If you are unable to create a new account, please email support@bspsoftware.com

 

News:

MetaManager - Administrative Tools for IBM Cognos
Pricing starting at $2,100
Download Now    Learn More

Main Menu

Optimizing EP Install

Started by kazanec, 26 Jun 2007 07:36:22 PM

Previous topic - Next topic

kazanec

I was wondering if someone can make some recommendations as to how to go about optimizing an EP installation. I had EP installed last week. We tested the installation by deploying the GO Tutorial example and everything seemed OK. Since then, I've begun developing my own model and have been running into some performance issues with Contributor. The cube isn't massive (110K cells), but things still move slowly. For instance, adding a value to cell and hitting enter results in having to wait upwards of 30 seconds before control is handed back. In the case that the modified cell is part of a calculation, it takes even longer. Is this normal? I'm not expecting Excel type speeds, but I was expecting it to be reasonably close.

My server is a 2x3.0 dual core with 4GB of RAM. It appears as though Planning only uses 1 of the 4 available cores. Is there anything I can do to have it take advantage of the other 3?

Thanks for your help.

ykud

Quote from: kazanec on 26 Jun 2007 07:36:22 PM
The cube isn't massive (110K cells), but things still move slowly. For instance, adding a value to cell and hitting enter results in having to wait upwards of 30 seconds before control is handed back.
Check out links and calculations -- you've got loops, it seems. It has nothing to do with server (Contributor has a thick client, using only one CPU and 1 GB of ram) -- it's all about your application model.

jluckhoff

Sounds like loops or conditionals.

Also make sure none of your links are tageting totals that really slows things down.

j

andrewbho

How much ram is available on the DESKTOP?  Contributor is PC based and not dependant on the server.  The other thing to look for is a lot of complex calcs, multiple breakbacks, and like the other said....circular links and conditionals.  Also check your virus protection, but that shouldn't be it because your model is small.  Test performance on the server and see if that makes a difference.

kazanec

The desktop has 2GB of RAM.

It appears as though my problem is that I have two dimensions nested together (Products -> Versions). As soon as I stop nesting the two, performance is good. This afternoon, I once again nested the two dimensions, but with a small set of products and the performance was good. I'm still a little confused as to why the original structure had such poor performance. My largest cube had a cell count of only 130K. I've got some people at Cognos looking into it, but even they seem baffled.

Does anyone have an experiencing with nesting and can pass along some tips?

ovo

If you have a large number of rows or columns on the grid that is being displayed the rendering can be very slow.  When you nest two dimension I suspect that the size of the display is growing significantly?