21:01:43 <martial> #startmeeting Scientific-SIG 21:01:44 <openstack> Meeting started Tue May 15 21:01:43 2018 UTC and is due to finish in 60 minutes. The chair is martial. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:01:45 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:01:47 <openstack> The meeting name has been set to 'scientific_sig' 21:01:54 <martial> #chair oneswig 21:01:55 <openstack> Current chairs: martial oneswig 21:01:57 <jmlowe> Isn't it like 22:00 there? 21:02:11 <martial> Good day everyone 21:02:14 <oneswig> Hi jmlowe - it is... 21:02:21 <jmlowe> I guess late dinners are the norm 21:02:21 <martial> Hi ike 21:02:28 <martial> Mike 21:02:37 <oneswig> ... only for the disorganised! 21:02:38 <jmlowe> Hello 21:02:43 <martial> #link https://wiki.openstack.org/wiki/Scientific_SIG#IRC_Meeting_May_15th_2018 21:03:05 <martial> here is our agenda (pretty much a repeat of last week) 21:03:19 <martial> #topic SIG Cycle Report 21:03:37 <martial> #link https://etherpad.openstack.org/p/scientific-sig-report-queens 21:04:16 <martial> inviting people to add to this document if they did do a presentation related to the work of the SIG 21:04:27 <martial> (thanks Mike, I know you added quite some content there already) 21:04:54 <jmlowe> All of our EOT activities count I think 21:05:58 <martial> yes it seems to me they do :) 21:06:02 <rbudden> hello 21:06:19 <oneswig> I've added some items (sure I could rack my brains for more but the list's looking pretty good) 21:06:27 <martial> Hi Robert :) 21:06:39 <martial> yeah the list is starting to be good indeed 21:06:50 <rbudden> apologies for missing a few of the last meetings 21:06:57 <oneswig> Hi rbudden - no problem 21:07:09 <rbudden> the time shift has put the meeting in the middle of daycare pickup ;) 21:07:49 <oneswig> That can be a wild time... 21:07:59 <martial> no worries, glad you can make it this time 21:08:28 <martial> Mike, Robert, we will likely want to talk about our "HPC using OpenStack" session of next Wednesday 21:08:37 <martial> but we can take it offline 21:08:48 <oneswig> (don't mind me...:-) 21:08:54 <martial> I expect Stig and Blair to join on the panel as well if they can join 21:09:11 <martial> (the system only let me do 3 people) 21:09:14 * oneswig goes to check on his pie 21:09:15 <jmlowe> We can meet about it Sunday, it's my excuse for flying in Saturday and totally not so I can go sight seeing 21:09:50 <jmlowe> I was kind of anticipating much more of a open discussion with the audience 21:10:08 <martial> same here, I listed it as a panel 21:10:09 <rbudden> ditto 21:10:18 <martial> and it shows as a panel 21:10:21 <jmlowe> inverse pannel 21:10:31 <rbudden> I feel like the last time we talked too much and could have had much better audience participation 21:10:47 <rbudden> just MHO though 21:11:15 <martial> we can think about it and make it better for us I am sure 21:11:31 <oneswig> Will you run an etherpad or have people with roving mikes in the audience? 21:13:02 <martial> you know we can still edit the descritpion 21:13:06 <jmlowe> This audience can probably handle a etherpad, unlike SC 21:13:07 <martial> maybe add an etherpad 21:13:20 <martial> to the panel description and that will help 21:13:26 <rbudden> jmlowe: +1 21:13:43 <martial> jmlowe: touche :) 21:14:03 <jmlowe> Suggestion of etherpads had the SC audience looking at us like we were aliens 21:16:01 <martial> started one https://etherpad.openstack.org/p/yvr18-hpc-using-openstack 21:17:51 <oneswig> I'll be interested to see what you guys come up with. Are you going to cover infrastructure or science workloads, or just see where the discussion goes? 21:18:42 <rbudden> i would imagine some talk of both. I know recently I’ve been in new territory for PSC regarding both. 21:18:48 <martial> #link https://www.openstack.org/summit/vancouver-2018/summit-schedule/events/20656/hpc-using-openstack 21:19:24 <martial> I was going to do a very small "HPC multimedia processing" into as a lighting talk as well as an intro to our work for the panel 21:20:20 <martial> to try to introduce my side (as a non large university) 21:20:54 <oneswig> martial: surely you've got war stories that involve actual war :-) 21:21:16 <rbudden> For better or worse we’ve delved into Windows Server VMs in OpenStack and are currently hosting a 1.5TB RAM VM for a CMU teaching workload/use case 21:21:49 <rbudden> I can atest to the war stories in the above regard ;) 21:21:56 <martial> sure upgrading openstack :) 21:22:21 <martial> oh yes same here, we have a couple people who wanted support for that ... sigh 21:22:55 <rbudden> glad I’m not the only one. We’ve had a recent influx of a lot of these requests 21:23:48 <oneswig> Can we cover the committee meeting? I think in Sydney we ought to have come in with more proposals to get the conversation started... 21:24:58 <martial> when is that? 21:25:22 <oneswig> The one on Wednesday morning. I'm not sure what it's meant to be called, but the one that isn't the BoF 21:25:43 <martial> oh I thought you meant the Sunday afternoon meeting' 21:26:15 <oneswig> no - sorry - confusing 21:26:35 <martial> yeah sure 21:26:53 <martial> if you look in your emails, I just sent a request to update the "verbiage" for it 21:27:01 <martial> (and add us as modertors) 21:27:11 <oneswig> What I'm interested in is ideas on what is deficient in OpenStack for research computing, and where (as a sig) we could set some activity areas. 21:27:28 <martial> let's start an etherpad for it as well? 21:27:29 <oneswig> Saw your mail martial - thanks, that needed doing. 21:27:42 <oneswig> martial: good idea 21:28:03 <martial> Jimmy is likely not happy with me, sent him like 3 or 4 emails in the past couple days about things to fix on the agenda 21:28:23 <martial> Mike, Robert: the HPC on OpenStack is moved to Wednesday afternoon btw 21:28:37 <oneswig> I'm sure it's a drop in the ocean of his inbox the week before the summit! 21:28:45 <martial> #link https://etherpad.openstack.org/p/scientific-sig-vancouver2018-meeting 21:28:53 <jmlowe> right 21:29:14 <rbudden> martial: thx, i wasn’t aware of the change! 21:29:23 <martial> conflicts 21:29:34 <jmlowe> Don't worry, I'll keep you in line 21:30:16 <martial> I got an email: you have a session from 8-5 Thu and you have a session from 10-11 Thu ... a scheduling snafu :) 21:31:19 <oneswig> martial: will you do a lightning talk update on ORCA at the BoF? 21:31:42 <martial> I requested for it to be Tuesday (so we could point people to our Wed sessions) but that was a no go 21:31:54 <martial> yes 21:32:08 <martial> #link https://etherpad.openstack.org/p/scientific-sig-vancouver2018-lighting-talks 21:32:59 <martial> I need to email our [scientific] friend with the info once the schedule is updated 21:33:21 <martial> which reminds me 21:33:28 <martial> #topic Call for Lighting Talks 21:33:54 <oneswig> martial: I think there is already an etherpad for lightning talks. 21:34:21 <martial> as always if you are able to join the Scientific SIG session on Wednesday and want to present a lighting talk on your work, see link above 21:34:29 <martial> oneswig as far as I know that is that link 21:35:14 <oneswig> weird, I'm sure there was a pad with ~3 on from a couple of weeks ago 21:36:33 <oneswig> I think I'm mixing it up with forum topics 21:37:31 <martial> if you find the info please follow up with Jimmy, I will email him the links soon to update the agenda 21:38:28 <martial> so far those are the topics I had listed for our meeting 21:38:43 <martial> there is also the dinner if we want to discuss it 21:38:47 <martial> should we announce it? 21:39:47 <oneswig> Well I think we are all going ... 21:40:10 <martial> very likely ;) I guess this solves this topic :) 21:40:23 <martial> remaining topic added to the pile 21:40:29 <martial> #topic PEARC 21:40:38 <martial> rbudden jmlowe going? 21:40:41 <oneswig> I wondered if we should include some folks on the AI/GPU/HPC track? 21:41:00 <martial> we can reach out to them indeed 21:41:22 <rbudden> unfortunately I can’t attend PEARC since it conflicts with my yearly vacation 21:41:49 <rbudden> that does remind me we are in search of a moderator to fill in for me for the Panel 21:42:08 <martial> happy to help, for PEARC? 21:42:12 <rbudden> yes 21:42:23 <rbudden> my boss was going to help out, but he has since left PSC 21:42:28 <martial> you have my email, reach out please and let's take this offline 21:42:36 <oneswig> J Ray? 21:42:38 <rbudden> yes 21:42:46 <oneswig> well well 21:42:54 <jmlowe> I am going to PEARC 21:43:09 <rbudden> oneswig: we can catch up about that offline if you’d like 21:43:28 <rbudden> short news is he’s moved up to computing services just up on CMU campus 21:43:38 <jmlowe> Are we taking AOB items? 21:43:47 <martial> sure 21:43:52 <rbudden> apologies for the off topic ;) 21:43:53 <martial> #topic AOB 21:43:54 <jmlowe> SIG social 21:44:42 <oneswig> jmlowe: are we at ~15? 21:45:07 <jmlowe> So far my count is 14 people for http://www.nuba.ca/ 21:45:11 <martial> am going to know if a couple ORCA/P2302 colleagues can join 21:45:28 <oneswig> jmlowe: is there a date for final numbers? 21:45:34 <jmlowe> I was shooting for Tuesday night 7:00ish 21:45:34 <martial> very soon (the note was hidden in its own email ... I should know better) 21:45:50 <oneswig> As in, about now? 21:45:51 <rbudden> might have missed it, but was there a place to sign up? 21:45:58 <jmlowe> until this afternoon there were no other events, now Redhat/EMC has something 6-9 21:46:04 <jmlowe> I signed you up 21:46:05 <rbudden> jmlowe: you might have just assumed i was coming 21:46:09 <rbudden> haha ok, thx 21:46:32 <jmlowe> We went there during the '15 summit 21:47:10 <oneswig> I've not heard back from the CERN team - will check up 21:47:12 <jmlowe> I need to make the reservation soon and secure it with my credit card 21:47:34 <jmlowe> we could all just show up and take our chances 21:48:09 <martial> (sending the email now ... again) 21:49:44 <martial> (sent) 21:50:24 <martial> maybe we have it after 8pm then :) 21:51:01 <martial> anything else? 21:51:34 <martial> wait I thought the RedHat thing was Wedneday 21:51:46 <oneswig> Not much from us. rbudden: are you aware of any activity to get Ceph working natively over OPA? 21:52:20 <rbudden> not offhand, though we bailed on Ceph for awhile due to not enough nodes to dedicate to it to be optimal 21:52:23 <jmlowe> There's a redhat ceph thing wednesday, https://www.redhat.com/en/events/crafting-your-cloud-reality?sc_cid=701f2000000Zx33AAC 21:52:28 <rbudden> we’re actually revisiting this though 21:52:30 <jmlowe> that's tuesday 21:52:53 <rbudden> our DXC project ends in August and I’m hoping to swipe 3-4 JBOD/Server pairs to do a proper Ceph Setup :) 21:53:02 <rbudden> So I am definitely interested in this 21:53:49 <oneswig> rbudden: I've been using IPoIB on OPA but I believe the performance would be significantly improved by the IB verbs messenger 21:55:03 <rbudden> we’ve been using IPoIB as well and we’ve had some rather quirky intermittent behavior 21:55:07 <martial> (5 minutes mark) 21:55:21 <rbudden> it would be great to have a native verbs solution 21:55:30 <oneswig> connected mode and big MTU seems to help but I get huge variations in performance all the same. 21:55:50 <oneswig> rbudden: if I get it going, I'll be sure to let you know... 21:55:59 <jmlowe> I'm keeping an eye on this, if we get a renewal it would be good to know if I need to get nics that can do RDMA 21:55:59 <oneswig> (likewise, I'm sure) 21:56:19 <rbudden> thx 21:56:35 <oneswig> jmlowe: have experimented with RDMA on 25GE and RoCE for Ceph - early days on that investigation still. 21:57:06 <oneswig> I also have 100G EDR IB but that didn't work on first attempt. Bah. 21:57:34 <jmlowe> I'm thinking we've got about 2 years before we would need to put a proposal together for a no compete renewal 21:57:58 <oneswig> a lot will happen in that time in the Ceph world. 22:00:13 <martial> okay, sounds like we are at the end of the meeting :) 22:00:17 <martial> thanks everyone 22:00:24 <oneswig> Thanks martial and all 22:00:35 <martial> see a lot of you (all the active ones Today) in Vancouver next week :) 22:00:47 <rbudden> absolutely! 22:00:49 <martial> #endmeeting