Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.
Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Array causes "out of memory" and other peculiar shutdowns
Posted Jun 29, 2012, 12:27 p.m. EDT Fluid & Heat, Geometry, Parameters, Variables, & Functions Version 4.2a 2 Replies
Please login with a confirmed email address before reporting spam
Hello,
I have a very simple model, single phase laminar flow through round aperture, that makes COMSOL 4.2a (.166) crash. (I know that 4.3 is out, but our licence server has not been updated yet.)
The system is: 2x AMD Opteron 6274, 64GB RAM, 64-bit Debian 6 (with nothing else than SSH, COMSOL+COMSOL requirements), Soft-rendering is active
The problem is as follows:
Two chambers are connected, one with large square inlet with varied length and the other with small array of round apertures (R=3mm). I have been able to get meaningful result using 2D modelling for this system, however in 3D I have not been able to get further than geometry definition.
Whenever I try to make the array of apertures COMSOL crashes in one way or the other. The array consists of 206x228 apertures with R=3mm, spaced 8mm apart. To begin with I tried making a rectangular array, this did not go well - COMSOL kept going for more than 24 hrs before I shut it down.
I then tried making a vertical (linear) array of 206 apertures, it worked. Using each point on the vertical array as input I made horizontal arrays with 228 apertures, this worked until I reached 180 arrays at which point COMSOL gave "Out of Memory" error. I used an earlier saved file with only 172 arrays on my other machine (16 Xeon cores and 48GB RAM), and it worked up until 182 arrays before crashing and giving the "JAVA SIGSEGV" error.
Unfortunately the system as a whole is not symmetrical, otherwise i would have used a half sized array. But I did a simulation with a smaller section of the system, it was successful.
No matter what I try I cannot make more than 182 arrays. All kinds of errors come up when I try, the "JAVA SIGSEGV", "Out of Memory", "Java heap", and "libiomp5.so" error and sometimes no errors just shutdown. In the COMSOL window RAM usage indicator does at no point exceed 10GB, nor does htop indicator.
I would like some help with finding out why the problem occurs. If anybody has a machine with more RAM than mine and is able to try, then I would like for you to try making the full sized array.
When I'm able to, I will make available the file with 172 arrays - currently all COMSOL able machines are busy with simulations.
Another thing, loading the saved files takes very long - approximately 6 hrs. for the file with 172 arrays. Is this normal?
regards,
Mohtadin
I have a very simple model, single phase laminar flow through round aperture, that makes COMSOL 4.2a (.166) crash. (I know that 4.3 is out, but our licence server has not been updated yet.)
The system is: 2x AMD Opteron 6274, 64GB RAM, 64-bit Debian 6 (with nothing else than SSH, COMSOL+COMSOL requirements), Soft-rendering is active
The problem is as follows:
Two chambers are connected, one with large square inlet with varied length and the other with small array of round apertures (R=3mm). I have been able to get meaningful result using 2D modelling for this system, however in 3D I have not been able to get further than geometry definition.
Whenever I try to make the array of apertures COMSOL crashes in one way or the other. The array consists of 206x228 apertures with R=3mm, spaced 8mm apart. To begin with I tried making a rectangular array, this did not go well - COMSOL kept going for more than 24 hrs before I shut it down.
I then tried making a vertical (linear) array of 206 apertures, it worked. Using each point on the vertical array as input I made horizontal arrays with 228 apertures, this worked until I reached 180 arrays at which point COMSOL gave "Out of Memory" error. I used an earlier saved file with only 172 arrays on my other machine (16 Xeon cores and 48GB RAM), and it worked up until 182 arrays before crashing and giving the "JAVA SIGSEGV" error.
Unfortunately the system as a whole is not symmetrical, otherwise i would have used a half sized array. But I did a simulation with a smaller section of the system, it was successful.
No matter what I try I cannot make more than 182 arrays. All kinds of errors come up when I try, the "JAVA SIGSEGV", "Out of Memory", "Java heap", and "libiomp5.so" error and sometimes no errors just shutdown. In the COMSOL window RAM usage indicator does at no point exceed 10GB, nor does htop indicator.
I would like some help with finding out why the problem occurs. If anybody has a machine with more RAM than mine and is able to try, then I would like for you to try making the full sized array.
When I'm able to, I will make available the file with 172 arrays - currently all COMSOL able machines are busy with simulations.
Another thing, loading the saved files takes very long - approximately 6 hrs. for the file with 172 arrays. Is this normal?
regards,
Mohtadin
2 Replies Last Post Jun 29, 2012, 5:45 p.m. EDT
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
indeed sometimes one have to accept that "brute force 3D models" become to complex to be easily solved.
But have you tried to reduce your model, by studying 1, 7 or 9 holes in an array to get some functional relationships on p and U and sensitivity to variation such that you might perhaps replace the full wall of 2nnx2nn holes by a few fonctions on the boundary related to the local p and U (and wahever else you need) ?
--
Good luck
Ivar
indeed sometimes one have to accept that "brute force 3D models" become to complex to be easily solved.
But have you tried to reduce your model, by studying 1, 7 or 9 holes in an array to get some functional relationships on p and U and sensitivity to variation such that you might perhaps replace the full wall of 2nnx2nn holes by a few fonctions on the boundary related to the local p and U (and wahever else you need) ?
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
I'm currently trying to minimize the array as much as possible while trying to maintain the physical significance. So far I'm down to a "unit cell" of 16x16 apertures, I will have to do simulations to find out whether or not this configuration is usable and if it makes sense; from the geometrical point I'm able to cover 4/5 of the system using this unit cell, the rest is too asymmetrical and needs to be done separately.
I'll report my findings, when I get that far.
Thanks for your quick response.
I'll report my findings, when I get that far.
Thanks for your quick response.