TM1 Forum

Hello,

I have a large project with multiple cubes (~20). When I edit a rule in one of main cubes, it takes 15minutes to save a rule. What's the best way to find long loading feeders? I want to comment out these feeders on my test environment to make development faster.

What's the best way to find long loading feeders?

That would have to be switching them all off and then back on one at a time and measuring the resulting load times.

You can also look into persistent feeders to speed up your cube loading process.

Edit: Before anyone points out - yes, you need to unload the cube after switching off a feeder statement.

There is a direct correlation between how many cells are being fed and how long it takes for feeders to process (duh!) so look for 1) the elements that contain the most data and turn off whatever they feed and 2) look for instances where you are feeding consolidated nodes (as this feeds all the children. Basically, there is no shortcut for knowing exactly what your feeders are doing and turn them off when desired because just turning them off without understanding what they are doing is going to lead to incorrect results and likely cause development problems and delays.

Just a note that once a cell is fed, it is permanently fed - so you will need to comment out your feeders and then bounce the server before testing.

Christopher Kernahan wrote:Just a note that once a cell is fed, it is permanently fed - so you will need to comment out your feeders and then bounce the server before testing.

Nah, it's enough to unload the cube.

Agree with qml on unloading the cubes.
Might not even be necessary though to simply find the lines that need a long time to save/compile/process.

Just adding to what has been said already about what takes a long time when compiling rules.
Conditional Feeders (with complex conditions) and Feeders that use a large consolidation (without really overfeeding the target) on the left side can take a long time when saving the rule, as well.

The other big factor in feeder time is the amount of data in the cube.
You are probably better off developing with a full set of rules and a subset of the live data rather than subset of the rules and all the data?
Cheers,

Steve Rowe wrote:The other big factor in feeder time is the amount of data in the cube.
You are probably better off developing with a full set of rules and a subset of the live data rather than subset of the rules and all the data?
Cheers,

There are a lot rules and small amount of data (when I zip all Data folder during nightly backup, zip file weights 11MB). So, I chose to delete all rules on some cubes which are unrelated to my current cube i'm fixing.

Basically, I have to skim through all feeders and disable longest loading ones, there is no easy way. Thank you everyone for your input.