Best way to optimize feeders

[Emixam]

Hello, I inherit a model with a cube of +/- 400 rules & feeders and the feeder file is 13gb. I suspect a lot of overfeed as there is not a lot of data in the cube. The server takes forever to start when the feeders are reprocess (they are using persistent feeders).

I will review the feeders but I was curious to know what are the steps you are doing when it comes to review and optimize feeders ?

[Adam]

Just some helpful information for me:
- How long does it take to start db without persistent feeders?
- How much RAM is being used?
- How many numerical cells are populated?

I’ve always thought this article is neat, though I’ve never used the approach: https://code.cubewise.com/blog/troubles ... th-emojis/

[Elessar]

Hello,

You can also use }Statsbycube for this (don't forget to start Performance monitor): calculate the ratio of Stored and Populated Cells (both Numeric and String) to Number of Fed Cells for the largest cubes (by Total Memory Used).

50 times is suspicious, and more that 100 — definitely overfeed.

(taken from Yuri's blog: https://www.ykud.com/blog/cognos/tm1-co ... nce-today/)

[Emixam]

Someone called me out on another post for not replying.

I drank a lot of coffee and reviewed and optimized the feeders one by one. Mistakes were made on the 'Next Period' feeders and the feeders on consolidation.

There was a "Day" dimension (ex: 2024-01-01, 2024-01-02, ...) and the feeders was something like this:

Code: Select all

['Amount']=>DB(Cube_Name, ..., ...,ATTRS('DAY', !Day, 'Next Year'), !Measure);

So each element in the day dimension was feeding the entire 'Next Year'. I fixed it and it looks better now !

Thanks to everyone for your help!

[Steve Rowe]

One thing to make sure of when you are recasting feeders is to make sure that.
1. If Persistent Feeders is turned on then delete the .Feeder files, with the server down.
2. Restart the database, even if you aren't using Persistent Feeders.

Otherwise you aren't really testing your current feeder set.