Hi,

I have a fact table that has a number of measures in it, and I have
allocated the data type to be "integer". I know this size if
sufficient at the detail level (ie. product by month by geography).
However, when aggregated to higher levels of each of the three
dimensions, there is every possibility that the measures will be
greater than the 2.147 billion limit that an integer allows.

Does this matter? Is the data type allocated to a measure in a cube
purely as a reference to the source fact table, or does Analysis
Services use it for allocating space for aggregations and therefore
require it to be large enough for the aggregations at the upper levels
of dimensions?

Thanks,
Steve.

Yes it matters. The data type must be capable of containing the fully
aggregated value.

public @ the domain below
www.tomchester.net

"Steve Jenkins" <sjenkins (AT) au (DOT) imshealth.com> wrote

Thanks for this Tom. It seems that the measure will overflow and go into negatives, as I've just found out in a cube of mine. I'll upgrade them to bigint.

Cheers,
Steve.