Page 1 of 1

Loading server with big-sized Cube and MaximumCubeLoadThreads

Posted: Tue Feb 20, 2018 1:35 pm
by EP_explorer
The model contains many objects and one big-sized Cube among them . I didn't use MaximumCubeLoadThreads parameter in tm1s.cfg and decide to try. Speed of Tm1 server loading became a bit faster but it didn't connect with loading of the big-sized Cube.

The big-sized Cube seems to load on one core and time for loading of the Cube (which is more essential for loading the whole model) is the same as before (I checked time in tm1server.log)

So - is it usual behavior for TM1 server (loading one object - i.e. cube on one core) and may be you know ways to load big-sized Cubes on several cores?

Re: Loading server with big-sized Cube and MaximumCubeLoadThreads

Posted: Tue Feb 20, 2018 4:29 pm
by mattgoff
I don't believe MaximumCubeLoadThreads does much, if anything, to improve stored data loads from disk. Its performance advantage is that it multithreads rule computation and feeder firing after stored values are loaded.

Matt

Re: Loading server with big-sized Cube and MaximumCubeLoadThreads

Posted: Tue Feb 27, 2018 5:07 pm
by ndivine
It seems IBM is still working through multi-threaded loading. As of PAL 2.0.3 multi-threaded loading has been disabled.
https://community.watsonanalytics.com/d ... reads.html

Also, at some point MaximumCubeLoadThreads was replaced by MTCubeLoad.

For large cubes, I believe MTCubeLoad did substantially speed up the data load portion.