EV Crashing TM1 9.5.2 When Using Non-Symmetrical Forumlas

[jim wood]

Guys,

This ones for reference. Basically we found when using EV 9.5 and TM1 9.5.2 that if you use non-symmetrical formulas in EV (formulas that span more than one dimension) the TM1 service crashed. Using Event viewer we found the service went down as non-paged memory had run out.

The customer I'm working for contacted IBM. It seems it was being caused by having the following setting in the CFG file: LockPagesInMemory=T. According to IBM some users are setting this as they feel they will get some performance benefit. This is only the case if you run out of physical memory. Setting LockPagesInMemory=F stopped EV crashing,

Jim.

[Andy Key]

According to IBM some users are setting this as they feel they will get some performance benefit. This is only the case if you run out of physical memory.

According to IBM, IBM are wrong - or it's another example of the left hand not knowing what the right hand is doing.
From the 10.1 infocentre: http://publib.boulder.ibm.com/...lockpa ... y_tm1.html
(or http://www-01.ibm.com/support/docview.w ... wg27024038 for a longer description.)

The important part of the link being:

If a Cognos TM1 server running on a Windows 64-bit operating system is idle for a long period of time, physical memory taken up by the Cognos TM1 server will page out to disk.

(my emphasis).
This has nothing to do with running out of physical memory and your resultant virtual memory paging to disk, this is Windows 'knowing best' and paging physical memory.

[jim wood]

I did wonder about that, but to be fair we are using 9.5.2 (As stated) rather than 10.1 and they may (but I doubt it) behave differently.

[Andy Key]

The second link shows the affected software as 9.5.2 and 10.1.

The paging that the parameter is trying to prevent is a 'feature' of 64-Bit Windows, so not really anything to do with the version of TM1.

[jim wood]

Thanks Andy.

[jim wood]

The group I have been working with on this EV includes a couple of IBM guys. I forwarded the link on to them and this was their response:

Please note that the parameter is useful only at times where the operating system pages memory. Also, that the parameter did not function correctly, up until 9.5.2 FP1. You should move forward without this parameter, and if there is a problem with the operating system paging memory to disk, it should first be investigated by the system administrators as memory paging degrades the WHOLE server and not just TM1.

Only when paging can not be mitigated, should the parameter be set to true.