Menu

Discussion Forum

New Discussion
Filter

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.

Why the same program sometime need 4000 secs and sometimes 300 secs

Posted Jan 4, 2017, 2:01 p.m. EST RF & Microwave Engineering, Studies & Solvers Version 5.2 4 Replies

lz Chen
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

May I ask if anyone know why the same comsol program sometime need 4000 secs and sometimes 300 secs to solve? Exactly the same one, I didn't change any parameters, mesh, solver, just run it
4 Replies Last Post Jan 9, 2017, 10:56 a.m. EST
Jeff Hiller COMSOL Employee

Please login with a confirmed email address before reporting spam

Posted: 8 years ago Jan 4, 2017, 3:32 p.m. EST
Hello LZ,
One possibility is that your model requires around the amount of RAM available on your machine to run in core. If some other programs are sometimes running on the same machine, it could force the solution to be done out of core, which is much slower.
That's only a guess.
Jeff
Hello LZ, One possibility is that your model requires around the amount of RAM available on your machine to run in core. If some other programs are sometimes running on the same machine, it could force the solution to be done out of core, which is much slower. That's only a guess. Jeff
lz Chen
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 8 years ago Jan 4, 2017, 4:35 p.m. EST
Dear Jeff,

Thank you very much for your quick and helpful reply. Your suggestion is very reasonable. I am now running the program on a server of my university, the RAM is 384G. My program takes like 24G. It is possible that many people are using it right now. The CPU occupation is now 100%, RAM occupation is 70%.

May I ask if there is any method to improve this situation, the mesh level is the lowest one to guarantee precision. It is in 3D RF module, using COMSOL with Matlab Liverlink control. Should I do any process like clear the cache when running the program? Thank you again for your kind help and suggestion.

Best,
LZ


Hello Iz,
One possibility is that your model requires around the amount of RAM available on your machine to run in core. If some other programs are sometimes running on the same machine, it could force the solution to be done out of core, which is much slower.
That's only a guess.
Jeff


Dear Jeff, Thank you very much for your quick and helpful reply. Your suggestion is very reasonable. I am now running the program on a server of my university, the RAM is 384G. My program takes like 24G. It is possible that many people are using it right now. The CPU occupation is now 100%, RAM occupation is 70%. May I ask if there is any method to improve this situation, the mesh level is the lowest one to guarantee precision. It is in 3D RF module, using COMSOL with Matlab Liverlink control. Should I do any process like clear the cache when running the program? Thank you again for your kind help and suggestion. Best, LZ [QUOTE] Hello Iz, One possibility is that your model requires around the amount of RAM available on your machine to run in core. If some other programs are sometimes running on the same machine, it could force the solution to be done out of core, which is much slower. That's only a guess. Jeff [/QUOTE]
Jeff Hiller COMSOL Employee

Please login with a confirmed email address before reporting spam

Posted: 8 years ago Jan 5, 2017, 8:33 a.m. EST
If the problem is indeed that other users of the machine are hogging its resources, you may want to consider running your models in batch at night when the machine is presumably in lower demand. See this blog as an introduction to the topic: www.comsol.com/blogs/how-to-run-simulations-in-batch-mode-from-the-command-line/ .

Also, make sure you are using the latest version of the software as the solvers are continually improved to use less memory. Talking about that, different solvers can use very different amounts of memory to solve the same problem, so you can experiment with that as well. See this blog post for an example where switching solvers decreases peak memory usage by as much as a factor 8 in a benchmark: www.comsol.com/blogs/using-the-domain-decomposition-solver-for-thermoviscous-acoustics/ .

Short of that, if your model is already meshed as sparingly as can be, you may need to consider more fundamental ways of making your model leaner. Here I am thinking of more fundamental changes such as using symmetries, ridding the geometry of details that do not significantly affect the solution, replacing thin solids with surface-level formulations, etc. In other words, review the underlying mathematical model that you're asking the software to solve. What such simplifications you can make is very much problem dependent.

Best,
Jeff
If the problem is indeed that other users of the machine are hogging its resources, you may want to consider running your models in batch at night when the machine is presumably in lower demand. See this blog as an introduction to the topic: https://www.comsol.com/blogs/how-to-run-simulations-in-batch-mode-from-the-command-line/ . Also, make sure you are using the latest version of the software as the solvers are continually improved to use less memory. Talking about that, different solvers can use very different amounts of memory to solve the same problem, so you can experiment with that as well. See this blog post for an example where switching solvers decreases peak memory usage by as much as a factor 8 in a benchmark: https://www.comsol.com/blogs/using-the-domain-decomposition-solver-for-thermoviscous-acoustics/ . Short of that, if your model is already meshed as sparingly as can be, you may need to consider more fundamental ways of making your model leaner. Here I am thinking of more fundamental changes such as using symmetries, ridding the geometry of details that do not significantly affect the solution, replacing thin solids with surface-level formulations, etc. In other words, review the underlying mathematical model that you're asking the software to solve. What such simplifications you can make is very much problem dependent. Best, Jeff
lz Chen
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 8 years ago Jan 9, 2017, 10:56 a.m. EST
Dear Jeff,

Thank you very much for your patient guidance, I am trying to learn how to improve my model, following your suggestions and website. Thank you again !

Regards,
Lz
Dear Jeff, Thank you very much for your patient guidance, I am trying to learn how to improve my model, following your suggestions and website. Thank you again ! Regards, Lz

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.

Suggested Content