|
[Sponsors] |
Cluster from pc having different processors and/or different mobo and/or different ra |
|
LinkBack | Thread Tools | Search this Thread | Display Modes |
February 11, 2018, 06:56 |
Cluster from pc having different processors and/or different mobo and/or different ra
|
#1 |
New Member
BALRAJ SINGH
Join Date: Feb 2018
Posts: 9
Rep Power: 8 |
I have five computers, one i5 dell workstation with 32 gb 1600hz ram, two Q 9650 hp with 16 gb 1300hz, two Q 9550 with 8 gb 1060hz each. Can I make a cluster of three or five for use in fluent, cfx, cfd softwares.
|
|
February 11, 2018, 06:58 |
|
#2 |
New Member
BALRAJ SINGH
Join Date: Feb 2018
Posts: 9
Rep Power: 8 |
I have z 800 with 1300hz 92 gb ram, two xenon 5650, quadro 4000 4 gb, two sas drives on raid zero. Can it be included in cluster as head.
|
|
February 12, 2018, 04:19 |
|
#3 |
Super Moderator
Alex
Join Date: Jun 2012
Location: Germany
Posts: 3,400
Rep Power: 47 |
You can create a cluster with ANY type of hardware, it does not have to be homogeneous.
However, on a cluster with heterogeneous hardware you will face problems with load balancing, i.e. the slowest node will limit total performance, faster nodes will have to wait until the slower ones are finished iterating. There should be ways to to circumvent this, for example assigning smaller MPI domains to slower nodes. But I have no experience how to do this in Ansys simulation software. Maybe someone else can give you a hint where to find this information. |
|
February 22, 2018, 11:50 |
|
#4 |
New Member
BALRAJ SINGH
Join Date: Feb 2018
Posts: 9
Rep Power: 8 |
Thank you Sir
|
|
February 23, 2018, 23:55 |
Cluster from pc having different processors and/or different mobo and/or different ra
|
#5 |
Member
Join Date: Jan 2015
Posts: 62
Rep Power: 11 |
I've linked multiple high end desktops up before and used CFX. You can in your hostinfo.ccl file assign relative processor speeds to each computer to account for the different speeds, which I believe biases the partition sizes.
However, this "cluster" didn't work great even with 10 Gig E interconnects. Adding the third computer didn't reduce solver time at all. Also, when we updated to a new CFX version and thus a new MPI version, I got segmentation violation errors trying to re run our benchmark file. You'll need expensive ANSYS HPC licenses to use multiple cores. |
|
February 24, 2018, 00:36 |
|
#6 |
New Member
BALRAJ SINGH
Join Date: Feb 2018
Posts: 9
Rep Power: 8 |
Thanks Sir for information in detail. I will get hpc for year from my college
|
|
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Socket 2011-3 processors - an overwiew | flotus1 | Hardware | 10 | June 17, 2017 09:47 |
v4.2.0 diverges with 32 processors but converges with 16. v4.0.0 worked with 32 | asthelen | SU2 | 0 | July 6, 2016 00:27 |
Pinning the processors in cluster to run a solver | coolcrasher | OpenFOAM Running, Solving & CFD | 0 | November 5, 2015 06:11 |
Why not install cluster by connecting workstations together for CFD application? | Anna Tian | Hardware | 5 | July 18, 2014 14:32 |
Question about using cluster processors to run .def file | Anna Tian | CFX | 1 | November 18, 2013 10:10 |