Re: [Condor-users] Condor and GPUs

[Steffen Grunewald <steffen.grunewald@xxxxxxxxxx>]

On Fri, Jul 04, 2008 at 06:14:02AM -0500, Miron Livny wrote:
> All,
> 
> Another question that will help me form a response to your very 
> thoughtful and useful input . What are your thoughts about time 
> limits on the execution. In other words, if a jobs says "I need 4 
> cores and 2.5 G of memory and  7.3 M of network bandwidth" will it 
> also say "for 2 hours"? If does, what are you going to do if the jobs 
> uses more or less than 2 hours?

While memory footprint and network bandwidth are predictable, wall-clock
time (is that what you're asking for?) won't be.

Pools can be very heterogeneous. Even with "comparable" hardware:
Imagine a mix of Intel and AMD processors, and apps built with Intels
compiler suite :-(
Or a saturated network or file server resulting in loooonger waits for
I/O.
And it's unrealistic to expect users to specify a limit on the number of
Flops (although Einstein@Home seems to know what they're doing, the
average user will at best have an estimate how long the app would 
take on her "development machine". Which can be off by an order of
magnitude). YMMV though.

I know that OpenPBS has mechanisms to limit jobs to a certain wall-
clock time...

Cheers,
 Steffen