Dear all,

We recieve a number of datafiles each day, these files contain both inserts
and updates. We need to generate a cube every day which has both the inserts
and updates in it in the fastest time possile.

Is there any documentation on having to perform updates on measures that are
already in a cube, or will this not be a problem ?

If possible could you point me to a document that covers it ?

Thanks
P

there is 2 ways to perform an update:
reprocess the cube with the current data
"delete then insert" the values using incremental updates:
to do this:
first you have to "delete" data from the cube, this operation is done by an
incremental update of the cube where you send the negative values to update.
For example, if your cube has 10K$ of sales, you'll send a -10K$ to your
cube. The step "remove" the current values
After this you can insert the new values.
You can optimize. if your cube has 10K$ and your new value is 15K$ just send
the +5K$ information.

I don't recommend the incremental update due to the complexity and the
impact on the cube performance. (incremental updates have bad impact on a
cube, also the stability of the cube could be impacted)

I recommend to play with your partitions.
Create a "weekly" partition, every day you process this partition which
contain only data for the current week.
Create another partition which contains historical data (or current year...)
and process this one once a week.
Simple, quick, easy to maintain and in case of a problem you have to
reprocess only 1 short partition.
You can also provides real time cubes.

which version of AS do you use?


"Patricia" <Patricia (AT) discussions (DOT) microsoft.com> wrote

Thanks Jéjé,

I'm currently running Analysis Services 2000.

Thanks for help
P

"Jéjé" wrote: