hi on a HPC server a RM process takes only a limited fixed amount of CPU time instead of utilizing all available CPU as if RM hunger for resources was capped? Why is that? ty
that was not a metaphysical or rhetorical question.
Please dont be shy and shoot, answer, discuss.....
I think it is a weird RM behaviour and it is a problem.
ty
fritmore wrote:
hi on a HPC server a RM process takes only a limited fixed amount of CPU time instead of utilizing all available CPU as if RM hunger for resources was capped? Why is that? ty
until now, the parallelization of RapidMiner has only limited capacity. We plan a major rework for the next major release (i.e. RapidMiner 6). Until then, you gain more if you start more than one instance of RapidMiner to handle several problems in parallel.
this happens regardless whether any parallelization features are on or off. The CPU utilization is simply capped for any type of operator or parallel settings for RM.
Sure one can run multiple jobs but this is not a solution.
f
Marius wrote:
Hi,
until now, the parallelization of RapidMiner has only limited capacity. We plan a major rework for the next major release (i.e. RapidMiner 6). Until then, you gain more if you start more than one instance of RapidMiner to handle several problems in parallel.
Yes, you are completely right, and as I said above we are planning true parallelization of all RapidMiner processes for the next major release. Until then we all have to live with the limited parallelization support of the current version.
Contributor II
Answers
that was not a metaphysical or rhetorical question.
Please dont be shy and shoot, answer, discuss.....
I think it is a weird RM behaviour and it is a problem.
ty
until now, the parallelization of RapidMiner has only limited capacity. We plan a major rework for the next major release (i.e. RapidMiner 6). Until then, you gain more if you start more than one instance of RapidMiner to handle several problems in parallel.
Best,
Marius
this happens regardless whether any parallelization features are on or off.
The CPU utilization is simply capped for any type of operator or parallel settings for RM.
Sure one can run multiple jobs but this is not a solution.
f
Best,
Marius