TM1 Forum

Hello,

Checking new parameters on tm1s.cfg on TM1 10.2.2 FP5 server, I found RulesOverwriteCellsOnLoad significantly effects TM1 server load time and memory consumption.
http://www.ibm.com/support/knowledgecen ... lang=en-us

Same model on current TM1 10.1.1 FP2 server load time is 02:13:00 and 117 GB memory consumption on startup.
Running same model on TM1 10.2.2 FP5 with RulesOverwriteCellsOnLoad=F result with 01:19:00 load time and 79 GB memory consumption on startup
Removing RulesOverwriteCellsOnLoad=F parameter on TM1 10.2.2 FP5 server showing same results as original TM1 10.1.1 server

I will appreciate any illumination and recommendations on this parameter.

Thank you,
Gabi

Hi Gabi,

The parameter was discussed a bit earlier in this thread: http://www.tm1forum.com/viewtopic.php?f=3&t=12414

The behaviour is really strange: with RulesOverwriteCellsOnLoad=F the server will store the inputed values "under rules", and so the server should consume more memory.
I think you should raise PMR with IBM.

Hello Elessar,

Thank you for your reply.
I'm re-checking this behavior as I did several changes on tm1s.cfg and might caught the wrong parameter.
I'll update with new results tomorrow

Thank you,
Gabi

Be aware that there is an open issue with IBM where RulesOverwriteCellsOnLoad=F can cause feeders to not fire properly.

You may find that your faster startup and smaller memory footprint is because your model is not completely calculated.

Hello again

Andy, Thanks for your reply - I'll verify RulesOverwriteCellsOnLoad impact before setting the prod env..

It seems I mistakenly caught RulesOverwriteCellsOnLoad as the cause for shorten load time and memory use.

Now, I verified 'SkipLoadingAliases' is the cause for shorten load time and memory consumption on startup.
Do you have any recommendations using this parameter?

Thank you,
Gabi

I think you should create a new topic with the subject "TM1 cfg parameter - SkipLoadingAliases". Our descendatns will appreciate this

IBM site tells us that "Please contact customer support to determine if this parameter is applicable to your Cognos® TM1® system" (This parameter you don't want, my young padawan)
I think it's better to delete the parameter from tm1s.cfg, if you do not know what it does.

Andy Key wrote:Be aware that there is an open issue with IBM where RulesOverwriteCellsOnLoad=F can cause feeders to not fire properly.

You may find that your faster startup and smaller memory footprint is because your model is not completely calculated.

Do you refer to the defect logged as APAR PI56768 which has been addressed by the IF1 for TM1 10.2.2 FP5 ?

https://www-304.ibm.com/support/entdocv ... wg1PI56768
PI56768: FEEDERS APPEAR NOT TO BE FEEDING THE CUBE IF RULESOVERWRITECELLS ONLOAD=F, PERSISTENTFEEDERS=F AND SUPPRESS ZEROS IT TURNED ON.

That looks to be a similar issue, but we found it in 10.2.2 FP4 rather than 10.1 as that link shows, and the setting of PersistentFeeders didn't make any difference for us. I don't have the precise APAR number to hand, so it is possible that it is just a mis-identification on that ticket. If it is correct in saying that it was 10.1, then it must be a regression bug (surely not!) as we were using the same code base and the same config file in base 10.1.1 without any problems.

IBM have assured us that it is fixed in 10.2.2 FP5 IF1 despite it not being mentioned in IBM's List of Fixes. We're too deep into getting an upgrade to 10.2.2.4.15 settled in to have a chance to install any new versions, and we'll probably be waiting for a few more critical fixes to be addressed as well.

I believe the PMR Andy is talking about is
PMR 70882,101,616 - Feeders do not fire up on startup in the TM1 Instance