Transformer

[EarthDog]

Hi!

I just got my 10.1 media yesterday and installed our development server. I uninstalled everything and reinstalled 10.1.

Everything is ok and it is much faster now BUT I have one problem rather serious:

-I tried to build a cube from Transformer that approx does 3 minutes to build in 8.4 Now it is around 10-15 minutes!

Does anyone has any idea about this? What could be the cause of the delay?

Elias

[MFGF]

Weird!  I was under the impression cube build times had been optimised for the C10 release.  I've not heard of them increasing before!  What options are set in the properties of the cube(s) in terms of number of passes etc?

Regards,

MF.

[EarthDog]

I am experiencing increase in speeds everywhere thats why i thought it was weird also.....

The settings are exactly the same as in the production cube..i just import the content and got the modules and tried the rebuild...

[cognostechie]

After upgrading 8.4 to 10, out testing sent e-mails to everybody saying that report performance improved significantly and everybody was a happy camper ! Guess what? They realised that during testing, there was no load on the server because the Live environment was still 8.4. Once all went live on 10, there was no difference in performance.

For the cube build, I too, expected to build faster becuause of 'in-memeory caching' in 10 but did not notice any difference. It didn't slow down though  compared to 8.4 !

[EarthDog]

brought back the snapshots and run the SAME cube with 8.4 nd 10.1 on the SAME computer and with the same DW on an SQL 2008.

Totals:

Timing, TOTAL TIME (CREATE CUBE),00:01:12 (8.4)

Timing, TOTAL TIME (CREATE CUBE),00:03:30 (10.1)

the diff i found:

10.1:
Πεμ 12 Μαϊ 2011 9:13:57 μμ 4 00000000 End processing 246786 records from data source 'Measures'.
Πεμ 12 Μαϊ 2011 9:13:57 μμ 4 00000000 Timing, READ DATA SOURCE,00:03:07

8.4:
Πεμ 12 Μαϊ 2011 10:23:28 μμ 4 00000000 End processing 246786 records from data source 'Measures'.
Πεμ 12 Μαϊ 2011 10:23:28 μμ 4 00000000 Timing, READ DATA SOURCE,00:00:58

now what?

In memeory caching is by default enabled or i can mess with it simewhere?

[cognostechie]

Try shutting down the 8.4 service, then re-boot the server, start only 10.1 service and build the cube.

[EarthDog]

Even if the servers are different? i mean 10.1 is installed on a different server than 8.4..

they have the same DW on an SQL though...

[cognostechie]

Yeah, shut down 8.4 because it's possible 8.4 processes are blocking those tables in the DW. I have had different versions
of Cognos running concurrently but 10.1 behaves strangely when working with 8.4. Atleast that's what we found.

By the way, how did you upgrade the Model? Did you upgrade it or running 8.4 model in 10.1 ? Did you re-compile the model using 10.1 ?

[EarthDog]

Did that (shutting down 8.4) no difference ...

As far as the models i didnt do anything..I mean i open the model in FM and republish the package but made no diff....

First i installed 10.1
then i got the content from 8.4 via all_content.zip file
and then o tried to create the cube..

Should i have done something else?

how do i "compile" the model?

[MFGF]

I think cognostechie is referring to the Transformer model - converting it from an mdl file to a pyj file.

MF.

[cognostechie]

Yes, that's what I meant.

By the way, the FM model does not really get upgraded  to 10.1 just be opening and saving the model and re-publishing the package. I mean it doesn't get the 10.1 functionality. This used to work in earlier upgrades but here is what hapenned in my case which leads me to believe that.

8.4 had an issue in FM which was later fixed in 8.4.1. I worked for a client who was using 8.4 and not 8.4.1.
The conformed dimension in FM (Two Facts joined thru a Dimension) was giving wrong results in the Report.
The determinants were set properly. When the FM Model was opened in 10.1 and saved and package re-published, it did not fix anything. I had to delete the Query subjects involved, save the model, add those Query subjects again and do the join again. After that, when I re-published the package, it fixed the data !!

Not sure if this has been fixed in the patch released for 10 recently.

[EarthDog]

Nice thing to know..So you are saying that probably i have to deltee and re import the queries in the "suspicious" models e?

Today i ll have over here a Cognos consultant and i will ask him about that..

Thanks.

[cognostechie]

No, you shouldn't have to delete things in thr Model. Only those things that were NOT supported in 8.4 due to a bug. Those things you may have to delete and create it again.

What I meant was that if a feature has a bug in the earlier version and if that bug has been fixed in the newer version, I would expect that feature to start working simply by upgrading the model (Open,save and publish again). It didn't work that way.