21:02:08 <oneswig> #startmeeting scientific_wg
21:02:08 <openstack> Meeting started Tue Jan 24 21:02:08 2017 UTC and is due to finish in 60 minutes.  The chair is oneswig. Information about MeetBot at http://wiki.debian.org/MeetBot.
21:02:09 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
21:02:11 <openstack> The meeting name has been set to 'scientific_wg'
21:02:33 <oneswig> #link agenda for today is https://wiki.openstack.org/wiki/Scientific_working_group#IRC_Meeting_January_24th_2017
21:02:42 <oneswig> Hello!
21:02:44 <armstrong> it seem I am on the wrong channel? scientifig-wg?
21:03:06 <oneswig> Hi armstrong - you're in the right place for the Scientific WG
21:03:13 <oneswig> Just starting the meeting now
21:03:16 <armstrong> ok thanks
21:03:30 <hogepodge> o/
21:03:36 <oneswig> Martial are you here?
21:03:42 <oneswig> Hi hogepodge
21:03:43 <priteau> Hello
21:03:50 <oneswig> Hi priteau
21:03:51 <martial> Hi David, Stig
21:03:54 <trandles> Hi everyone
21:03:59 <martial> oneswig: fine thanks
21:04:05 <oneswig> #chair martial
21:04:05 <openstack> Current chairs: martial oneswig
21:04:09 <dfflanders> o/
21:04:11 * cdent lurks
21:04:15 <oneswig> Hi dfflanders
21:04:29 <oneswig> In fact, hi all, lets get this show on the road
21:04:43 <oneswig> #topic Reproducible Science Frameworks
21:05:02 <oneswig> So there was a little discussion earlier this week on the idea of reproducibility
21:05:18 <oneswig> Thought it might be interesting to hear the group's thoughts on how this is achieved
21:05:35 <oneswig> martial: want to provide some context on work at NIST?
21:05:47 <martial> oneswig: sure
21:05:49 <priteau> oneswig: link to the discussion if it was online?
21:06:11 <martial> there is a PhD student that is working on a model to create reproducable experiement results
21:06:21 <oneswig> priteau: alas it was in the semi regular "what shall we put into the agenda" mail...
21:06:27 <martial> https://mgi.nist.gov/cloud-reproducible-records
21:06:54 <martial> and he presented his tool to me last week
21:07:10 <priteau> martial: this page doesn't load for me
21:07:10 <martial> the idea is that you have an API to allow you to create records of runs
21:07:26 <oneswig> martial: I'm having trouble with it too - https issues
21:07:26 <martial> (try without the s ?)
21:07:28 <armstrong> the page is broken
21:07:34 <trandles> ditto
21:07:42 <priteau> SSL error
21:07:49 <priteau> http redirects to https
21:07:56 <martial> great it worked yesterday :)
21:08:16 <martial> I will have to talk to the student on this one
21:08:35 <priteau> text-only google cache copy: http://webcache.googleusercontent.com/search?q=cache:RRp1-4OeX04J:https://mgi.nist.gov/cloud-reproducible-records&num=1&hl=en&gl=uk&strip=1&vwsrc=0
21:08:36 <armstrong> is reproducibility here same as replication?
21:08:43 <oneswig> The Demo Effect is strong with this one...
21:08:59 <martial> armstrong: not entirely
21:09:07 <armstrong> ok
21:09:16 <oneswig> armstrong: I think the idea is that results can be regenerated by others (or indeed by the same person twice)
21:09:26 <oneswig> To get independent corroboration of research
21:09:38 <martial> oneswig: exactly
21:09:48 <martial> (sorry was getting int touch with student)
21:10:11 <oneswig> ... and the opportunity here is that cloud techniques enable us to do this in more effective ways
21:10:42 <armstrong> i got you
21:10:47 <martial> oneswig: yes I am trying to see about setting up a collaboration with the student to test on our stack
21:10:51 <martial> among this discussion we talked about a few tools of interest for such tasks
21:11:08 <martial> and part of this effort was mentioning a few of those
21:11:12 <oneswig> martial: I circulated the links among the group in Cambridge and apparently there is a data-centric approach evolving there: https://www.cl.cam.ac.uk/research/dtg/fresco/
21:11:33 <priteau> martial: We have got a similar requirement for Chameleon. It's great to see that you're working on it!
21:11:47 <martial> priteau: yes I remember
21:11:55 <oneswig> priteau: did you have a specific approach in mind?
21:12:42 <martial> so a couple links of tools for people who share this interest: https://pythonhosted.org/Sumatra/ https://github.com/ViDA-NYU/reprozip
21:12:53 <priteau> oneswig: I was thinking about something more integrated with OpenStack: describe your experiment as a kind of workflow with Heat stacks, which would include some helpers to persist results somewhere (e.g. Swift)
21:13:33 <priteau> martial's approach has the advantage of being portable across clouds
21:13:43 <martial> as well as CDE http://www.pgbovine.net/cde.html and Singularity  http://singularity.lbl.gov/about
21:13:53 <priteau> martial: thanks for the links
21:14:15 <oneswig> Might tie in well with jmlowe and rbudden's work on using Murano to manage application images without storing away entire qcows
21:14:28 <trandles> https://github.com/hpc/charliecloud
21:14:31 <jmlowe> ooh, I like that
21:14:34 <trandles> throwing charliecloud in the mix too
21:14:44 <martial> If there is some interest, I will try to get the student to join a meeting
21:14:44 <oneswig> Aha, as if by magic :-) Hi jmlowe
21:15:03 <jmlowe> don't mind the smell of brimstone
21:15:05 <martial> he can explain his work far better than I can :)
21:15:29 <oneswig> martial: would be good to follow up I think, it's a relevant issue
21:15:34 <martial> but let us just say we have an interest in the subject and wanted to judge interest
21:16:07 <martial> oneswig: okay putting it out there (because tons of topic for today) and we can follow up
21:16:36 <oneswig> martial: can you aggregate the links (eg for holding on the wiki for data?)
21:16:39 <jmlowe> more or less you can think of murano as a service wraps up heat templates and presents them in point and click launch catalog in horizon
21:16:45 <martial> oneswig: yes
21:16:57 <oneswig> Thanks martial :-)
21:17:33 <oneswig> jmlowe: I may be on your case for Murano experience in the coming weeks, we are starting in that direction and would love to benefit from your experience
21:18:05 <oneswig> OK, move on?
21:18:24 <oneswig> #topic Scientific OpenStack at SC 2017
21:18:39 <jmlowe> reproducing experiments is not getting the same results on the exact same lab equipment but rather following the same steps to construct similar lab equipment and getting similar results
21:18:54 <oneswig> The SC2016 show seemed to go well with plenty of interest
21:19:03 <oneswig> How to build on that and take it to the next level?
21:19:03 <jmlowe> +1
21:19:20 <oneswig> dfflanders and hogepodge I think have an idea
21:19:36 <oneswig> Which was to do a hands on workshop
21:20:07 <hogepodge> oneswig: what would be the content?
21:20:24 <oneswig> Ah, well here's a worthy discussion
21:20:38 <jmlowe> rbudden was getting crazy right before barcelona with a one click openstack deployment inside a vm, maybe I can get a piece of jetstream and everybody can stand up their own openstack?
21:21:02 <oneswig> One idea could be to do something like a clinic for optimisation / integration of HPC
21:21:14 <cdent> jmlowe++
21:21:25 <martial> oneswig +1
21:21:27 <oneswig> jmlowe: also sounds good to me for a structured introductory session
21:22:01 <jmlowe> yeah, that's more of a tutorial idea
21:22:05 <hogepodge> so, we would be focusing on infrastructure over applications?
21:22:12 <oneswig> hogepodge: for such a session, where could we get infrastructure?
21:22:17 <trandles> oneswig: I think optimization/integration would be generate a lot of interest, especially when anything "cloud" as not being very performant
21:22:25 <hogepodge> It seems like there could be two proposals, one for operators one for users
21:22:30 <oneswig> hogepodge: that might only be first prejudice but most of us here are that way inclined...
21:22:44 <hogepodge> oneswig: one of the partner universities, or possibly OSIC
21:23:16 <hogepodge> OSIC is sometimes available as a tenant cloud for short term projects. We can work out other options too
21:23:18 <dfflanders> lots of public clouds would be willing to provide free credits
21:23:25 <dfflanders> for training
21:24:03 <oneswig> OSIC has the advantage of regularly exposing the underbelly of infrastructure to the great unwashed
21:24:29 <oneswig> I'd be very unpopular doing that on the systems at work :-)
21:25:06 <oneswig> dfflanders: that would work for cloud applications.  I think the discussion's diverging onto infrastructure and application paths
21:25:39 <martial> dfflanders: if only there was a public scientific cloud for all ... :)
21:25:41 <jmlowe> would chameleon be available this far out?
21:25:49 <hogepodge> OSIC probably wouldn't work for the infra part, just because the RAX and Intel teams do the provisioning for those systems
21:26:03 <martial> I take it trystack is out too ?
21:26:14 <oneswig> Another possibility might be "bring your own cloud" - ie for people to bring issues for WG members to hack on
21:26:48 <dfflanders> martial, ask me again in 9 months ;-)
21:26:59 <priteau> jmlowe: depends how much resources we're talking about, it may be a possibility
21:27:02 <hogepodge> bringing own cloud is challenging, but one option might be to get some basic hardware like NUCS that are easily transportable and can mimic a data center
21:27:07 <oneswig> hogepodge: au contraire - isn't OSIC all about testing infrastructure?  Or are there two parts to it
21:27:11 <jmlowe> you could fill the gaps with "I'm ____ and this is how I work on my cloud" lighting talks
21:27:35 <hogepodge> oneswig: it has the bare metal cloud allocations and the tenant cloud. I was thinking the tenant cloud.
21:27:52 <oneswig> ah ok - I've only ever used the former
21:27:58 <hogepodge> oneswig: I guess it couldn't hurt to submit a request and see if it could be available for a workshop as an installation exercise
21:28:24 <martial> SC is in 9 months, we can talk to Intel in Boston too
21:28:45 <oneswig> hogepodge: what's the harm in asking, eh?
21:29:06 <jmlowe> tell them we are considering pine64's as an alternative
21:29:09 <oneswig> The reason for discussing it now is that there's a deadline for workshop submission - 7th Feb
21:29:22 <martial> jmlowe: or the new asus board :)
21:29:38 <martial> oneswig: I meant for bring your own cloud
21:30:22 <jmlowe> I'm fairly confident we can do the booth talk thing again at our booth as long as we are talking sc17 stuff
21:30:57 <oneswig> jmlowe: thanks.  BTW my son's basketball hoop has broken, where do I send it for refund :-)
21:31:31 <jmlowe> good question, sc'16 was the first booth we didn't do ourselves
21:31:47 <oneswig> So I gauge there's a good deal of interest in a workshop session (of some form)
21:32:15 <jmlowe> +1
21:32:19 <oneswig> Or perhaps a day of two halves (infrastructure then cloud apps)?
21:32:28 <hogepodge> oneswig: +1 to two halves
21:33:02 <oneswig> hogepodge: have you done this kind of workshop before?
21:33:56 <oneswig> I think there are also viable alternatives for finding infrastructure / instances for lab resource
21:34:06 <hogepodge> oneswig: I taught an intro to openstack seminar once, many years ago. I've given some demos, though, and have some materials ready for doing an installation using Ironic and Kolla
21:34:30 <oneswig> hogepodge: saw that at the Barcelona keynote, nice job
21:34:33 <hogepodge> I don't have the expertise for tuning that I'm sure others here have, though
21:35:05 <hogepodge> (fwiw I'm also attending the board meeting right now so I may be a bit distracted)
21:35:20 <hogepodge> dfflanders: how would you feel about helping with app dev content?
21:36:32 <oneswig> (he's taking a long time to type out 'yes'...)
21:36:59 <hogepodge> My proposal for right now is to get concrete ideas down to start formulating a proposal
21:37:07 <hogepodge> I set up an etherpad
21:37:09 <hogepodge> #link https://etherpad.openstack.org/p/SC17WorkshopWorksheet
21:37:18 <oneswig> hogepodge: great, thanks
21:37:21 <jmlowe> I'll go ahead and volunteer some significant number of instances on jetstream for either part or both
21:37:52 <hogepodge> jmlowe: +1 :-D
21:37:59 <armstrong> can you also mention some usecases?
21:38:16 <martial> FYI: https://mgi.nist.gov/cloud-reproducible-records works again
21:38:33 <oneswig> yes it does!
21:39:06 <armstrong> yes web site working now
21:39:17 <oneswig> armstrong: interesting question.  HPC cloud apps - ones that actually use cloud technologies rather than run unaware of them - any thoughts?
21:39:58 <oneswig> There's some good stuff in the bioinformatics world around portals
21:40:06 <armstrong> sure
21:40:18 <dfflanders> hogepodge, yip
21:40:23 <jmlowe> I met with these guys last week, they were doing some very interesting "wrap your code and run on cloud" stuff, http://genapp.rocks/
21:41:35 <oneswig> It might be interesting to look at deployment of "application topologies" - heterogeneous pipelines effectively
21:41:53 <martial> is this still SC17 ?
21:42:05 <oneswig> martial: yes - the workshop ideas (apps this time)
21:43:26 <martial> should we just keep iterating on the etherpad for interested party; we have two topics left to discuss
21:43:27 <jmlowe> time is getting a little short for both the meeting and the submission, should we solicit interested parties and continue oob?
21:43:47 <oneswig> martial: jmlowe: yes to both
21:43:56 <oneswig> #topic Boston planning
21:44:05 <martial> reminder: https://etherpad.openstack.org/p/SC17WorkshopWorksheet
21:44:33 <oneswig> The major question here was how to organise WG activities at the summit
21:45:26 <oneswig> And the Boston Cloud Declaration activities on policy-level intercloud federation: would that enrich the summit or over-egg it to be held in the same concourse
21:45:45 <hogepodge> enrich
21:46:01 <hogepodge> (from osf point of view)
21:46:02 <martial> enrich but the concern was the length needed to discuss its content
21:46:08 <jmlowe> either way, it certainly simplifies travel arrangements
21:46:14 <oneswig> #link Background context https://wiki.openstack.org/wiki/Boston-Cloud-Congress
21:46:37 <trandles> is BCC as a follow-up the next week still up for discussion?
21:46:53 <trandles> or is it scheduled for 15-16th?
21:47:08 <oneswig> I agree on co-scheduling, although there's an EGI event in Europe that conflicts with the summit, which is very unfortunate
21:47:16 <oneswig> dfflanders: what's the latest?
21:47:34 <martial> according to the link: "Venue being discussed (98ppl) with MIT for 'day and half event' for May 15th and part of May 16th, 2016."
21:48:05 <martial> wait we are in 2017 are we not
21:48:09 <martial> ok disregard
21:49:30 <martial> no wait it says May 15/16 in the header, so it is actual I guess
21:49:46 <trandles> lol exactly...I'm confused now
21:49:48 <hogepodge> dfflanders: and I can take space requests to our events and management team, and see how the congress could fit into the schedule. It's definitely something that's been on our minds so it wouldn't be a shock
21:49:57 <oneswig> I don't know if it has been agreed either way.  I think Flanders may be elsewhere - dfflanders?
21:50:14 <hogepodge> We just need a solid request, so space and time requirements
21:51:12 <oneswig> OK so lets cover the direct WG activities
21:51:32 <oneswig> Last time, there was a committee meeting, a BoF and an evening social
21:51:55 <oneswig> The feedback was that the committee meeting should be bigger than 30 - noted
21:52:06 <martial> and in a bigger room :)
21:52:38 <oneswig> The discussion from a couple of weeks back on BoF centred on the idea of two sessions, back-to-back possibly, right?
21:52:54 <martial> oneswig: that is also my recollection
21:53:52 <martial> oneswig: there was also the issue/question of the BCC as part of the SWG meetings, but we decided it needed it own time
21:53:59 <oneswig> I don't recall the room sizes for the BoF but it was pretty big - 100 attendees approx?
21:54:45 <oneswig> martial: agreed, I think it's a discussion at a higher level
21:55:05 <martial> the BoF was well attended and we had (in Barcelona) one of the bigger tutorial room, which did help
21:55:29 <oneswig> OK, we should put in for a double session of space for the BoF.  What else, anyone?
21:56:07 <oneswig> Final item was a social.  Any Bostonians in the room?
21:56:13 <martial> I have an item that might be of interest for the BCC, but I will email people directly for now to discuss interest
21:56:44 <oneswig> Last summit, 57 people bought tickets - something about that size.  Any thoughts?
21:57:43 <jmlowe> seems like we should go to an oyster bar, but my nightmare is having food poisoning on a plane
21:57:43 <oneswig> I haven't been there for 15 years, my local knowledge may be too dated...
21:58:15 <oneswig> Union Oyster House was the choice downtown - it's a bit of a warren inside though.
21:58:28 <oneswig> More of a lobster place nowadays anyway :-)
21:58:54 <martial> oneswig: might have to have options too for people who do not eat oysters :)
21:59:09 <oneswig> OK, we might need to widen it to anyone who knows a Bostonian...
21:59:43 <trandles> I know two but I'm not sure I trust their judgement
21:59:55 <oneswig> martial: agreed.  Wasn't equal with the Barcelona choice!
22:00:18 <oneswig> We'll need to defer this discussion, alas.  the thief of time strikes again
22:00:19 <fungi> there are some awesome italian-american restaurants in the "little italy" area of town
22:00:36 <oneswig> fungi: thanks, now you're making me hungry!
22:00:37 <martial> darn, time flies
22:00:47 <oneswig> OK, got to clear the channel
22:00:56 <oneswig> Thank you everyone - and welcome to the WG armstrong
22:01:00 <trandles> bye
22:01:01 <martial> thanks all
22:01:05 <oneswig> #endmeeting