Hi Jeff, as a tools provider I would also like MPI_T to expose other sensors, as this means I don't have to support yet another metric interface. Not stacking a multitude of different metric interfaces on top of each other might also cut down on the overhead as you said. But could an implementation not expose such info right now? I thought the whole idea of not standardizing anything about counters, their names and semantics, is that implementations can expose whatever they care about. Cheers, Marc-Andre On 17.12.13 21:39, Jeff Squyres (jsquyres) wrote:
We talked briefly in the meeting last week about using MPI_T to gather non-MPI information, such as temperatures, fan speeds, CPU counters, ...etc.
I was originally in favor of such things, but I think I got talked down by the argument of "it seems like there are already other tools to do this; why would we [also] expose these kinds of values through MPI_T?"
Which I think I agree with.
But I had a thought on the way home from Chicago as to why we would want to do this: using a single tool to gather these values could give a more consistent set of numbers. I.e., if you're gathering all the values through a single interface (e.g., MPI_T) you can get a consistent view of *all* sensor values, to include MPI introspective sensors (average queue depths, number of sends, ...etc.). If you're using multiple tools to gather these values, there's variability introduced by using multiple difference interfaces, each with different costs of reading values, etc.
Is this a worthwhile observation? Or should MPI_T just stick to gathering MPI-related information?
-- Marc-Andre Hermanns German Research School for Simulation Sciences GmbH c/o Laboratory for Parallel Programming 52062 Aachen | Germany Tel +49 241 80 99753 Fax +49 241 80 6 99753 Web www.grs-sim.de Members: Forschungszentrum Jülich GmbH | RWTH Aachen University Registered in the commercial register of the local court of Düren (Amtsgericht Düren) under registration number HRB 5268 Registered office: Jülich Executive board: Prof. Marek Behr, Ph.D | Prof. Dr. Sebastian M. Schmidt