US ATLAS HL-LHC computing discussion

[Torre Wenaus <wenaus AT gmail.com>]

Hi Pavlo,

His talk was no-slides, not everyone could do that but Miron can :-) I missed his talk.

  Torre

On Thu, Mar 21, 2019 at 11:18 AM Pavlo Svirin <pavlo.svirin AT cern.ch> wrote:

Hi Torre,

are there any slides available from Miron's talk? This already sounds interesting but some details about their current development will be greatly appreciated.


Best regards,
Pavlo Svirin.



> On Mar 21, 2019, at 4:12 PM, Torre Wenaus <wenaus AT gmail.com> wrote:
>
> Hi,
> I had a chat with Miron here at JLab. He's interested to know how we can collaborate. I mentioned one area that came to mind, please comment on it and how we might get something going if you think it's reasonable. We want to co-schedule CPU and GPU payloads on big HPCs. We're currently thinking of doing that at the PanDA/pilot level. Condor doesn't even function at this point on HPCs without outbound connectivity from worker nodes. But as Miron mentioned and as we've heard here in the meeting, Condor is working on this, in particular at Barcelona, using the file system to communicate with workers via an edge service just as we do. So Condor is on its way in the door on these systems, and Miron is interested in the CPU/GPU co-scheduling use case.
>   Torre