My first Publication Agile-Data-Warehouse-Design-eBook | Page 41

20
Chapter 1
logical and physical, star schemas used by database administrators (DBAs) and
ETL/BI developers.
Data modeling
sessions (model-
storms) need to be
quick: hours rather
than days
Agile modelers must
balance JIT and
JEDUF modeling to
reduce design
rework
Evolutionary DW
development
benefits from ETL/BI
tools that support
automated testing
DW designers must
embrace change
and allow their
models to evolve
Agile dimensional
modeling techniques
exist for addressing
these requirements
To encourage collaboration and support iteration, agile data modeling needs to
be quick . If stakeholders are going to participate in multiple modeling sessions
they don’t want each one to take days or weeks. Agile modelers want speed too.
They don’t want to wear out their welcome with stakeholders. The best results
are obtained by modeling with groups of stakeholders who have the experience
and knowledge to define common business terms (conformed dimensions) and
prioritize requirements. It is hard enough to schedule long meetings with these
people individually let alone in groups. Agile data modeling techniques must
support modelstorming: impromptu stand up modeling that is quicker, simpler,
easier and more fun than traditional approaches.
Stakeholders don’t want to feel that a design is constantly iterating (fixing what
they have already paid for) when they want to be incrementing (adding func-
tionality). They want to see obvious progress and visible results. Agile modelers
need techniques that support JIT modeling of current data requirement in details
and JEDUF modeling of ‘the big picture’ to help anticipate future iterations and
reduce the amount of design rework.
Developers need to embrace database change . They are used to working with
(notionally) stable database designs, by-products of BDUF data modeling. It is
support staff who are more familiar with coding around the database changes
needed to match users’ real requirements. To respond efficiently to evolution-
ary data warehouse design, agile ETL and BI developers need tools that support
database impact analysis and automated testing.
Data warehouse designers also need to embrace data model change . They will
naturally want to limit the amount of disruptive database refactoring required
by evolutionary design, but they must avoid resorting to generic data model
patterns which reduce understandability and query performance, and can al-
ienate stakeholders. Agile data warehouse modelers need dimensional design
patterns that they can trust to represent tomorrow’s BI requirements tomorrow,
while they concentrate on today’s BI requirements now.
If agile dimensional modeling that is interactive, inclusive, quick, supports JIT and
JEDUF, and enables DW teams to embrace change seems like a tall order don’t worry;
while there are no silver bullets that will make everyone or everything agile over-
night, there are proven tools and techniques that can address the majority of these
agile modeling prerequisites.