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.
Running COMSOL in distributed-memory mode
Posted 16 mai 2011, 12:43 UTC−4 3 Replies
Please login with a confirmed email address before reporting spam
Hello everyone,
So far, I have only used a cluster for solving my COMSOL problems in shared-memory mode and then I use several processors but only one computational node. However, now I have a very large problem (almost 3,000,000 DOF) and I need more memory in order to get a solution. Then I am thinking of using the distributed-memory mode but I have some doubts about it:
1) I don't know if the license I have allows me to use this mode. Currently, I have a floating network license for 1 concurrent user and I can only execute a file at a time. Let me explain this: if I send a file to the cluster (shared-memory mode), it is going to be solved by one sole machine. However, I can't send a different file to be executed at the same time on another different machine because I get a license error. Then I am wondering if I will be able to use the distributed-memory mode since, to my knowledge, COMSOL will run in several physical nodes (machines). Then, can I use an unlimited number of nodes (I mean as many nodes as the cluster has)?
2) Do I have to build my models in a different way in order to solve them using the distributed-memory mode?
3) I don't want to make a parametric sweep, I only have a rather memory-consuming problem that I want to solve using a direct solver. Can I use any of the direct solvers? Do yo reccommend me one of them in particular?
4) I am still using COMSOL 3.5 because I am adapting to version 4.1. Would there be many differences (regarding calculation speed and so on) if I used version 4.1?.
Thank you very much for your help
So far, I have only used a cluster for solving my COMSOL problems in shared-memory mode and then I use several processors but only one computational node. However, now I have a very large problem (almost 3,000,000 DOF) and I need more memory in order to get a solution. Then I am thinking of using the distributed-memory mode but I have some doubts about it:
1) I don't know if the license I have allows me to use this mode. Currently, I have a floating network license for 1 concurrent user and I can only execute a file at a time. Let me explain this: if I send a file to the cluster (shared-memory mode), it is going to be solved by one sole machine. However, I can't send a different file to be executed at the same time on another different machine because I get a license error. Then I am wondering if I will be able to use the distributed-memory mode since, to my knowledge, COMSOL will run in several physical nodes (machines). Then, can I use an unlimited number of nodes (I mean as many nodes as the cluster has)?
2) Do I have to build my models in a different way in order to solve them using the distributed-memory mode?
3) I don't want to make a parametric sweep, I only have a rather memory-consuming problem that I want to solve using a direct solver. Can I use any of the direct solvers? Do yo reccommend me one of them in particular?
4) I am still using COMSOL 3.5 because I am adapting to version 4.1. Would there be many differences (regarding calculation speed and so on) if I used version 4.1?.
Thank you very much for your help
3 Replies Last Post 17 mai 2011, 16:53 UTC−4