Running Multiple Scenarios Simultaneously on One Computer

[Anonymous]

Hi All,

This is a long shot, but I was wondering if there was a way to run simulations simultaneously on one computer. The computer I have is miles ahead of anything else avaiable on the network, and has 12 cores. I know CFD runs on 2^n cores, so 4 are wasted right now, so I was wondering if there was a way to split to processing power through just one computer, since it will still be better than remote solving to another computer.

If I use the remote solver (guidelines link below), can I go into the cfdserv application and specify the same computer, but one as 4 cores and one as 8 cores?

http://forums.autodesk.com/t5/Autodesk-Simulation-CFD/Setting-up-Remote-Solve-Fast-Track/td-p/321199...

I assume not, given the IP address is the same, but I figure I'd ask.

Thanks.

[Royce_adsk · Accepted solution]

Hi Max,

Just as you guessed you can only run 1 job at a time on 1 computer.  I suspect if you tried to do what you suggested it would still get queued.

-Royce

[Anonymous]

Ahh, shame, but thanks for the response. 

[OmkarJ]

Well I use two combinations of cores on the cluster : 8 cores and 16 cores. Now instead of running jobs on 16 cores serially, with one at a tine, I typically run two jobs running parallely with 8 cores each.

I have seen performance improvement of 25% while going from 4 to 8 cores, with anther 25% improvement from 8 to 16 cores. If you do the math, you will realise that two jobs with 8 cores runnign parallely finish in 75% time as that of two jobs running in series with 16 cores.

Now, I am being told that even with two licences, it is not always possible to run two jobs with 8 cores on the cluster simultaneously. In my case, I can so UI often do. As a disclaimer, there may be a risk of erraticv behaviour from cluster while communicating the solution to client. But the economisation is worth the risk.

Cheers

OJ

[OmkarJ]

Additionally, specific to your case, the two combnations can be 4 cores and 8 cores etc. Even 4 cores running parrallely might be quicker than 8 cores running in series!