21:01:13 <oneswig> #startmeeting scientific_wg 21:01:14 <openstack> Meeting started Tue Mar 21 21:01:13 2017 UTC and is due to finish in 60 minutes. The chair is oneswig. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:01:15 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:01:18 <openstack> The meeting name has been set to 'scientific_wg' 21:01:26 <oneswig> Alas, all this talk of currywurst has made me hungry 21:01:40 <oneswig> #chair martial_ 21:01:41 <openstack> Current chairs: martial_ oneswig 21:01:46 <martial_> hello stig 21:01:50 <trandles> yeah...mid-afternoon doldrums have taken hold and currywurst isn't helping 21:01:51 <oneswig> Greetings 21:02:03 <martial_> hi tim 21:02:05 <rbudden> hello everyone 21:02:10 <trandles> hell 21:02:14 <trandles> +o 21:02:14 <oneswig> #link OK here's an agenda https://wiki.openstack.org/wiki/Scientific_working_group#IRC_Meeting_March_21st_2017 21:02:15 <trandles> oops 21:02:30 <oneswig> Good afternoon America :-) 21:03:06 <oneswig> #topic WG round-up from Milan 21:03:17 <oneswig> Was anybody else present or was it just me? 21:03:52 <martial_> oneswig: nope sorry, not me 21:03:53 <oneswig> ... in which case I'll do my best to share what I'd learned from a WG perspective 21:04:02 <flanders_> Spill the goss ;) 21:04:08 <Guest81586> Hi there 21:04:16 <oneswig> First up, our hosts in Milan were excellent, just great 21:04:17 <hogepodge> o/ 21:04:23 <oneswig> hi y'all 21:04:56 <oneswig> There was one discussion which may require a little follow up 21:05:14 <oneswig> The team at OSIC were looking for test case workloads for testing live migration 21:05:24 <oneswig> And they didn't have anything representative for science. 21:05:47 <Guest81586> Do you have more details? 21:06:06 <oneswig> I am not sure of the finer details. Unfortunately the guy I spoke to is on vacation this week. 21:06:18 <b1airo> hello! 21:06:38 <oneswig> I believe what they want is a heat stack (or equivalent) that deploys some busy busy workload that runs for long enough for them to experiment with migrating 21:06:45 <oneswig> And at the end, a yes/no pass fail 21:06:48 <oneswig> Hi b1airo! 21:06:52 <oneswig> #chair b1airo 21:06:53 <openstack> Current chairs: b1airo martial_ oneswig 21:07:28 <oneswig> #link might be further details here (but I don't recall it) https://etherpad.openstack.org/p/MIL-ops-live-migration 21:07:30 <martial_> oneswig: any use of ansible in that model? 21:08:03 <oneswig> martial_: I'd have thought that's a good fit. Create a long-ish running workload - easy thing to initiate the infrastructure for that with ansible 21:08:37 <oneswig> So... how are we all doing with our cloud-native deployment of science workloads ... ? 21:08:45 <oneswig> yes it's on my list too 21:09:56 <b1airo> oneswig, if i wanted cloud native i'd go to azure 21:10:23 <martial_> oneswig: related to ansible, they just posted their videos from their recent meet in DC, see: https://www.ansible.com/videos-ansible-automates-2017 21:10:28 <oneswig> Interesting you'd say that. From what I've heard recently Oracle bare metal's the new king of the hill 21:11:06 <b1airo> i took "cloud-native" to be more about application architecture... 21:11:16 <mrhillsman> oneswig izaak can speak more to the workload when he gets back 21:11:33 <mrhillsman> in particular the framework 21:11:35 <oneswig> Hi mrhillsman, thanks for dropping in! 21:12:00 <oneswig> mrhillsman: can you give examples of other test cases you're working with? 21:12:44 <Guest81586> Oneswig: are you planning a dinner event like in Barcelona? 21:12:45 <oneswig> b1airo: I was thinking of the packaged deployment, cluster-as-a-service creation etc. 21:12:45 <mrhillsman> we defined the workloads ourselves to get the framework to a "v1" status 21:13:18 <mrhillsman> details can be found here - https://github.com/osic/ops-workload-framework 21:13:28 <oneswig> Guest81586: would very much like to - Boston coming up on the agenda 21:13:45 <mrhillsman> izaakk has more detail than me however 21:14:10 <mrhillsman> hopefully the readme sheds some light for you in prep for that meeting when he gets back 21:15:24 <oneswig> mrhillsman: how would you like people to contribute - pull requests? 21:16:56 <b1airo> i'm confused by the VM types described in that readme 21:17:43 <b1airo> "Large VM ( CPU: 6, RAM: 4g, Disk: 6g)" - doesn't seem particularly large? 21:18:21 <oneswig> b1airo: I wonder if it's large for something that can be live-migrated? 21:18:34 <b1airo> hmm not in our experience 21:18:50 <Guest81586> Maybe it's european 21:18:58 <b1airo> lol 21:19:09 <oneswig> Are you doing live migration with NUMA passthrough - does that work? 21:19:57 <oneswig> Guest81586: European VMs, compact and bijoux... 21:20:13 <b1airo> oneswig, no we're not, though i believe it is supposed to work if the destination can meet the topology requirements, however in our case we are also doing SRIOV VF passthrough so live-migration is out due to that 21:20:45 <b1airo> also this sounds odd: "Spans 60 disk workers that fills up the disk by creating 1 gb blocks of data." 21:21:01 <b1airo> in the context of a VM with on 6G of disk how is that supposed to work...? 21:21:08 <b1airo> s/on/only/ 21:21:13 <martial_> (looking at the agenda, I think we will need to move to the next topic soon, even if we come back to this one later if time permits) 21:21:25 <b1airo> mrhillsman, i must be misunderstanding something? ^^^ 21:21:50 <mrhillsman> i think there will be a discussion with izaakk about the ask 21:22:11 <oneswig> martial_: thanks, good prompt. I think we'll need to follow up with izaakk and mrhillsman 21:22:33 <mrhillsman> i do not have detail unfortunately but hopefully looking at the repo gives you some context rather than having to wait a week to get the same info 21:22:40 <mrhillsman> it is a good starting point 21:22:50 <martial_> mrhillsman: thanks :) 21:23:16 <mrhillsman> stepped away, apologies 21:23:23 <mrhillsman> the sizes are related to the defaults 21:23:50 <martial_> #topic Boston Summit Forum - contribution to discussion 21:23:59 <mrhillsman> hope in discussion with different stakeholders is that things will change based on need 21:24:23 <martial_> #link https://etherpad.openstack.org/p/BOS-UC-brainstorming-scientific-wg 21:24:39 <martial_> so the link for the brainstorming etherpad is above 21:24:57 <oneswig> Thanks martial_ - so the activity on the list indicates it's time to collect these diverse thoughts together 21:25:18 <oneswig> #link http://lists.openstack.org/pipermail/user-committee/2017-March/001856.html 21:25:32 <martial_> yes, I wanted to try to go from the etherpad to a list of topic to be discussed and a list of people assigned to those topics 21:25:47 <martial_> I could not find the 1-to-1 match from the etherpad history 21:26:43 <martial_> I figure we ought to have a person working on each topic present this discussion 21:27:08 <martial_> what do you think? 21:27:29 <martial_> (in a way that was my question to flanders_ yesterday) 21:27:30 <oneswig> Seems like a good idea to find advocates for each 21:27:56 <b1airo> recall the original instructions were to include a leader for each 21:28:38 <flanders_> +1 21:28:52 <trandles> "Please add contact information next to your topic, so we can contact you for formal "forum" session posting. Thank you" 21:29:25 <oneswig> flanders_: Can people participate in this stage who will not be present at the forum? 21:29:43 <flanders_> And have that leader identify which upstream contributors to invite, who can act as advocates for actions arising from forum sessions. 21:29:49 <martial_> trandles: yes, I added this earlier today 21:30:14 <flanders_> oneswig: not sure, would post that Q publucally on UC ML 21:30:28 <b1airo> oneswig, can't see why not - at least half the effort is just in getting it organised 21:30:36 <martial_> looking at the history, I think "Supporting Pre-emptible instances" is "Adam Huffman" 21:30:37 <oneswig> OK I can take a couple of these up provisionally 21:31:11 <flanders_> b1airo: +1 but would be great for other WGs to understand this. 21:31:26 <oneswig> martial_: equally likely that aloga_ would advocate it 21:31:59 <oneswig> Can we pick up tasks using #action? 21:32:10 <Guest81586> I got to go now but I'll check the meeting log for details on any Boston event 21:32:22 <oneswig> OK thanks Guest81586 21:32:29 <b1airo> i am guessing Tim is still happy to take "Quota management pain points, workarounds, cross-project needs" 21:33:17 <martial_> Maybe we ought to email the SWG ML with the link to the etherpad to try to determine our "champions" 21:33:46 <martial_> because reading the pad/history I am not sure who to contact directly mostly 21:34:41 <oneswig> martial_: agreed, it's not a reliable method 21:35:42 <martial_> okay, I will send an email to the scientific wg with this request post our meeting 21:36:02 <martial_> this ought to help us concretize a list of topics 21:36:02 <oneswig> Thanks martial_ 21:36:10 <martial_> April 2nd is coming fast after all :) 21:37:52 <martial_> I have not seen must additional discussion on the Cloud Declaration 21:37:58 <martial_> (must -> much) 21:38:21 <martial_> #topic Other Boston Summit planning 21:39:01 <oneswig> In general the lightning talks format of the BoF seemed popular. 21:39:05 <oneswig> Any thoughts on that? 21:40:09 <martial_> oneswig: seems feasible, although if the foundation is taking care of hte schedule it might be good to ask people how long they expect to discuss a topic for? 21:40:20 <martial_> (sure there are the after summit hours :) ) 21:41:08 <oneswig> martial_: the guillotine on overrunning talks? 21:41:34 <martial_> oneswig: my worry exactly :) 21:41:52 <martial_> flanders_: any opinion on this topic? 21:42:21 <oneswig> Surely easy enough to enforce 21:42:49 <martial_> oneswig: the question is are those going to be during the SWG discussions? 21:43:28 <oneswig> I haven't had confirmation but my understanding is we have a meeting session and a BoF session. 21:43:32 <martial_> #link https://www.openstack.org/summit/boston-2017/call-for-presentations/preview/17698 Scientific Working Group planning meeting 21:43:34 <oneswig> This would be the Bof 21:43:46 <b1airo> i think judging from feedback on how we ran things in barcelona people wanted more time for discussion 21:44:02 <martial_> #link https://www.openstack.org/summit/boston-2017/call-for-presentations/preview/17699 Scientific OpenStack BoF 21:44:12 <b1airo> they liked the lightening talks, but there was a general feeling we need more time in the planning session 21:44:32 <martial_> b1airo: and a bigger room :) 21:44:49 <b1airo> that also 21:45:20 <oneswig> my bad - thought a 100-person meeting was best avoided! :-) 21:45:49 <martial_> oneswig: many scientists :) 21:46:01 <b1airo> actually i didn't mind, makes it easier to hear people and others are more likely to shutup and listen :-) 21:47:06 <martial_> okay, so no consensus on this at this point 21:47:25 <martial_> we can open this for question to the champions 21:47:29 <oneswig> martial_: no I think a bigger room for the WG meeting makes more sense 21:47:49 <martial_> I can add it to the email (format + time needed) 21:48:38 <martial_> I think there was also a question about a dinner for the SWG? any plans at this point? 21:48:58 <b1airo> "Martial Michel: To be extended." 21:49:06 <martial_> trandles: you were involved in this discussion, I remember 21:49:06 <oneswig> I'm looking for a volunteer on the ground for the evening social 21:49:26 <trandles> martial_: which discussion? dinner? 21:49:33 <martial_> do we have a boston local? 21:49:44 <martial_> trandles: yes, was I incorrect? 21:49:51 <trandles> I passed along info that a Boston-local friend provided me 21:50:08 <martial_> I remember, thanks Tim 21:50:10 <b1airo> though you won't actually be there trandles ? 21:50:17 <trandles> yeah, I'll be there 21:50:31 <trandles> I get the feeling that I'm being volunteered :P 21:50:31 <b1airo> ah ok great 21:50:45 <b1airo> it is feeling that way isn't it... weird 21:50:48 <flanders_> Volontold ;p 21:50:49 <trandles> lol 21:50:53 <trandles> yeah 21:50:55 <martial_> trandles: not really, just asking politely :) 21:50:57 <trandles> I'll submit 21:51:20 <b1airo> awesome! i love it when oneswig doesn't have to do everything 21:51:42 <martial_> b1airo: stig is not going to be there either sadly 21:51:46 <oneswig> I'm pretty pleased with that - great work all 21:51:48 <oneswig> :-) 21:52:03 <trandles> oneswig: if you have any likely sponsors for me to hit up... 21:52:08 <martial_> look at that, one topic, 10 minutes :) 21:52:23 <oneswig> trandles: I am sure, we can work that. 21:52:31 <b1airo> no, some nonsense about a child's birthday - my own family are guilting me about being away for my birthday 21:52:55 <oneswig> it was my birthday in Barca ... 21:52:58 <jmlowe> didn't keep me from the Austin summit 21:52:58 <hogepodge> With time running out, can I have the floor for the last topic? 21:53:12 <oneswig> take it away hogepodge 21:53:14 <b1airo> please hogepodge..? 21:53:17 <martial_> oneswig: mine as well :) 21:53:20 <martial_> #topic Future Online Analysis Platform Workshop 21:53:35 <hogepodge> In a couple of weeks I'm going to a workshop to talk about OpenStack and scientific computing 21:53:58 <hogepodge> I have a 30 minute talk, and I'm hoping to incorporate examples from existing systems. 21:54:05 <b1airo> is that the future platform thing Kate Keahey was looking for someone to attend? 21:54:13 <hogepodge> Yes. 21:54:28 <jmlowe> I have a quick question, any chance anybody will be at the HPC Advisory council meeting in Lugano next month? 21:54:29 <b1airo> cool, looking for slideware etc? 21:54:38 <oneswig> jmlowe: see you there/ 21:54:41 <hogepodge> I was hoping that members of the WG could forward me some talks or slides they've given that I could incorporate into my talk 21:54:43 <hogepodge> yes :-D 21:54:54 <jmlowe> oneswig: excellent 21:54:57 <hogepodge> you can send it to chris at openstack.org 21:55:15 <jmlowe> hogepodge: we have a pile of stuff for our roadshow 21:55:20 <oneswig> hogepodge: I'll see what I've got that might help. 21:55:30 <martial_> should we crowdsource the ML as well ? 21:55:31 <hogepodge> some background on the event: "developing capabilities that would help us transform HPC resources to be more usable to experimental communities running analytics rather than simulation codes " 21:55:36 <b1airo> hogepodge, sure will share a deck or two and you can just ping me any questions 21:55:40 <rbudden> hogepodge: i can deliver some slides as well 21:55:46 <hogepodge> " those researchers need more on-demand access and more data management capabilities than the traditional data centers allow. The attendees will be primarily people designing various elements of these solutions — plus a panel of application scientists." 21:55:50 <hogepodge> fantastic, thank you everyone! 21:56:19 <hogepodge> I'm going to be in Berlin the week before the event, so I'm trying to get the slides put together this week. 21:56:20 <oneswig> Tough crowd hogepodge 21:56:21 <b1airo> hogepodge, so basically sounds like they are saying, "we want IOPS oriented filesystems now please" ...? 21:56:38 <martial_> :) 21:56:48 <oneswig> b1airo: is that on our etherpad yet? 21:56:55 <jmlowe> hogepodge: going to pitch openstack extras for scientific clouds at Lugano a few days later 21:56:57 <flanders_> Serverless compute and frictionless data for everyone! 21:57:00 <b1airo> oneswig, which? 21:57:08 <hogepodge> oneswig: b1airo: et. al. I'm also interested in any advice on how to communicate to the audience 21:57:16 <oneswig> b1airo: forum topics... 21:57:32 <hogepodge> That's it for me. thanks everyone! 21:57:34 <oneswig> jmlowe: me too - don't steal the show :-) 21:57:41 <martial_> #link https://etherpad.openstack.org/p/BOS-UC-brainstorming-scientific-wg 21:57:42 <b1airo> oneswig, are you referring to my IOPS comment or the general topic? 21:58:07 <b1airo> anyone planning to attend ISC ? 21:58:17 <jmlowe> oneswig: no worries, I barely get by on my public speaking abilities 21:58:21 <oneswig> b1airo: iops. I made a comparison between bare metal RBD and virtualised Cinder a couple of weeks ago. ouch 21:58:27 <hogepodge> (I'll be taking copies of the book too, so that work will also be represented) 21:58:34 <martial_> hogepodge: great thanks 21:58:40 <oneswig> b1airo: no plans as yet for ISC. 21:58:48 <martial_> hogepodge: I owe you this conversation for DockerCon as well 21:58:56 <b1airo> or the HPC Knowledge Meeting (http://www.hpckp.org/) ? 21:59:14 <oneswig> hogepodge: you have more copies? Can I get ~50? 21:59:40 <b1airo> oneswig, sounds fun - what was the virtualised backend ? 21:59:44 <hogepodge> I ordered for dockercon 21:59:47 <oneswig> ceph 21:59:58 <martial_> hogepodge: see you there then 22:00:18 <b1airo> so basically, baremetal ceph versus virtual ceph? 22:00:20 <hogepodge> 100 copies 22:00:25 <oneswig> hogepodge: let me know if you have more - conference is iirc 10-12 april 22:00:46 <oneswig> b1airo: correct - quite different (as you might expect) 22:00:55 <oneswig> Ah, times up 22:01:00 <b1airo> interesting thing to figure out tuning for though 22:01:02 <hogepodge> oneswig: yeah, we'll have 100 copies available 22:01:11 <martial_> hogepodge: seems reasonable. Dcon will be big, so we just have to make sure not everybody walks away with a copy right away, otherwise you will be done in minutes 22:01:32 <oneswig> going... going... 22:01:36 <oneswig> #endmeeting