Our Story

Who We Are
BEM Partners Inc. was started with the sole purpose of assisting clients with their Cognos Business Intelligence and Cognos TM1 applications. With this being our primary focus we provided the most experienced resources in Southern California.

BEM Partners Inc.

info@bempartners.com

(619) 839-9250

Top
 

Improving Cognos TM1 Feeder Performance with MTFeeders Configuration

BEM Partners Inc.Cognos TM1 Improving Cognos TM1 Feeder Performance with MTFeeders Configuration

Improving Cognos TM1 Feeder Performance with MTFeeders Configuration

In IBM Planning Analytics (aka Cognos TM1) there is a new configuration setting called MTFeeders. This allows multiple cores to be used in the processing of feeders versus the single threaded processing in Cognos TM1 version 10. For larger applications this can be a game changer for your development/sustainment team that are making changes to your application.

 

Similar to Multi-Threaded Quarying (MTQ) where you specify the number of cores that a query can use you do the same thing with MTFeeders where you specify the max number of cores that can be used to process feeders when you save your rule file or run a CubeProcessFeeders function.

 

What should I be aware of with this setting?

There are a few things to be aware of with this setting:
1. IBM says that it will significantly consume more memory. From our testing of a few applications we have not seen any significant memory spikes.
2. This is the biggest consideration, you cannot have conditional feeders. Think about it, when you process feeders with a single core everything runs in order one after the other. If a feeder is dependent on a condition you have control in the order that they are processed. With multi-threaded feeding your feeders can jump around a bit with a dependency of a feeder being processed to late.

 

What type of performance should I expect?

From our testing we took a rule file that was taking 6 minutes to save with single threaded processing of feeders. We implemented a setting of MTFeeders=8, meaning 8 cores will be used for feeder processing, and our re-processing of feeders took 60 seconds for an 83% improvement. We recommend that you play around with this setting and your model as you eventually reach a threshold of diminishing returns. For example, if I had MTFeeders set to 12 and my rule save completed in 40 seconds, is that 20 second gain really worth locking up 4 additional cores, that is a question only you can answer about your model.

 

Let us know what you think about MTFeeders. We have helped a number of companies upgrade from TM1 10 to Planning Analytics and can assist you with your upgrade process.

Brandon
No Comments

Leave a Comment