Cognos Performance Issues

[Kellyc]

Hey All,

I am new to this forum so please bear with me! lol

Where do I start?! We used to have Applix and TM1 used to run absolutely fine but since upgrading to Cognos (Aug 2011) we have had nothing but problems. We are currently using version 9.5.2 on our Production (Live) server with excel 2007 and version 10.1.1 on our Development (Test) Server with Excel 2010. Both are run on a virtual enviroment and I am wondering if Cognos is fully supported on virtual as there seems to be conflicting information on this from IBM? We were also running with only 4GB memory at the beginning, which we managed to get up to 6GB. Now we have fought IT and got 12GB on both on our Live and Test Servers and the Live is running much better with fewer perfromance issues but our Test Server is still not right. We are concerned that it is down to upgrading to 10.1.1 and so are a little apprehensive to upgrade live to the same version if we are going to encounter the same issues.

Whilst doing our testing ahead of upgrading Live to 10.1.1 we have come across some problems. Firstly am I correct in saying that version 10.1.1 does not support excel 97-2003 files? As we still have some Budget files saved in this format.

Also I know it is advised not to pull and send at the same time but as the majority of our reports are in DBRW some with DBS formulas too, am I also correct in thinking that, although this was not a problem for Applix, it is causing major performance issues in Cognos to have both formulas "active" at the same time? As I read about TM1 getting confused as to which formula to process first? We have been experiencing slow processing and TM1 "locking up"

I have been looking at ways to prevent these issues and we are changing the DBRW formulas to DBR as we need to have both formulas "active" at the same time in the spreadsheets. We are also changing the format of some of our files to Excel 2007.

Other than this I'm not really sure what else we can do that is not IT related? I know when the test Server went down last week there was nothing on the Log file or any errors being reported so it is just really strange.

If anyone can answer any of the above or provide any other advice/info I would really appreciate it!

Many Thanks!

[David Usherwood]

Given you are mixing 9.5.2 and 10.1.1, I would check out this;

[Kellyc]

Thanks for your reply David.

I read that PI is turned off by default in version 9.5.2, is this the same for 10.1.1? In which case it might already be off. Do we just check the TM1 config file to see if it is on or not?

[declanr]

Thanks for your reply David.

I read that PI is turned off by default in version 9.5.2, is this the same for 10.1.1? In which case it might already be off. Do we just check the TM1 config file to see if it is on or not?

Kellyc,

That would be logical... but NO... IBM decided to confuse the hell out of all of us on that one.

In 10.1.1 (unlike 9.5.2) if not specified in the CFG file, TM1 will automatically enable PI

[Kellyc]

Thanks for your reply David.

I read that PI is turned off by default in version 9.5.2, is this the same for 10.1.1? In which case it might already be off. Do we just check the TM1 config file to see if it is on or not?

Kellyc,

That would be logical... but NO... IBM decided to confuse the hell out of all of us on that one.

In 10.1.1 (unlike 9.5.2) if not specified in the CFG file, TM1 will automatically enable PI

Thanks Declanr

If this will improve performance on 10.1.1, how do you turn it off in the file?

[declanr]

From memory I believe its just:

Code: Select all

ParallelInteraction=F

And then you will need to bounce the server for it to take effect.

I recommend looking in the guides though to see all available config parameters as you may find some other things of use and there is a chance memory has failed me on the correct statement, I know I have got it wrong at least once before (but that was partly because IBM released it incorrectly in a sample template that I copied.)

[Kellyc]

Hi Declan

I've just googled that and it is what you have put so I will see if we can get the config file amended.

Out of interest have you any experience with Cognos TM1 in a virtual environment? And what memory are you running TM1 with? I believe we are running 12GB and 2 cores.

Thanks!

[David Usherwood]

IF PI is the issue you should look at getting the right patch to make it work properly as it can be a great help with lots of writers. For memory usage, look at the memory footprint of your server via Windows Task Manager, also turn Performance Monitor on and see which cubes are taking the space and why (}StatsByCube). If you are blowing 'physical' memory then yes, TM1 will tank unless you have very very fast swap.

[declanr]

Out of interest have you any experience with Cognos TM1 in a virtual environment? And what memory are you running TM1 with? I believe we are running 12GB and 2 cores.

The majority of installations I have been involved with have been through virtual environments and I have worked with anything in memory terms from the tiddly (4GBish) up to some monstrously impressive setups, it all depends on the model and it would be hard to recommend any specific requirements to you without seeing the model. David's suggestion is a good idea.

One thing to be careful of with VMs is making sure that TM1 always has access to the allocated memory, on occasion when set up in a VM, the management system could say TM1 Should have 16GB but it's only now using 8GB so we can allocate the remaining 8GB to another slice on the same box... then when TM1 runs up and wants to get at that remaining 8GB... it's not there anymore! Your infrastructure guys should be able to set it so that you are guaranteed access to X amount of RAM.

[Kellyc]

I don't think we are exceeding memory now as TM1 seems to be using around 8GB so I don't think this is the issue but I don't have access to the Server to turn on performance Monitor or look at the Task manager. We have previously looked at the Cube optimizer to improve memory used, but can look back at this for our test environment.

We can look into PI or whether there is a suitable patch as our TM1 consultants are coming in to discuss performance next month, its just that I think even they are a little baffled by the issues we are experiencing with 10.1.1. on our Test Server.

We had issues on Live too (9.5.2) but this seems to be running much better since memory was upped to 12GB....

Wish things were more straightforward!!