Yes, we are using MSAS SP3. How did you get around this
problem?

Thanks,
Sue

Here are the answers to your questions.

Are you using MSAS SP3? YES
What's the memory usage after Analysis Services initiallly
starts? 840 MB
What's the MemoeryConservationThreashold setting? 1.6 GB
How much physical memeory does the computer have? 4.0 GB


Thanks,
Sue

It seems you still have plenty of memory. Just to be sure, how did you get
the memory consumbed by Analysis Service?

What we did is try to process each dimension individually, then process each
cube individually. Another thing you could try is to restart Analysis
services after you finish processing all the dimensions. Our hang occurs
during dimension processing when we try to process all dimensions in one
transaction. I'm not sure when yours occur. Analysis Manager should indicate
where yours hang. If you're using DSO, try to report all the events to see
where your processing stops.

HTH,

Yubo

"Susan" <sdegrendele (AT) aol (DOT) com> wrote

Our's hangs when we attempt to commit all our changes from
performing "Process the database" If we process the
dimensions and then the cubes, it works fine. We used
pmon to determine the amount of memory consumed by the
Analysis Service.

Thanks for all your help,
Sue

What you are seeing is a common problem with "Process the database".

If you search in the Analysis Services Performance Guide
http://www.microsoft.com/technet/tre...e/AnSvcsPG.asp
under the topic 'shadow dimensions', you will see that just like when you
process a partition that 'shadow folders' are created for the new
aggregates, so does dimension processing. It creates shadow dimensions which
represent the in-memory images of what dimensions will look like once the
processing transaction completes. Then the old dimension in-memory images
are deleted.

When you do "Process the database", all of the dimensions are processed in
*ONE* transaction. If you have a large number of members in the dimensions,
this can easily take a LOT of memory to hold the shadow dimensions
(effectively double the amount taken afterwards), and then there is the
memory for the process buffers (although that is reclaimed after each
cube/partition). Then after that one transaction commits -- things go back
to normal.

I specifically mention this in the PG. Although it is just one line :-)
Sorry if I didn't emphasize it more :-)

If you run out of memory in that process, you may hang.
And this is why if you do it individually, it doesn't -- because each
individual dimension's shadow dimension is being reclaimed when it
individually commits.

Hope this helps.

--
Dave Wickert [MSFT]
dwickert (AT) online (DOT) microsoft.com
Program Manager
BI Practices Team
SQL BI Product Unit (Analysis Services)
--
This posting is provided "AS IS" with no warranties, and confers no rights.

"Susan" <sdegrendele (AT) aol (DOT) com> wrote