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.

out of core options - the difference

Please login with a confirmed email address before reporting spam

What is the difference between:

1) Direct solver PARDISO (out of core) *
2) Store the solution on file **

* cl.ly/O4B
** cl.ly/OrP

It would be great if you share your experience with these options.

Many thanks.

4 Replies Last Post 2010年4月16日 GMT-4 05:23
Jim Freels mechanical side of nuclear engineering, multiphysics analysis, COMSOL specialist

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2010年4月15日 GMT-4 22:17
Saving the solution on file does just that. In COMSOL when you click the "store a solution" button on the solver manager, the default behavior is to store the solution into memory. The COMSOL developers recognized that the problem of running out of memory could be improved upon by allowing the stored solution to also be saved to a disk file as an option. So, now there is a box to click in the advanced settings when you want to do that. This increases the memory available for solving your problem.

The PARDISO (out of core) solver, will allocate a reserved amount of disk space to add to the memory available for PARDISO to solve a problem. PARDISO is desirable by users because it is a direct solver and very easy to implement a solution (compared to say an iterative, geometric multigrid solution for example). You might think of it as a separate "swap space" or "virtual memory" just there for PARDISO to use instead of the operating system. Keep in mind that the PARDISO (out of core), while increasing the memory available, will also slow down the computations significantly while the data is written and reading from the hard disk. This is because data is transferred much faster to memory than to a hard disk. In my opinion, it is much quicker to learn how to use the iterative solvers then to wait on the out of core solver. Or perhaps, while you are waiting, you can read up on the iterative solver for the next time.

Incidently, you can store the solution to a file AND solve out of core simultaneously if you want to. That would be a interesting experiment to see how much memory you could save for a given problem.

Saving the solution on file does just that. In COMSOL when you click the "store a solution" button on the solver manager, the default behavior is to store the solution into memory. The COMSOL developers recognized that the problem of running out of memory could be improved upon by allowing the stored solution to also be saved to a disk file as an option. So, now there is a box to click in the advanced settings when you want to do that. This increases the memory available for solving your problem. The PARDISO (out of core) solver, will allocate a reserved amount of disk space to add to the memory available for PARDISO to solve a problem. PARDISO is desirable by users because it is a direct solver and very easy to implement a solution (compared to say an iterative, geometric multigrid solution for example). You might think of it as a separate "swap space" or "virtual memory" just there for PARDISO to use instead of the operating system. Keep in mind that the PARDISO (out of core), while increasing the memory available, will also slow down the computations significantly while the data is written and reading from the hard disk. This is because data is transferred much faster to memory than to a hard disk. In my opinion, it is much quicker to learn how to use the iterative solvers then to wait on the out of core solver. Or perhaps, while you are waiting, you can read up on the iterative solver for the next time. Incidently, you can store the solution to a file AND solve out of core simultaneously if you want to. That would be a interesting experiment to see how much memory you could save for a given problem.

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2010年4月16日 GMT-4 02:49
Thank you James.

In time dependant case, when I enable the "store on disk" option, my memory usuage does not change much during let's say 10k timesteps. However, if I disable it, then RAM usuage goes around 32GB in 10k time steps. So, it seems during the time-stepping it also keeps the previous time-steps on disk -- does it keep it usually in memory? I don't know. That's why I still don't see the difference of out-of-core and store-on-disk options clearly. I understood from your post that out-of-core option is also "another" RAM saving option other than store-on-disk.
Thank you James. In time dependant case, when I enable the "store on disk" option, my memory usuage does not change much during let's say 10k timesteps. However, if I disable it, then RAM usuage goes around 32GB in 10k time steps. So, it seems during the time-stepping it also keeps the previous time-steps on disk -- does it keep it usually in memory? I don't know. That's why I still don't see the difference of out-of-core and store-on-disk options clearly. I understood from your post that out-of-core option is also "another" RAM saving option other than store-on-disk.

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2010年4月16日 GMT-4 03:44
Is comsol "swap" more eficient than system swap on linux system?

the reason I ask is that I guess that if I dont use the store on disk option comsol memory growing requirement will lead to using VM on swap.
I just happen to have my system swap file stripped on 4 drives and I wonder if COmsol will take advantage of the gain .
so what will be better use store on file or let the OS manage the memory requirement..?
Is comsol "swap" more eficient than system swap on linux system? the reason I ask is that I guess that if I dont use the store on disk option comsol memory growing requirement will lead to using VM on swap. I just happen to have my system swap file stripped on 4 drives and I wonder if COmsol will take advantage of the gain . so what will be better use store on file or let the OS manage the memory requirement..?

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2010年4月16日 GMT-4 05:23
Hi

I have read somewherer (perhaps on the forum) that people managed to get more speed by using "memory drives" swaps too, but this is anyhow related to HW design, RAM cash, disk cash and how many parasite processes are running. Definitively UX/linux is far more efficient than MS VISTA on swap management

Ivar
Hi I have read somewherer (perhaps on the forum) that people managed to get more speed by using "memory drives" swaps too, but this is anyhow related to HW design, RAM cash, disk cash and how many parasite processes are running. Definitively UX/linux is far more efficient than MS VISTA on swap management Ivar

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.