Hi Scott,

 

  Are you’re saying a facility panel is completely uninteresting to you, and you’d rather hear from users?

 

  My expectation is that the facility panel would tell you much more about what the systems they design/install/manage are likely to look like rather than about specific users (though I find that I’ve had to do a lot of user interviews at LANL, and most facilities likely have similar people). And I recognize that there is a decent amount of diversity between ALCF, OLCF, and NERSC and further that I have the advantage of being able to see that diversity while a lot of attendees will not be as aware of it – so a panel with a rep from those centers would be broadly useful.

 

  What I’m less certain about is how much diversity there is in data acquisition and processing at instruments. And I don’t really know anyone who knows more than one instrument DAQ system very well. And I wonder if there are new interesting requirements related to differences between these instruments. But if no such person exists, then no such person exists.

 

Cheers,

Brad

 

From: "Klasky, Scott A." <klasky@ornl.gov>
Date: Tuesday, July 31, 2018 at 3:26 PM
To: "Settlemyer, Bradley Wade" <bws@lanl.gov>, "Mohror, Kathryn" <mohror1@llnl.gov>, "Ross, Robert B." <rross@mcs.anl.gov>, "ssio-2018@lists.mcs.anl.gov" <ssio-2018@lists.mcs.anl.gov>
Subject: Re: [Ssio-2018] notes from call

 

My 2 cents

I care about scale

 

Number of simultaneous data sets accessed during a campaign

Number of simultaneous users

Size of the datasets 

Frequency of the data streams

Size of the runs for pre, post and during runs

Sizes of metadata because of more provenance 

Size of the workflow tasks

Length (Time) of the experiment/ computation

 

 

Scott

 


From: Settlemyer, Bradley Wade <bws@lanl.gov>
Date: July 31, 2018 at 5:09:48 PM EDT
To: Mohror, Kathryn <mohror1@llnl.gov>, Ross, Robert B. <rross@mcs.anl.gov>, ssio-2018@lists.mcs.anl.gov <ssio-2018@lists.mcs.anl.gov>
Subject: Re: [Ssio-2018] notes from call

 

For me personally (which is slightly unfair since I am a facilities person), the differences between ALCF, OLCF, and NERSC are far less interesting than the differences between Open Science Facilities, NNSA, ESNet, and Instruments (a light source, a telescope, etc.).

 

To briefly describe how I’m thinking about it as a storage set of requirements:

  • NNSA – large simulation jobs that use experimental data sets, but aren’t coupled with experiments in real time
  • Open science facs – lots of users, diverse workloads, may couple with running experiments
  • ESNet – a completely different type of facility – no true storage requirements, but part of science workflows
  • Instruments – data acquisition pipelines, processing pipelines, and fast-feedback coupling

 

Part of my interest is I don’t understand what the instruments people desire to do versus what they feel like they can afford to do. These workflows still seem in their infancy with respect to data processing (I’m not saying they are unsophisticated). It’s also possible that NNSA is coupling simulations and I just don’t have that information.

 

  I felt like with the EOD emphasis in your problem proposals surprised me, and I think it may surprise others as well. These problems have lots of surface area, and while its probably hard to grapple with all of that surface, there might be some big wins available in those areas.

 

Cheers,

Brad

 

From: ssio-2018 <ssio-2018-bounces@lists.mcs.anl.gov> on behalf of "Mohror, Kathryn" <mohror1@llnl.gov>
Date: Sunday, July 29, 2018 at 2:43 PM
To: "Ross, Robert B." <rross@mcs.anl.gov>, "ssio-2018@lists.mcs.anl.gov" <ssio-2018@lists.mcs.anl.gov>
Subject: Re: [Ssio-2018] notes from call

 

On the facilities panel, should we have an NNSA facilities representative as well?

 

Kathryn

 

From: ssio-2018 <ssio-2018-bounces@lists.mcs.anl.gov> On Behalf Of Ross, Robert B.
Sent: Friday, July 27, 2018 1:37 PM
To: ssio-2018@lists.mcs.anl.gov
Subject: Re: [Ssio-2018] notes from call

 

Note to self, put a "respond by XXX" date on invites in the future. Just in case. -- Rob

 

On Jul 26, 2018, at 3:21 PM, Ross, Robert B. <rross@mcs.anl.gov> wrote:

 

Hi all,

 

Some quick notes from the call today:

 

Invitations. First 60-ish have gone out, per discussion on last call. So far we have about 20 accepts (including the organizers) and no declines. I'm updating the attendees spreadsheet in the top-level folder on Google drive.

 

Panels. We were talking about speakers, and we seem to have converged on two panels.

1. Facilities panel session. There's some sort of roll-up of requirements from applications (presented by one of us?), followed by a panel session with the facilities (ALCF, OLCF, NERSC, ESNet, anyone else?). We would need to iterate with the facility folks to make sure that they agree with the roll-up. Still needs a little thinking.

 

2. Tech panel session. This one might have short talks from each person followed by Q&A. Tim Feldman, Bruce Jacob, and Dmitry Katramatos were suggested as panelists, possibly a DB person also (Remzi or David Andersen as a possible person from the pool of existing invitees).

 

Please keep writing. We will revisit attendees next week and further discuss this panels idea.

 

Thanks,

 

Rob