[GE users] 'serial' gaussian03

Reuti reuti at staff.uni-marburg.de
Wed Apr 6 16:48:36 BST 2005

    [ The following text is in the "ISO-8859-1" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some special characters may be displayed incorrectly. ]

Hello Tom,

Tom Kuppens wrote:
> Hi there,
> I use guassian03 on a self-assembled beowulfy-like cluster. It's a 
> heterogeneous environment with different processors.
> I do not use the parallel linda approach because of the heterogeneous
> 'flavor' of my cluster. I distribute my jobs to the different CPU's, so 
> that 1 CPU has 1 job at hand. To distribute my g03 jobs I created a tool    
> that allows me to automate it (rather then do it by hand, like I used to  
> do ...) based on proc/loadavg, availability etc.

you can also integrate Linda, as long as for one parallel job you stay 
on one type of machine. You just need Linda for all types of machines 
you have - or just limit parallel jobs to run only on one platform. This 
can be configured in SGE.

> Now I was wondering if it is also possible with SGE to do this in an easy
> way, or does it only manage to parallelize calculations with linda.   
> The number of Gaussian jobs I am discussing is around 200 ? 300 per
> problem, so it is important that each job should not be submitted by hand.

What automatism are you looking for? You could a) use a shell loop to 
submit the whole bunch of jobs, or b) use a so called array-job, if 
there is the chance to select an input file by a varying parameter - 
depends on your workflow. How did you create and submit your jobs up to now?

Cheers - Reuti

> As I am a SGE dummy; is it hard to install (for a simple chemist) and is 
> there a lot of work needed on the slave nodes (like services, daemons, 
> dependencies ...), where to start in the pile of documentation ...
> I'm happy with my 'way of doing it' but I try to keep my horizon open if
> other methods can be applied and how they work, and perhaps apply them
> eventually ;-)
> Thank you for your time,    
> Tom Kuppens
> Quantum Group Ghent
> Department of Inorganic and Physical Chemistry WE06V
> Ghent University
> Krijgslaan 281 - S3, 9000 Ghent, Belgium
> Tel  +32(0)9 264 4444
> Fax  +32(0)9 264 4983

To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net

More information about the gridengine-users mailing list