21:00:32 <martial__> #startmeeting Scientific-SIG
21:00:33 <openstack> Meeting started Tue Oct 16 21:00:32 2018 UTC and is due to finish in 60 minutes.  The chair is martial__. Information about MeetBot at http://wiki.debian.org/MeetBot.
21:00:35 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
21:00:37 <openstack> The meeting name has been set to 'scientific_sig'
21:00:44 <mrhillsman> o/
21:00:52 <oneswig> hello!
21:01:11 <oneswig> #link agenda for today https://wiki.openstack.org/wiki/Scientific_SIG#IRC_Meeting_October_16th_2018
21:01:44 <oneswig> It's somewhat short today
21:02:34 <martial__> (shoot too late was looking for that :) )
21:02:45 <oneswig> martial__: are you travelling today?
21:02:51 <b1air> o/
21:02:54 <janders> g'day - sorry for being late
21:03:02 <martial__> #topic Topics for the agenda for Berlin
21:03:16 <trandles> I'm here, sorta
21:03:24 <oneswig> hey b1air janders trandles
21:03:36 <b1air> I was an hour early :-(
21:03:38 <martial__> stig: nope am talking to a colleague at NIST who is looking at deploying an OpenStack for Public Safety type work
21:03:57 <martial__> am trying to get him to meet people and come to Berlin :)
21:04:04 <oneswig> b1air: always the keener
21:04:27 <martial__> Will introduce him after, maybe he can setup a couple meetings with people to see how to best do his work :)
21:04:29 <b1air> Not sure which timezone I'm in, but it's not the same as my calendars
21:04:30 <oneswig> martial__: be very happy to meet them
21:04:42 <oneswig> b1air: you're in Sydney?
21:04:49 <b1air> Melbourne
21:05:16 <oneswig> nice!
21:05:27 <oneswig> So - we should get on track I guess.
21:05:33 <janders> yeah daylight savings made the evening session tricker, but the morning session is so much easier now :)
21:05:38 <martial__> (see guest speaker and all that :) )
21:06:07 <oneswig> We have SIG sessions in the schedule now
21:06:10 <martial__> So team, the dates have been set
21:06:16 <martial__> Stig, you go :)
21:06:28 <oneswig> They are not consecutive, but at least on the same day (and in the right order)
21:07:04 <oneswig> Very happy to see it, in the back of my mind I had a nagging feeling I needed to jump through one more hoop to get the paperwork in order :-)
21:08:14 <martial__> have you thought of your co-hosts? :)
21:08:31 <oneswig> martial__: fancy a video link? :-)
21:08:59 <oneswig> I hadn't thought of co-hosts tbh
21:09:01 <martial__> if it is not at the same time as the SC18 panels/BoF or similar, I am happy to participate if I can
21:09:25 <oneswig> Dallas - I think it may be 5am...
21:09:40 <martial__> For you, sure
21:09:56 <martial__> Which reminds me, the dates in questions are
21:10:01 <martial__> https://www.irccloud.com/pastebin/Cdz6bL1D/
21:10:14 <martial__> SIG Meeting Wed 14, 11:00am - 11:40am
21:10:14 <martial__> SIG BoF Wed 14, 1:40pm - 2:20pm
21:10:14 <martial__> SIG Social Wednesday evening, TBD
21:10:23 <martial__> (that was interesting)
21:10:57 <trandles> those times are all CET?
21:11:37 <oneswig> As in, Central European Time, I am not sure what they do for DST in Germany
21:12:00 <janders> I just looked it up, in November it should be CET
21:12:12 <janders> summer would be CEST
21:12:14 <trandles> US switches to standard time night of November 3-4
21:12:40 <oneswig> Our summer is not so long and neither is our summer time :-)
21:12:53 <trandles> so Berlin should be 7 hours ahead of Dallas
21:13:23 <trandles> I guess we'll see just how dedicated martial__ is ;)
21:13:30 <oneswig> hmmm. I know martial__ is a night owl but that's maybe too far!
21:14:06 <martial__> 5am ... no prob-zzzzzzzz
21:14:07 <janders> Berlin (Germany - Berlin) Wednesday, 14 November 2018 at 11:00:00 am CET UTC+1 hour   Dallas (USA - Texas)      Wednesday, 14 November 2018 at 4:00:00 am  CST UTC-6 hours  Corresponding UTC (GMT)   Wednesday, 14 November 2018 at 10:00:00
21:14:16 <oneswig> b1air: martial__: I shall expect a video message as minimum (recorded in some Dallas bar, perhaps)
21:14:25 <b1air> Just need to find a bar that stays open 24 hours ;-)
21:14:28 <martial__> I can set up the GoToMeeting link even ;)
21:14:36 <martial__> actually why not ...
21:14:53 <janders> Berlin (Germany - Berlin) Wednesday, 14 November 2018 at 2:00:00 pm CET UTC+1 hour   Dallas (USA - Texas)      Wednesday, 14 November 2018 at 7:00:00 am CST UTC-6 hours  Corresponding UTC (GMT)   Wednesday, 14 November 2018 at 13:00:00
21:15:09 <janders> at least BoF falls within more civilised hours
21:15:27 <martial__> Sitg, let's talk after the end of the meeting
21:15:54 <martial__> Btw, Jim said he can speak about his cloud approach to Public Safety, so we have a guest speaker :)
21:16:18 <oneswig> sounds good.  The final matter on the SIG events is that Nick from our team has offered/been conscripted to get us a beer social on Wednesday evening.
21:16:26 <oneswig> It is Germany, after all :-)
21:16:40 <oneswig> martial__: That Jim Golden?
21:16:48 <martial__> but you got there in "November"
21:16:59 <martial__> another Jim
21:17:13 <martial__> the first Jim will be in Germany btw
21:17:14 <oneswig> Good to have him, all the same
21:17:27 <janders> oneswig: subtle hint: how about a venue with craft beer + open fire roasted meat? :)
21:17:36 <janders> always a highlight at ISCs
21:17:52 <janders> unfortunately I don't know Berlin at all
21:17:56 <oneswig> janders: seems like you've done this before :-)
21:18:03 <martial__> surprised Jim (Lowe) did not jump on this one ;)
21:18:09 <janders> yeah but only in Frankfurt and Leipzig
21:18:12 <trandles> Mike Lowe?
21:18:40 <trandles> sounds like janders is volunteering to organize the outing, good man
21:18:49 <martial__> yes Mike (is it already that late? :) )
21:19:01 <martial__> yeah janders!!!!
21:19:27 <janders> trandles: I don't know Berlin well enough (or at all) so just hinting the idea :)
21:19:32 <janders> can ask around though
21:19:52 <oneswig> Berlin attendees, I'll be receptive to any google-based suggestions of local places on the theme of das-beer-and-roasted-meat-haus
21:20:16 <oneswig> (although at pains to point out that other dietary requirements must also be met)
21:20:22 <trandles> janders: I'll be in Dallas unfortunately :(  Definitely be at the Denver Summit in the spring though...
21:20:40 <martial__> asked my German friend Lukas for recommendations :)
21:20:52 <janders> martial_: good man! :)
21:21:08 <oneswig> martial__: couldn't ask for more than that.  Provided Lukas is not a monk.
21:21:40 <oneswig> BTW how are your panel questions coming on for SC?
21:21:41 <martial__> nope
21:21:57 <martial__> I worked on the "panel" questions, not the "BoF" ones yet
21:22:03 <martial__> need to get this started
21:22:17 <martial__> will start a Google Doc or Etherpad and share it here
21:22:51 <oneswig> This came up among our team this week and may be interesting for your crowd: https://kubernetes.io/blog/2017/08/kubernetes-meets-high-performance/
21:23:20 <martial__> Lukas via chat recommend looking at https://awesomeberlin.net/attractions/biergarten-berlin-best-beer-garden
21:24:03 <martial__> Jim is next to me btw, so when we get to AOB, I will give him my keyboard
21:24:38 <oneswig> I get the impression Docker's runtime is getting squeezed out of the Kubernetes ecosystem - lots of energy going into projects like podman and cri-o
21:25:06 <janders> oneswig: VERY interesting link
21:25:12 <janders> thanks for sharing
21:25:13 <oneswig> Might be a good opener for Christian :-)
21:25:22 <martial__> (ouch)
21:25:37 <janders> on the SIG sessions front - I did have a quick chat to Bright last week
21:26:15 <janders> as it stands, they don't seem to be targetting running on top of a bare-metal OpenStack, they (still) insist on running their own OpenStack (and I don't think they are far with baremetal on that)
21:26:56 <janders> I wonder if you guys see value in trying to get some HPC cluster management solution vendors to join us, so that they get a better understanding of what we're doing?
21:27:21 <janders> I can ping Bright again asking if they have anyone at the Summit - we could also maybe try with others?
21:27:48 <oneswig> janders: that is always good.
21:28:05 <janders> we have (or have access to) four systems - two are bright, one is SGI stack and one is a Cray
21:28:17 <janders> so Cray would be my second thought
21:28:26 <oneswig> The Cray's probably half Bright anyway?
21:28:38 <janders> it's hard to keep up with them
21:28:52 <janders> two years ago they were going all-OpenStack, now they backed out
21:28:55 <oneswig> Much of the management infra is deployed with that.
21:29:08 <trandles> AFAIK is Cray even doing any OpenStack any more?
21:29:14 <janders> they are making crazy changes to their interconnects (details - NDA but you likely heard)
21:29:29 <janders> trandles: last time I heard they admitted failure and dropped the ball
21:29:36 <trandles> janders: +1
21:30:18 <janders> I hope maybe if they listen to us talking they'll realise what they're missing out on
21:30:20 <oneswig> trandles: no
21:30:36 <janders> and we'll be fine without them, but it would be good to have more options for running HPC on OpenStack
21:31:02 <janders> ok.. I will ping Bright
21:31:11 <janders> do you guys know anyone at Cray that could be interested?
21:31:23 <janders> I only talk to salesy people on their end unfortunately
21:31:44 <bollig> I know a couple guys across the river that used to work on OS at Cray. I’ll ask
21:31:47 <oneswig> janders: I know some people in Europe but I don't think they will be in Berlin - I can ask
21:32:00 <janders> cool!
21:34:01 <janders> I dropped Martijn a quick note, let
21:34:10 <janders> 's see if anyone from Bright is heading to Berlin
21:34:43 <janders> how about SchedMD? Do you know if they'll have any presence in Berlin?
21:35:09 <trandles> janders: highly doubtful since SC is the same week
21:35:17 <janders> very fair point
21:35:58 <oneswig> I've pinged my contacts in Cray but suspect it'll be a similar story
21:36:07 <trandles> does IBM have any openstack things going on?  You might use that as a way to get in touch with LSF folks.
21:36:24 <trandles> although since IBM != x86 any more, I doubt it
21:36:41 <janders> all I know is that their GPFS integration is struggling to keep up with the new OpenStack releases
21:36:47 <janders> which isn't a good sign
21:37:23 <trandles> that basically leaves Adaptive (torque/moab) and Altair (PBS)
21:37:41 <trandles> I wouldn't be surprised if Altair doesn't have something going on.
21:37:41 <janders> RHATs are doing some interesting work with Power and Ironic here in Canberra, but that's completely non-HPC
21:37:56 <trandles> s/wouldn't/would/
21:40:05 <martial__> seems we are slowing down on this topic
21:40:16 <martial__> should I get Jim Horan to introduce himself?
21:40:26 <oneswig> go for it
21:40:28 <janders> please
21:44:21 <martial__> Hi my name is Jim Horan. I am working at NIST in Gaithersburg, MD in the area of public safety analytics. We are currently working with NIST Public Safety Communications Research Lab in our Bolder facility to further the use of analytics by public safety stakeholders and first responders. One of the areas we have been investigating is having HPC capability at the edge for use by public safety operators. We are
21:44:21 <martial__> also looking at cloud-based analytics and hybrid cloud capabilities to optimize the distribution of high and low computational intensity analytics.
21:45:29 <martial__> We are funding several research grants to develop analytics as well as internal projects to support reference frameworks to test and evaluate analytic performance.
21:46:43 <martial__> We are currently developing a stack to support this reference capability and Openstack has been an area of interest to support our research.
21:47:05 <oneswig> Sounds good.  There was a keynote by AT&T (IIRC) on a similar theme at the last summit.  Did you see that?
21:47:34 <martial__> No, but if I can track it down, I'll check it out.. Thanks!
21:48:36 <oneswig> It was a movie clip, started with a car going off a bridge
21:50:12 <oneswig> It will be very interesting to hear about the use case - edge computing work is a hot subject these days!
21:50:48 <martial__> Ah.. I think somebody did tell me about that movie clip.  Our PS folks have been working to refine requirements of use cases (such as car or person off a bridge) and put them against state of the art development of analytics to see what can be done in the near term.. and what capabilities are required to support them.
21:51:58 <janders> few years back, we built a POC where we had a small HPC OpenStack in Singapore which would also manage resources around the world
21:51:59 <martial__> we have been keeping a close eye on the progress in edge computing.. since with the onset of 5G the edge can go from being a group or cluster of functionality to a fabric
21:52:29 <janders> controllers + few computes in Singapore (along with central storage ) and then more computes in Australia, Europe, US
21:53:03 <janders> worked reasonably well, was quite interesting. not sure if this is directly relevant to your use case but thought I will mention in case it is
21:53:27 <martial__> how did the POC hold up in terms of robustness with all the variables of comms/network etc?
21:53:45 <oneswig> Jim would you be able to talk on this some more with the SIG in Berlin?
21:54:27 <martial__> Martial has been trying to get me to talk about it. I am checking to see if I can make it out.. but I would be glad to talk more about it.
21:54:28 <janders> it was quite solid. wan setup was different for different sites and those without QoSed circuits would sometimes cause trouble. Those with solid QoS were pretty good.
21:55:30 <oneswig> janders: sounds like an idea ahead of its time
21:55:38 <martial__> did you do any network active adaptation based on the QoS changing? I think this will become super important as the edge clusters start to load balance etc
21:56:10 <janders> martial_: nope
21:56:47 <janders> we would have different host aggregates mapping to AZs and depending on the requirements the user could spin up resources within a specific site, or scattered around the World
21:57:30 <janders> but there wasn't any other form of load balancing or placement optimisation
21:57:35 <martial__> sometimes not the luxury of having spare capacity/circuits..   that sounds like an interesting system
21:59:19 <martial__> (and we are reaching the end time wise :) )
21:59:25 <martial__> (taking back my keyboard)
21:59:50 <martial__> so I am trying to get Jim to come to the Berlin Summit :)
21:59:58 <janders> +1! :)
22:00:14 <oneswig> martial__: be good to see him!
22:01:27 <martial__> will try to get him to come for sure :)
22:01:44 <oneswig> OK we should wrap up for today
22:01:56 <martial__> and on that, thanks all for attending our Scientific SIG meeting
22:02:02 <janders> thanks all! have a good day / good night :)
22:02:12 <oneswig> Cheerio
22:02:16 <oneswig> #endmeeting