driven
approaches that are applied when an organization
has fixed warehousing requirements. Such require-
ments correspond to typical or predictable results
expected from an XML document warehouse or
frequent user-query patterns.

RELATED WORK

In this section, we first recall a couple of funda-
mental definitions before detailing literature work
related to XML data warehousing and OLAP.

Definitions

A data warehouse is a copy of transaction data
specifically structured for query and analysis
(Kimball, 2002). More formally, a data warehouse
is a subject-oriented, integrated, time-variant and
non-volatile collection of data in support of manage-
ment's decision making process (Inmon, 2005). A
single-subject data warehouse is typically referred
to as a datamart , while data warehouses are gener-
ally enterprise in scope (Reed et al. , 2007).
A star schema is the simplest data warehouse
schema. Shaped like a star, it consists of a single,
central fact table linked to peripheral dimensions .
A fact represents a business measurement (Kim-
ball, 2002) and is constituted of references to its