Memory issues in full annulus simulation

evan247 · March 14, 2016, 15:20

Hi all,

I'm currently trying to do a steady, single stage, full annulus simulation using frozen rotor model for interface. (due to the existence of non-axisymmetric parts of the other geometries) My (single passage) blade rows are generated in Autogrid 5 and then loaded in CFX-pre in .cgns format. I then used the 'turbo rotation' option to copy each row to full annulus. Total mesh count is around 26m which is considered coarse for my case.

When run on a local linux workstation (16 cores, 32GB RAM), I've been given memory deficiency warnings after trying serial, local MPI, increasing the memory allocation factors, using large problem partitioner option, etc, following the exisiting posts in this forum.

I suspect that it's the GGI that causes massive memory usage but I don't see why it should be as my interfaces are fairly simple. Also I feel sth is wrong to use more than 32G RAM for a 20m ish calculation.

ghorrocks · March 14, 2016, 20:58

Are you using double precision? That uses extra memory, so if you don't need it turn it off.

The memory useage you report is about what I would expect for a simulation of that number of nodes. So I think this model is just too big for the RAM you have available. I don't think the GGIs are contributing significantly to the problem, just the node count.

alirezame · March 16, 2016, 08:47

Did you try to increase the '' Memory Alloc Factor'' by x2 or more?
in the partitioner part option of the CFX-Solver Manager.

evan247 · March 16, 2016, 09:30

Quote:

Originally Posted by alirezame

Did you try to increase the '' Memory Alloc Factor'' by x2 or more?
in the partitioner part option of the CFX-Solver Manager.

I did, but my experience is that the actual allocated memory doesn't change much. I later on reduced my mesh from 26m to 21m and was just able to run it (with 4 or 8 'cores' as per CFX) using 95% of my 32GB RAM.

-Maxim- · March 16, 2016, 10:34

I think your model is too big for just 32GB RAM. Can you update to 64GB?

evan247 · March 16, 2016, 10:40

Quote:

Originally Posted by -Maxim-

I think your model is too big for just 32GB RAM. Can you update to 64GB?

Yes that's the solution I'm taking now: all the way to 128G now!

March 14, 2016, 15:20	Memory issues in full annulus simulation	#1
evan247 New Member Join Date: Jan 2015 Posts: 29 Rep Power: 11	Hi all, I'm currently trying to do a steady, single stage, full annulus simulation using frozen rotor model for interface. (due to the existence of non-axisymmetric parts of the other geometries) My (single passage) blade rows are generated in Autogrid 5 and then loaded in CFX-pre in .cgns format. I then used the 'turbo rotation' option to copy each row to full annulus. Total mesh count is around 26m which is considered coarse for my case. When run on a local linux workstation (16 cores, 32GB RAM), I've been given memory deficiency warnings after trying serial, local MPI, increasing the memory allocation factors, using large problem partitioner option, etc, following the exisiting posts in this forum. I suspect that it's the GGI that causes massive memory usage but I don't see why it should be as my interfaces are fairly simple. Also I feel sth is wrong to use more than 32G RAM for a 20m ish calculation.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
restarting paused transient simulation using reactingFoam	JMDag2004	OpenFOAM Running, Solving & CFD	1	August 10, 2015 10:15
enGrid to OpenFOAM full export issues	coanda	enGrid	1	May 4, 2013 09:31
ICEM Meshing - Memory issues.	DamjanStefkov	ANSYS Meshing & Geometry	0	May 25, 2011 01:38
Memory Issues with AMP	subrew	ANSYS Meshing & Geometry	0	August 25, 2010 15:50
boundary layer simulation : basic issues	amol palekar	Main CFD Forum	0	July 27, 2004 18:22

March 14, 2016, 20:58		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	Are you using double precision? That uses extra memory, so if you don't need it turn it off. The memory useage you report is about what I would expect for a simulation of that number of nodes. So I think this model is just too big for the RAM you have available. I don't think the GGIs are contributing significantly to the problem, just the node count.

March 16, 2016, 08:47		#3
alirezame Member Join Date: Nov 2013 Posts: 57 Rep Power: 12	Did you try to increase the '' Memory Alloc Factor'' by x2 or more? in the partitioner part option of the CFX-Solver Manager.

March 16, 2016, 10:34		#5
-Maxim- Senior Member Maxim Join Date: Aug 2015 Location: Germany Posts: 415 Rep Power: 12	I think your model is too big for just 32GB RAM. Can you update to 64GB?