In the book online, it said max dimensions per cube 128. Is this the limit of
regular dimension, not include virtual dimension? Or is 128 the sum of
regular dimensions and virtual dimensions?

Thanks

I *think* the 128 is the sum of regular and virtual dimensions, but perhaps
someone from MS can confirm this?

Anyway, the point to make here is that no sensible cube design will ever
approach these limits (at least in AS2K), so they shouldn't be a problem.
Generally speaking, any more than 10-15 dimensions will thoroughly confuse
your users; additionally, very few client tools will be able to display more
dimensions than this very clearly, and you'll also find that aggregation
design becomes a bit of a nightmare.

Regards,

Chris

"Frank" wrote:

I can concur with Chris's point about the users not understanding it. I once
gave a a user group a cube of 20+ dimensions and it baffled them beyond
belief! In my experience users rarely ever use more than 3 dimensions (and I
don't mean just at any one time). Those dimensions are usually Time, Product
& Customer.

You can give them loads of wonderful demographic breakdown of their sales
but they generally never use it!!! What a waste!

Regards
Jamie Thomson


"Chris Webb" wrote:

I have 5 dimension tables with 300+ attributes. User needs to be able to
filter every attribute. I'm going to build a virtual dimension for each
filter. This is a special request, any better solution other than virtual
dimension?

Frank

"Jamie" wrote:

Frank,
Surely in a dimension with 300+ attributes there must be some natural
hierarchies emerging?

The simple answer is that AS can't do this (and neither, I'm sure, can any
other OLAP tool) and even if it could the time it would take to query it
would render it useless.

Without wanting to annoy you too much - you may be interested to know that
doing this in AS2005 is a snap. Each of these attributes could form their own
1-level hierarchy in a dimension, each of them browsable

Regards
Jamie Thomson

"Frank" wrote:

This reminds me of something I read about SAP BW (possibly in one of Nigel
Pendse's OLAP Surveys), that the reason so many projects with that tool
failed was because they went for the 'big bang' approach and tried to build
every single possible dimension - which I understand could mean having well
over a hundred dimensions. This led to cubes which were so slow and so
complex that no-one wanted to use them.

I suspect that in the above scenario, and possibly yours, the 'users' who
are requesting this are either not representative the people who would
actually be using the system (eg an unusually technical superuser, or a SAP
consultant), or haven't considered what the impact of their request would be
on the UI.

My advice would be to pick a few of the most commonly-used attributes, or
pick the subset of attributes used for one particular business process, and
build a smaller cube first. Show it to the users in their chosen front-end,
gain their trust and explain the issues. Then either add a few more
dimensions as and when necessary until the users are satisfied, or consider
building multiple cubes for different scenarios.

Regards,

Chris

"Jamie" wrote: