|
[Sponsors] |
What about Thin Clients as terminals for CFD? |
|
LinkBack | Thread Tools | Search this Thread | Display Modes |
April 23, 2011, 10:31 |
What about Thin Clients as terminals for CFD?
|
#1 |
New Member
Romulo Heringer
Join Date: Jul 2009
Posts: 8
Rep Power: 17 |
Hi all,
We intend to install a new CFD lab for didactic purposes. We think that a server(s)/(30)clients configuration may be a better solution (maintenance in mind). The clients will be minimal terminals with some graphical capabilities while the server take the hard processing work (as server/client should be). The question: THIN CLIENTS could be used as client terminals in this case (numerical simulation, visualization)? Any suggestion will be appreciated. Thanks all, rheringer |
|
April 24, 2011, 09:57 |
|
#2 |
Retired Super Moderator
Bruno Santos
Join Date: Mar 2009
Location: Lisbon, Portugal
Posts: 10,982
Blog Entries: 45
Rep Power: 128 |
Greetings rheringer,
It's theoretically possible, but:
Of course that if you're trying to save some cash with this set-up, than the server should use a free Linux distribution Another thing you should take into account is what type of thin-client you are planning to use, or more importantly, if you're going to use Atom based clients or ARM based clients (I'm talking about their CPU ). ARM based clients in CFD is something I'm not familiar with, but at least ParaView should work well with them, which is something I think can't be said for commercial CFD applications... Best regards, Bruno
__________________
|
|
April 24, 2011, 12:17 |
|
#3 |
New Member
Romulo Heringer
Join Date: Jul 2009
Posts: 8
Rep Power: 17 |
Thank you Bruno, for your answer.
I think we chose a good video card for the server (nVidia QUADRO FX4800). But, an ordinary video card (even on-board) installed in each terminal couldn't make this work? (cfd post-processing in college course level) I mean, couldn't the graphical acceleration be processed in the client terminal? The system we have in mind is: - Software (open source) - OS: Ubuntu GNU/Linux or another convenient linux distribution CFD package: OpenFOAM (1st option) Visualization: Paraview - Hardware (server) - CPU: 2 Processors Intel Xeon X5690 HT - 6x 3.46GHz, TDP 130w, Cache 12MB Mobo: Intel Workstation Board Dual Xeon, Modelo S5520SC Chipset: Intel S5520 (Tylersburg) RAM: 48 GB SSD: 120 GB X25-M Intel SSDSA2MH120G2K5 Hard Disk: 2 TB SATA3Gb/s, Seagate Barracuda LP, Cache 32MB Video: PNY nVidia QUADRO FX4800, 1.5 GB Best regard, Romulo |
|
April 24, 2011, 23:14 |
|
#4 |
Super Moderator
Takuya OSHIMA
Join Date: Mar 2009
Location: Niigata City, Japan
Posts: 518
Blog Entries: 1
Rep Power: 20 |
Hi Romulo, I've been teaching CFD and stress analyses for several years in an undergraduate course of around 45 students. Our system is:
Thin clients: Celeron 2.8 GHz (cutdown version of Pentium4), 1GB RAM, integrated Intel graphics Windows XP Professional OpenFOAM 1.4 for Cygwin, Gmsh 2.0.8, ParaView 3.4 (the software are more or less customized to suit the system and to meet our specific needs) Server: Hardware unknown, houses boot image of the thin clients, provides disk space of 120 MBytes(!) per student The hardware are becoming outdated (> 5 years old) but even with the system we are barely able to run cases of around OpenFOAM tutorial sizes (~30 000 cells) and pre/post-processings on the thin clients. The optimal hardware/software configuration quite depends on what you plan to do, but in any case, I would say that what really count are good teaching assistants...! Takuya |
|
April 25, 2011, 23:29 |
|
#5 |
New Member
Romulo Heringer
Join Date: Jul 2009
Posts: 8
Rep Power: 17 |
Hi Takuya,
Thank you for sharing your experience with us. The system we are looking for seems quite similar to your lab (hardware/software). It's useful to known this config work well somewhere. Maybe the real challenge will be take all the software working fine on the server/terminals, instead the particular hardware used... Do you think some linux distribution could be easily used in your case, instead Windows XP with Cygwin? Romulo |
|
April 26, 2011, 05:17 |
|
#6 | ||
Super Moderator
Takuya OSHIMA
Join Date: Mar 2009
Location: Niigata City, Japan
Posts: 518
Blog Entries: 1
Rep Power: 20 |
Hi,
Quote:
Quote:
Takuya |
|||
April 26, 2011, 23:05 |
|
#7 | |
New Member
Romulo Heringer
Join Date: Jul 2009
Posts: 8
Rep Power: 17 |
Quote:
Good tip. I'll take this into account. Thanks again, Romulo |
||
June 8, 2011, 18:14 |
|
#8 |
New Member
Romulo Heringer
Join Date: Jul 2009
Posts: 8
Rep Power: 17 |
Unfortunately we could not make conclusive tests with thin client / server (as proposed by Bruno). Thus, we chose a solution similar to that of Takuya (above).
However, if anyone knows a successful solution using (real) thin clients, please let us know. Thank you both. Romulo |
|
June 9, 2011, 17:25 |
|
#9 |
Retired Super Moderator
Bruno Santos
Join Date: Mar 2009
Location: Lisbon, Portugal
Posts: 10,982
Blog Entries: 45
Rep Power: 128 |
Hi Romulo,
I haven't gotten my hands on a little thin client yet, but I did test yesterday a very sweet innate ability that ParaView has: I used a ParaView 3.8.0 32bit on Windows to connect to a 64 bit 3.8.0 ParaView server (pvserver) running on Linux! Unfortunately, I didn't have time to do some serious testing, but at least it connected very well and I could even browse the remote folders from within my local ParaView! If I get another chance to do some tests, I'll write about it here By the way, if this works well, it's not very difficult to imagine users using terminal windows to access remotely their areas in the server to edit their OpenFOAM cases, running them and then plugging into their own session of pvserver from their ParaView client! Best regards, Bruno
__________________
|
|
Tags |
cfd lab, thin client |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Thin Walls using blockMesh | varun | OpenFOAM Running, Solving & CFD | 6 | December 8, 2020 14:18 |
Reversed flow error with thin film heater | MK | FLUENT | 1 | February 12, 2009 11:26 |
Thin surfaces within a body using DesignModeler? | Forrest | CFX | 7 | April 5, 2005 09:21 |
Meshing thin surfaces | Louwrens | CFX | 8 | August 18, 2003 18:49 |
Modelling thin surfaces in 5.6 | Louwrens | CFX | 0 | August 16, 2003 11:47 |