Is it possible to use a calculated measure to avoid one distinct count
per cube limitation ?
Any example of MDX query would be very helpful.

Thanx,

Radek

Or... you could combine several cubes with distinct counts into a virtual
cube

Be aware however that distinct counts are real performance hogs if you
cannot aggregate you cube to 100%

Yours sincerely
Thomas Kejser

<suspi (AT) lycos (DOT) com> wrote

Are distinct count cubes slow to query or slow to process/build?

Are there any alternates to a distinct count? I guess i could set a 1/0 flag and sum it up, but setting the 1 and 0 to simulate a distinct count scenario will be somewhat complex.

thanks in advance

--
Message posted via http://www.sqlmonster.com

Hi Tarana

Distinct counts are generally very slow til process because of their nature.
If you do not aggregate to 100% they can also be very slow to query,
depending on the amount of indexing you have done on you relational database
fact table (since every aggregate that does not exist must be explicitly
queried)

If you can get away with a 1/0 flag (which you sometimes can) you should
definately try that approach - especially so if you cube is large. However,
be aware that distinct count in a multi dimensional world can only rarely be
simulated in the relational world with 1/0 flags. The problem is - that the
count will normally not be distinct across all dimensions - however, you may
be able to live with this in your scenario.

Yours sincerely
Thomas Kejser





"tarana halsall via SQLMonster.com" <forum (AT) SQLMonster (DOT) com> wrote