|
[Sponsors] |
October 4, 2007, 10:29 |
Paradox in parallel computations
|
#1 |
Guest
Posts: n/a
|
Hi CFD users,
I am running a 3D simulation of a centrifugal pump in Fluent 6.2.16. First i used the MRF method and all the cases were simulated in a single machine. There were no convergence problems. Recently we made a network cluster of four machines. The OS is Linux SUSE 10. I am novice in parallel simulations. I am running the previous simulations with SM method using the MRF as an initial solution in parallel mode. I partition the grid in a serial solver and then I read the cas/dat in the parallel solver. Here is the strange: When I group all the interfaces in a single partition by activating the "encapsulated grid interface " in Partition panel (This is recommended by Fluent manual for non-conformal meshes), my solution is diverged in the parallel solver. When de-activate the "encapsulated grid interface" my solution converges!!! Can anyone know why this happens? Isn't it strange? Thank you in advance, |
|
October 9, 2007, 12:09 |
Re: Paradox in parallel computations
|
#2 |
Guest
Posts: n/a
|
Yes, this may happen, because the encapsulation is causing very long partition interfaces, which is very bad for convergence.
|
|
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Script to Run Parallel Jobs in Rocks Cluster | asaha | OpenFOAM Running, Solving & CFD | 12 | July 4, 2012 22:51 |
Cumulative continuity error large in parallel simulations | xiao | OpenFOAM | 3 | March 7, 2011 04:13 |
Problems of running the parallel computations | lorraine | FLUENT | 17 | July 6, 2007 08:08 |
Paradox in parallel processing | Vagelis | FLUENT | 0 | October 26, 2005 05:36 |
Parallel Computing Classes at San Diego Supercomputer Center Jan. 20-22 | Amitava Majumdar | Main CFD Forum | 0 | January 5, 1999 12:00 |