Re: On what OLAP can and what OLAP can't

"Cimode" <cimode_at_hotmail.com> wrote in message news:1157702527.539724.32840_at_m73g2000cwd.googlegroups.com...
>
> David Cressey wrote:
> > It's clear from the above transcript that you and the data architect
were
> > talking past each other rather than holding a genuine dialogue. Looking
> > back, I can see myself in years gone by being on either side of the
> > discussion, at different points in my career.
> For a dialogue to happen, there's a need for a common basis. While I
> did set up OLAP cubes, I thought it obvious that the *Architect* did
> not measure up the consequences of loosing data integrity at definition
> level. As all idiots who try compensate such loss of integrity by
> costly and unreliable filtering processes at applicative level. OLAP
> layer is not exception.
>
>
> > I can, if you like, shed some light on what the data architect might
have
> > been trying to show you. However, I can only surmise that the
architect's
> > opinions and knowledge were somewhat like my own. They may have been
quite
> > different. What you might get out of such a discussion in here depends
on
> > what you are seeking.
> He later acknowledged that he had to assume SOME level of correcteness
> on the raw data extracted from the source databases to be able to go
> any further. On the other side, he admitted that most data sources
> returned incorrect results because not normalized. In a word this is
> a contradictory attitude. Which basically makes OLAP layer nothing
> but a semi presentation layer on the perspective of the database. The
> concept of OLAP could be perceived as a concentrator/connector of
> flexible targeted querying capabilities nothing more nothing less.
>
> > Since the DA isn't here, I can't rephrase what you were trying to
> > communicate to the DA for that person. However, it's just vaguely
possible
> > that the DA had been through the same learning curve you have, and had
> > grown to appreciate multidimensional modeling in addition to relational
> > model, rather than instead of it.
> Quite frankly I don't believe there's actually such thing as a
> multimensional model. OLAP technologies which I have to maintain on a
> dayly basis are nothing but a hierarchical hangover.
>
> > If you are seeking to genuinely understand how multidimensional modeling
and
> > star schema design might be genuinely useful additions to your own set
of
> > tools and methods, the discussion could go one way. If, on the other
hand,
> > you are merely looking for validation of your own opinions, and want
> > justification for disregarding anything novel (to you) that the data
> > architect might have said, then the discussion is likely to go nowhere,
as
> > many prior discussions in c.d.t. have.
> I perfectly know what a star schema is for having built several data
> warehouses using Analysis Services. Too much ado about nothing.
>
> > Before I invest the time and effort at trying to summarize and explain
the
> > benefits of star schema design, I want to have some inkling about
whether
> > it's a waste of time.
> Thanks for the intention. The point I was trying to make was not that
> I did not understand what OLAP is. It was sharing a discussion I
> supposing not to be uncommon.

Thanks for clearing that up. I guess I missed the point you were trying to make by sharing the discussion.
What was your point? Received on Fri Sep 08 2006 - 15:27:11 CEST