From: Malcolm T. <mt...@wu...> - 2009-12-01 19:24:06
|
Wilfred, On Tuesday 01 December 2009 01:16:44 pm Wilfred Li wrote: > It's probably most portable if you have your application report such > statistics as one of the outputs in the Opal Job output directory. I thought about that, but we're supporting multiple applications, and even if they could all be convinced to add the output, there's no guarantee that they would report things in a consistent way (i.e. max. physical vs. virtual memory), nor is there any guarantee that the output wouldn't change at some point in the future. Finally, there are occasional 'run away' jobs that have to be killed so would never report their usage even thought they had consumed resources. Since the queuing system already has access to all this information I'd like to find a way to leverage it (assuming it's easy enough to do). Thanks for the response. Malcolm -- Malcolm Tobias 314.362.1594 |