Robert Koslover
Certified Consultant
Please login with a confirmed email address before reporting spam
Posted:
3 years ago
2022年3月26日 GMT-4 23:46
I think the easiest way to leverage multiple PCs is in the case of long runs, in which you are simply sweeping through a range of many parameter values, basically running a model over and over as certain parameter(s) change. In that case, you can divide up your problem space into N separate smaller sets of runs, on N separate computers. In that case, the computers don't have to share any memory (you could simply copy initial files between them to set up your separate runs). Of course, ideally, you (and everyone else) would likely prefer to link your multiple computers together to have them act as if they were just one big/amazing computer, sharing (and addressing) all of the combined RAM and disk space, and distributing all computational tasks too, optimally divided among all the processors. Well... in that case, I suggest you ask Comsol's in-house experts directly for detailed advice about what you actually might be able to configure with this software and how. I'm sure they have some ideas to share about how to implement multi-platform computing most efficiently. Good luck.
-------------------
Scientific Applications & Research Associates (SARA) Inc.
www.comsol.com/partners-consultants/certified-consultants/sara
I think the *easiest* way to leverage multiple PCs is in the case of long runs, in which you are simply sweeping through a range of *many* parameter values, basically running a model over and over as certain parameter(s) change. In that case, you can divide up your problem space into N separate smaller sets of runs, on N separate computers. In that case, the computers don't have to share *any* memory (you could simply copy initial files between them to set up your separate runs). Of course, *ideally*, you (and everyone else) would likely prefer to link your multiple computers together to have them act as if they were just one big/amazing computer, sharing (and addressing) all of the combined RAM and disk space, and distributing all computational tasks too, optimally divided among all the processors. Well... in that case, I suggest you ask Comsol's in-house experts directly for *detailed* advice about what you actually might be able to configure with this software and how. I'm sure they have some ideas to share about how to implement multi-platform computing most efficiently. Good luck.
Please login with a confirmed email address before reporting spam
Posted:
3 years ago
2022年3月27日 GMT-4 14:48
@robert Thats a good idea, thank you.
@robert Thats a good idea, thank you.
Jeff Hiller
COMSOL Employee
Please login with a confirmed email address before reporting spam
Posted:
3 years ago
2022年3月29日 GMT-4 16:33
Updated:
3 years ago
2022年3月30日 GMT-4 16:11
Hi Sanjay,
This blog post is a bit old but it deals with your question directly.
From a licensing perspective, cluster computing is permitted under the FNL, CSL and ASL license agreements, and not under the NSL and CPU license agreements.
Best regards,
Jeff
-------------------
Jeff Hiller
Hi Sanjay,
[This blog post](https://www.comsol.com/blogs/building-beowulf-cluster-faster-multiphysics-simulations/) is a bit old but it deals with your question directly.
From a licensing perspective, cluster computing is permitted under the FNL, CSL and ASL license agreements, and not under the NSL and CPU license agreements.
Best regards,
Jeff