[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Condor-users] Just curious how you....



Hello Ian,

A Good day to you!

Yes.... this is more or less what we do at our work place too.....

Basically, we do CFD Simulations, but not the really hard-core ones which require 10s of Processor Cores and two days to get some useful results, but relatively small ones which range from 500,000 to around 6 Million Cells.

For something of this nature, it would make no sense to have a dedicated cluster of Linux systems, and I would never get the funds to implement something like that. Hence what we have done is to have one dedicated Linux "Master" which is a Quad-Core. This system acts as the Master for a Condor Cluster, and hosts all the Simulation tools which are exported via NFS.

The Quad-Core also acts as the "base-load" system in case we need to do some urgent simulations during the normal working hours.

The rest of the Cluster of "Slave" computers are basically all our normal "Day Computers". During the working hours, we use it as a usual office computer with Windows.... After office hours, we reboot the systems (number depending on the Simulation load we have) into the dual-boot Linux "personality".

During boot-up these systems automatically mount the NFS exports, our shared storage system, and advertises themselves to the Condor Pool running on the Master computer. The Cluster then starts working through the queue of simulations which were submitted to the Master over the course of the day.

So far this concept has worked very well, and I am quite satisfied with it..... the only downside being that I don't have access to a high performance Network interconnect between the computers (such as Infiniband, etc...) which would help a lot when we perform OpenMPI based Parallel simulations across cores on multiple computers.


I think Condor is designed with such use-cases in mind.... and if I am not mistaken, this is basically some form of "CPU / Cycle - scavenging" right?

Have a nice day ahead!

Philippose





On Wed, Jan 19, 2011 at 6:09 PM, Ian Cottam <Ian.Cottam@xxxxxxxxxxxxxxxx> wrote:
Just curious if (any) other sites run their Condor Pool the way I (and my
colleagues) do?

Basically, we have PC Cluster rooms all over campus. They run Windows when
students have access to them.
Overnight, at weekends and student vacs we automatically reboot them into
Linux with Condor.
When students are around they, of course, automatically reboot back into
Windows every weekday morning.

We have some separate Linux PCs --about 80 processor cores-- that keeps
things at 'tick over' during the student working day.

The drivers for this approach include: that students never see a slow PC;
and that we don't need rock hard security (if such exists) within Condor
to stop someone destroying our Windows teaching image.


Assuming you don't do it this way, I'm also interested in why not and what
you do do.
-Ian


_______________________________________________
Condor-users mailing list
To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
subject: Unsubscribe
You can also unsubscribe by visiting
https://lists.cs.wisc.edu/mailman/listinfo/condor-users

The archives can be found at:
https://lists.cs.wisc.edu/archive/condor-users/