21:01:47 <martial> #startmeeting scientific-wg 21:01:48 <openstack> Meeting started Tue May 30 21:01:47 2017 UTC and is due to finish in 60 minutes. The chair is martial. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:01:49 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:01:51 <openstack> The meeting name has been set to 'scientific_wg' 21:01:59 <martial> #topic Collecting together research papers: user stories of Scientific OpenStack 21:02:02 <priteau> better :-) 21:02:03 <martial> #chair oneswig 21:02:03 <openstack> Current chairs: martial oneswig 21:02:05 <oneswig> And now for something completely different! :-) 21:02:10 <martial> Ni 21:02:10 <trandles> haha 21:02:15 <oneswig> too good! 21:02:26 <oneswig> #link agenda https://wiki.openstack.org/wiki/Scientific_working_group#IRC_Meeting_May_30th_2017 21:02:33 <oneswig> OK, lets get this show on the road. 21:02:44 <oneswig> Blair sends apologies - he is in flight 21:03:55 <martial> #topic Scientific datasets blog from SWITCH 21:04:12 <martial> #link http://lists.openstack.org/pipermail/user-committee/2017-May/002051.html# 21:04:28 <martial> so switch people? 21:04:32 <oneswig> right - this was a blog post by Saverio from SWITCH 21:04:41 <oneswig> it's late night for them. 21:04:58 <oneswig> I put it on the agenda because I thought it might interest some people (including me) 21:05:10 <martial> #link https://cloudblog.switch.ch/2017/05/22/hosting-and-computing-public-scientific-datasets-in-the-cloud/ 21:05:18 <martial> #topic Collecting together research papers: user stories of Scientific OpenStack 21:05:41 <martial> do we have anybody else able to contribute such user stories 21:05:43 <martial> ? 21:06:31 <oneswig> I've not previously used Zenodo 21:06:53 <trandles> we've been talking about public datasets in the cloud internally but that's as far along as we are 21:06:58 <oneswig> How do we check a paper is appropriate for sharing? 21:10:56 <martial> I would guess it has to do with the use of the dataset 21:11:04 <TheMistyMay> hello! 21:11:08 <martial> is it public, can it be used, etc ... 21:11:17 <oneswig> Hello TheMistyMay, welcome 21:12:00 <martial> as for the paper, "user stories" might also be good ways for people to share technology work / success they are having 21:12:13 <martial> maybe we ought to have those peer reviewed 21:12:41 <martial> the plan is still to have those added to a possible v2 of the book? 21:12:53 <oneswig> martial: for sure, user stories is another level. The papers I've seen up on Zenodo are more academic in style 21:13:16 <oneswig> martial: I'd like to collect user stories, each being approximately a chapter in scope. 21:13:39 <oneswig> eg, identity federation (the current work in 'progress' - although I blush to call it that) 21:14:00 <martial> how long are we talking about here? 21:14:50 <martial> Mike (TheMistyMay) made a demo at the lighting talks, would a 3-4 pages description of why/how/links be a good user story for example? 21:14:51 <oneswig> martial: no fixed constraint but a working rule of thumb was about 8 pages (plus some images) in a google doc... 21:15:48 <oneswig> That sounds good, but the other sections bring in ~3 related user stories into a common theme - what was the demo on Mike? 21:15:54 <TheMistyMay> I'd be happy to write something up as a "Helping to Prevent Bit Rot" in the style of a user story 21:16:56 <oneswig> #link existing work for Scientific WG study here: https://github.com/openstack/scientific-wg/tree/master/doc/source 21:17:07 <TheMistyMay> It was a demo of CI/CD build base images for openstack (and other common IaaS). As well as an nginx reverse proxy plugin that took care of authentication so researchers (users) didn't have to worry about edge security. 21:18:00 <martial> that actually sounds pretty cool to me 21:19:22 <oneswig> That might be a useful contribution on a wider section on how people go about doing research computing development on cloud - what people sometimes call "ResOps"? 21:19:53 <martial> oneswig: too early for me and Dmoni and such ... a public beta planned for a couple of weeks from now 21:20:46 <oneswig> What's interesting to me is that there isn't really an analogy to working at the system image level on conventional infrastructure. It's a new level of freedom - conversely a new level of burden :-) 21:21:42 <oneswig> martial: plenty of time for that though, right? I think we are targeting November... 21:22:22 <martial> oneswig: I know but a user story based on it would be great 21:23:27 <martial> (ie will make it happen after we introduce it here) 21:24:45 <oneswig> Mail just in from Lauren Sell - http://lists.openstack.org/pipermail/user-committee/2017-May/002091.html 21:25:13 <oneswig> Should we have people's video picks of outstanding talks they saw in Boston? 21:25:45 <martial> that sounds like a good idea to me 21:25:57 <oneswig> #topic summit video picks 21:26:04 <martial> (remembering the presentation name is going to be the tricky part) 21:28:54 <oneswig> #link Manila, CephFS at CERN - watched this video, it's quite similar to some things we are doing, was intrigued by the question at the end on RDMA (thought that was not active) https://www.openstack.org/videos/boston-2017/manila-on-cephfs-at-cern-the-short-way-to-production 21:29:11 <martial> #link Container as a Service on GPU Cloud: Our Decision Among K8s, Mesos, Docker Swarm, and OpenStack Zun https://www.openstack.org/videos/boston-2017/container-as-a-service-on-gpu-cloud-our-decision-among-k8s-mesos-docker-swarm-and-openstack-zun 21:29:49 <oneswig> One thing that we are stuck on is that there's no Manila equivalent of provider networks - 'provider filesystems' if you will - I want to manage access to existing filesystems, which can be big and long-lived 21:30:19 <trandles> oneswig: +1 on 'provider filesystems' 21:30:20 <priteau> +1 for the CephFS talk at CERN, that was a good one 21:30:51 <martial> #link OpenStack + OpenHPC = HPC Cloud https://www.openstack.org/videos/boston-2017/openstack-openhpc-hpc-cloud 21:31:27 <oneswig> Cool - something we are interested in too! 21:31:52 <Georgem_> https://youtu.be/EhrXALI_I-Y 21:32:07 <trandles> What's the deadline on providing some input on videos? I haven't even had a chance to start through my watch list. :( 21:32:15 <oneswig> That you George? :-) 21:32:26 <Georgem_> Yeap 21:32:42 <Georgem_> On the train, poor connection:( 21:33:47 <oneswig> trandles: lets scrape these links and keep going via openstack-operators in the next few days 21:34:14 <oneswig> I haven't had a good look through myself yet, not even made it through the HPC/Research track 21:34:30 <trandles> Ok. I'll try to provide some feedback by the end of the week. 21:34:36 <martial> #link The U.S. Army Cyber School OpenStack Use Case: Saving Millions and Making Changes for the Benefit of National Defense https://www.openstack.org/videos/boston-2017/the-u-s-army-cyber-school-openstack-use-case-saving-millions-and-making-changes-for-the-benefit-of-national-defense 21:35:17 <martial> (that last one was very interesting, an extension of the keynote presentation) 21:35:32 <oneswig> I love the fact they are presenting in uniform! 21:36:07 <TheMistyMay> fun fact: cyber schools architecture was build around a deployable mobile cloud architecture that was mission oriented 21:36:54 <martial> #link Towards a Platform for Science - Initial Lessons from the Science Accelerator Platform at LBL https://www.openstack.org/videos/boston-2017/towards-a-platform-for-science-initial-lessons-from-the-science-accelerator-platform-at-lbl 21:38:29 <martial> that's it for me for now 21:39:21 <oneswig> martial: How was the Cyborg talk in the end - I know that's one you're interested in 21:39:45 <martial> oneswig: it was very interesting and a good introduction for people to join the team 21:39:54 <priteau> I haven't watched it yet, but there was another CERN presentation, about containers 21:39:57 <priteau> #link https://www.youtube.com/watch?v=VOh0dwPGeSM 21:40:29 <martial> but my colleague goldenfri is on the team for the presentation, so I will let him talk about it 21:41:16 <oneswig> priteau: very interested in new work from CERN OpenLab 21:42:08 <martial> #link OpenStack Acceleration Service: Introduction of Cyborg Project https://www.openstack.org/videos/boston-2017/openstack-acceleration-service-introduction-of-cyborg-project 21:44:16 <oneswig> OK, I'll collect these together and follow up to the list, we'll get a thread going there 21:44:27 <martial> sounds like a good plan 21:44:35 <martial> #topic Working Group activities for the new release cycle 21:45:44 <oneswig> There's definite interest in updating and extending the studies done so far. 21:46:54 <martial> stig: extending those would make a lot of sense I agree 21:47:28 <oneswig> What's bugging everyone else? 21:48:34 <martial> stig: as you know, I want to make our set of tools open to the SWG for testing, and hopefully have a stable version for the next in person meeting in November 21:49:11 <martial> My team and I will welcome comments, recommendations and bug reports (and fixes :) ) 21:49:32 <oneswig> martial: good for you, sounds like a very worthwhile contribution 21:49:46 <oneswig> You think it'll be public in a couple of weeks? 21:50:00 <martial> we are packaging it internally as an alpha 21:50:07 <trandles> bare metal is still bugging me and it looks like progress is being made on that front 21:50:23 <trandles> if that's the case I'll be happy to show off our use case in November 21:50:29 <trandles> but it's nothing new for the WG I'm afraid 21:50:41 <martial> going to try to have a few people to use it, comment on the "docs" ... we all know we write our docs for people who already know how to use it 21:51:16 <martial> then release a beta on github 21:51:35 <martial> tim: although it is not new, it is still relevant 21:51:37 <oneswig> trandles: one area I am hopeful for progress is in the new roles in the Ironic state machine and how that might enable us (with the new Nova scheduler capabilities) to find a more elegant way of reprogramming BIOS and RAID for bare metal instances 21:52:07 <martial> tim: looking forward to progress or best case definition 21:53:08 <oneswig> I think Mike Lowe was interested in continuing the project on handling datasets 21:54:51 <martial> stig: I think this is correct 21:55:55 <martial> does not seem there is much to be discussed today 21:56:07 <martial> #topic Any other business 21:56:17 <martial> open floor... 21:56:24 <trandles> sorry I'm being pulled in two directions at the moment so only about 50% present here 21:57:07 <martial> tim: no worries, we understand. Feel free to follow up if you are able to 21:57:07 <oneswig> Anyone using CephFS in bare metal? We've got curious problems with metadata latency 21:57:25 <trandles> in fact, I have to run... I'll follow up on the mailing list thread with video feedback 21:57:28 <oneswig> bandwidth is good but directory operations really seem to hurt 21:57:29 <trandles> later! 21:57:33 <oneswig> see you trandles 21:58:26 <oneswig> I think I've got a blog article to make on it, but only if I can find an answer... 21:59:22 <themisty_> We're experimenting with bare metal Ceph too 21:59:55 <oneswig> themisty_: cool - lets keep in touch - #scientific-wg is the place to be :-) 22:00:09 <themisty_> :) I plan to try and be around more 22:00:21 <oneswig> I leave you with this: https://www.youtube.com/watch?v=dMwP7nEuDGA 22:01:00 <martial> sweeeet 22:01:15 <oneswig> OK folks, got to pop 22:01:21 <oneswig> until next week 22:01:28 <oneswig> #endmeeting