MRFsimpleFoam + Porosities in parallel

bastil · September 9, 2009, 09:15

Dear all,

I have massive trouble with the parallel scaling of MRFsimpleFoam on our cluster. The solver has been slightly modified (added porosities) and is from the 1.5-dev. It contains some ggis - this might be important for parallel performance. I compare with FLUENT runs from 1 to 32 cores on a myrinet cluster and so far I have (still running):

FLUENT:
2Core: 13,739s
4Cores: 5,594s
8Cores: 3,458s
16Cores: 1,609s
24Cores: 1,215s
32Cores: 950s
This looks quite reasonable for me. Same model on OpenFOAM:
8Cores: 12,752s
16Cores: 14,647s
32Cores: 20,444s
Very strange - slowdown instead of speedup and in general very slow.
Are there some specialities for ggi, MRF and porosities in parallel? What might be wrong? May it be worth trying a benchmark case? which? Any ideas?

Regards BastiL

bastil · September 9, 2009, 10:39

Ok after investigating little further this improves rapidly as soon as tuning of ggis (even keeping same decomposition). GGI-performence in parallel still seems to be an issue for large cases running on more than 8 cores.

Regards BastiL

September 9, 2009, 09:15	MRFsimpleFoam + Porosities in parallel	#1
bastil Senior Member BastiL Join Date: Mar 2009 Posts: 462 Rep Power: 9	Dear all, I have massive trouble with the parallel scaling of MRFsimpleFoam on our cluster. The solver has been slightly modified (added porosities) and is from the 1.5-dev. It contains some ggis - this might be important for parallel performance. I compare with FLUENT runs from 1 to 32 cores on a myrinet cluster and so far I have (still running): FLUENT: 2Core: 13,739s 4Cores: 5,594s 8Cores: 3,458s 16Cores: 1,609s 24Cores: 1,215s 32Cores: 950s This looks quite reasonable for me. Same model on OpenFOAM: 8Cores: 12,752s 16Cores: 14,647s 32Cores: 20,444s Very strange - slowdown instead of speedup and in general very slow. Are there some specialities for ggi, MRF and porosities in parallel? What might be wrong? May it be worth trying a benchmark case? which? Any ideas? Regards BastiL Last edited by bastil; September 9, 2009 at 09:39.

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
MRFSimpleFoam Tutorial	bastil	OpenFOAM Running, Solving & CFD	48	August 1, 2012 10:00
Script to Run Parallel Jobs in Rocks Cluster	asaha	OpenFOAM Running, Solving & CFD	12	July 4, 2012 22:51
HP MPI warning...Distributed parallel processing	Peter	CFX	10	May 14, 2011 06:17
MRFSimpleFoam with cyclicGgi in parallel	hannes	OpenFOAM Bugs	7	April 29, 2010 12:50
Parallel Computing Classes at San Diego Supercomputer Center Jan. 20-22	Amitava Majumdar	Main CFD Forum	0	January 5, 1999 12:00

September 9, 2009, 10:39		#2
bastil Senior Member BastiL Join Date: Mar 2009 Posts: 462 Rep Power: 9	Ok after investigating little further this improves rapidly as soon as tuning of ggis (even keeping same decomposition). GGI-performence in parallel still seems to be an issue for large cases running on more than 8 cores. Regards BastiL