Cluster 2 server units to run star with mpi?

RCB0B · August 16, 2017, 02:58

Hello Everyone,

I got 2 server units each with centos and star ccm+. I have a network switch with gigabit ethernet capability (as do the servers). They have 10gbit ethernet cards too just in case.

I want to cluster them together to run star but I'm not exactly sure on where to start. From some internet searching it looks like this will have to be done via setting up MPI with file sharing and such on each server. I see a lot of support and instructions on how to set up MPI or openMPI etc but I'm not sure which one to go for exactly and how it'll be implemented with star, wondering if there's a specific one that works best with star perhaps? I'm hoping to cluster them such that one is a frontend node and one is a compute node and it would be nice to still use the star gui.

Tried using Pelican HPC and it doesn't seem to be ideal for what I'm looking for.

I'm using a POD key too btw.

Any input would be appreciated!

Bob

me3840 · August 16, 2017, 21:05

You can either use OpenMPI or Platform MPI, which comes with star. You must install OpenMPI yourself if you want to use it.

You will have to have both machines share a directory somehow. This can be accomplished with NFS or with SSHFS, both are free. There are lots of guides around on doing this.

Install star on both machines in the same directory (or if not at least provide an alias for the starccm+ binary on both machines).

Both machines must be connected with a public-key authenticated ssh session. Again, lots of tutorials on this around the internet.

Then just make a text file with a list of the machines hostnames for each processor. For example, two quad-core machines would have a file:

machine1
machine1
machine1
machine1
machine2
machine2
machine2
machine2

Then just launch star with whatever script you want to use, you will need to provide the hosts file made above as an argument. Check the user guide on parallel execution for the syntax, don't know it offhand.

You can of course enter this info into the GUI instead, but I wouldn't recommend that for this type of machine. Better to have no frontend on a computing machine.

August 16, 2017, 02:58	Cluster 2 server units to run star with mpi?	#1
RCB0B New Member Bob Comstock Join Date: Aug 2017 Posts: 4 Rep Power: 9	Hello Everyone, I got 2 server units each with centos and star ccm+. I have a network switch with gigabit ethernet capability (as do the servers). They have 10gbit ethernet cards too just in case. I want to cluster them together to run star but I'm not exactly sure on where to start. From some internet searching it looks like this will have to be done via setting up MPI with file sharing and such on each server. I see a lot of support and instructions on how to set up MPI or openMPI etc but I'm not sure which one to go for exactly and how it'll be implemented with star, wondering if there's a specific one that works best with star perhaps? I'm hoping to cluster them such that one is a frontend node and one is a compute node and it would be nice to still use the star gui. Tried using Pelican HPC and it doesn't seem to be ideal for what I'm looking for. I'm using a POD key too btw. Any input would be appreciated! Bob

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Run Java macros in Star Ccm+	peter parkour	STAR-CCM+	2	November 6, 2015 03:48
How to run custom solvers in parallel (CLUSTER) ?	iy-a	OpenFOAM Running, Solving & CFD	0	August 3, 2015 07:45
does anyone run a Windows Cluster?	Eike	STAR-CCM+	0	December 3, 2012 10:06
Working directory via command line	Luiz	CFX	4	March 6, 2011 21:02
[Commercial meshers] Trimmed cell and embedded refinement mesh conversion issues	michele	OpenFOAM Meshing & Mesh Conversion	2	July 15, 2005 05:15

August 16, 2017, 21:05		#2
me3840 Senior Member Join Date: Nov 2010 Location: USA Posts: 1,232 Rep Power: 25	You can either use OpenMPI or Platform MPI, which comes with star. You must install OpenMPI yourself if you want to use it. You will have to have both machines share a directory somehow. This can be accomplished with NFS or with SSHFS, both are free. There are lots of guides around on doing this. Install star on both machines in the same directory (or if not at least provide an alias for the starccm+ binary on both machines). Both machines must be connected with a public-key authenticated ssh session. Again, lots of tutorials on this around the internet. Then just make a text file with a list of the machines hostnames for each processor. For example, two quad-core machines would have a file: machine1 machine1 machine1 machine1 machine2 machine2 machine2 machine2 Then just launch star with whatever script you want to use, you will need to provide the hosts file made above as an argument. Check the user guide on parallel execution for the syntax, don't know it offhand. You can of course enter this info into the GUI instead, but I wouldn't recommend that for this type of machine. Better to have no frontend on a computing machine.