TM1 Forum

I was asking a question on http://www.tm1forum.com/viewtopic.php?f=3&t=10042 on how to use multiple processors and was given advice by David to start tuning with ViewConstruct. To follow one topic per forum thread, I opened a new thread.

1. I created very simple report in Cognos BI (one column and one row of data) in crosstab report.
2. In Cognos TM1 Architect I created view with the same data as in Cognos BI report and in Turbointegrator created ViewConstruct function.
3. Then I have restarted both server (Cognos BI and Cognos TM1) to make sure all of the caches are empied.
4. Started both servers.
5. I did a controled test. No turbointegrator process executed, so default server start-up without anything special.
6. Executed a report from Cognos BI. Results is 30 seconds.
7. Second test. Restarted both servers to empty cache. Executed turbointegrator process with ViewConstruct.
8. The same as step 6 and time is also 30 seconds.

In both cases I have started tm1top to see what is going on and as I see in both cases the same commands are executed and it takes the same time.

Do you have any step-by-step tips, how to make sure that report executed from Cogons BI is using Cognos TM1 precalculated views?

Which version of BI are you using? Are you using Dynamic Query Mode or Compatible Query Mode?

We've confirmed in the past that BI does leverage the TM1 calculation cache.
You may not have the ViewConstruct settings quite right. I suggest you export your ViewConstruct view to a file to see whether it includes the content in the BI report. Check the timings for the export, then run it again to confirm the TM1 cache is speeding up the export.

As has been explained quite a few times on this forum you also need to make sure that the VMM and VMT parameters are set to sensible values that will allow the view to be cached.

David Usherwood wrote:We've confirmed in the past that BI does leverage the TM1 calculation cache.

I agree it does, yes. However, correct me if I'm wrong, but I believe this is only the case if the query goes through DQM. In the case of CQM (or, more broadly the TM1 MDX query interface in general) the cache is not used.

Other points about making sure the stargate view is actually created and contains the relevant data are of course valid. I would try and confirm in pure TM1 first that the view is being created and used before including the BI layer.