11:01:27 <martial_> #startmeeting Scientific-SIG
11:01:28 <openstack> Meeting started Wed Mar 14 11:01:27 2018 UTC and is due to finish in 60 minutes.  The chair is martial_. Information about MeetBot at http://wiki.debian.org/MeetBot.
11:01:29 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
11:01:31 <openstack> The meeting name has been set to 'scientific_sig'
11:01:53 <martial_> Good morning
11:01:56 <oneswig> Morning martial_ thanks for driving today
11:01:59 <priteau> Good morning Martial
11:02:16 <martial_> hey Stig, Pierre
11:02:23 <martial_> stig: sure thing
11:02:23 <daveholland> o/ (might get pulled to another meeting but hoping to hear/talk about hierarchical projects)
11:02:37 <verdurin-alt> Morning.
11:02:42 <martial_> well let's get started with that then :)
11:02:50 <martial_> #topic hierarchical projects
11:02:59 <daveholland> :)
11:03:25 <martial_> daveholland: want to give us the brief on the forum post?
11:03:56 <daveholland> I haven't seen it but I have seen past proposals I think, I can outline the use case for us if you want?
11:04:48 <oneswig> #link discussion on user-committee list http://lists.openstack.org/pipermail/user-committee/2018-March/002636.html
11:05:01 <oneswig> That the one Martial?
11:05:09 <martial_> The forum post contained: "[...] an enforcement model is an opinionated way of how a quota,    or limit, should behave with respect to other parent, sibling, or child    projects. It's possible to think of multiple ways in which enforcement    can be done, and it's not that there is one right way and the rest are    wrong, just a difference in how quotas might need to behave for different deployments."
11:05:43 <martial_> #link http://lists.openstack.org/pipermail/openstack-sigs/2018-March/000309.html
11:05:50 <martial_> I had this one (thanks Stig)
11:06:18 <daveholland> Thanks, I hadn't caught up with email that far yet
11:06:59 <martial_> this one link and this one link
11:07:02 <martial_> #link http://lists.openstack.org/pipermail/openstack-dev/2017-February/111999.html
11:07:43 <daveholland> FWIW our users would like nested/heirarchical projects to look as much as possible like the current implementation; but to allow a higher-level project to be a reporting method for the subprojects. If it was an implementation detail that a project could only contain either projects or other resources that would be OK.
11:08:17 <daveholland> (i.e. a top-level "container" project with leaf projects containing instances, volumes etc)
11:08:17 <martial_> so "We'd like to see if the User Committee can help us collect more    information from users and operators about how they expect enforcement    to be done. Do your deployments use hierarchical projects? How do you    manage quota today? Do you have expectations about how quotas and limits    work across related projects (e.g. setting quota on a parent project   affects the children in X ways)?"s
11:09:21 <daveholland> we don't do hierarchical projects today (just migrating off Liberty, to Pike) and there's frustration around having to track many different quotas
11:10:06 <belmoreira> CERN has a blog post about our use case: http://openstack-in-production.blogspot.fr/2017/07/nested-quota-models.html
11:12:00 <martial_> looking at the options section of this post helps as well, thank you belmoreira
11:12:09 <daveholland> Maybe I'm getting ahead a bit, but I can imagine user desire for both absolute and proportional sub-project quota settings ("I want this CI subproject to have 10 VCPUs", vs "I want my two PhD students to each have 50% of this project's quota")
11:13:40 <martial_> that is also how we would use it for our users but can this work with cpu overprovisioning for example?
11:14:43 <daveholland> We are only just getting into overprovisioning/oversubscription, previously have set cpu_allocation_ratio=1 everywhere. Should a VCPU on a potentially-oversubscribed hypervisor count for 1 or 1/N VCPUs?
11:14:44 <belmoreira> daveholland cpu overprovisioning is related with the allocation that you allow. I don't see a problem
11:15:40 <martial_> priteau: does this affect you as well?
11:15:42 <daveholland> I see a need to distinguish, because we intend to allow users to choose oversubscribed or not, based on flavour
11:17:33 <priteau> martial_: being mostly bare-metal, we don't use overprovisioning, but we may want to use hierarchical projects for teachers who then use sub-projects for groups of students. We could work with absolute quota settings but it would surely be easier if proportional was possible.
11:18:06 <martial_> so both models have apllications
11:19:27 <daveholland> martial_: yes, definitely
11:19:56 <martial_> the picture at the end of the cern blog post and the explaination are a good example of the complexity
11:21:01 <martial_> we have had a couple users groups explain to us how they use the model, and both cases are represented
11:21:29 <daveholland> Also a vote from here for "It should be possible for an administrator to lower the limit on a child project below the current usage." - sometimes it's necessary to get people to tidy their rooms :)
11:22:14 <martial_> daveholland: or release a GPU :)
11:22:50 <martial_> any other point of view on this subject?
11:24:11 <martial_> because this is a conversation in an open forum, we would like to invite people to explain their use case.
11:24:32 <martial_> as a reminder, the link for this topic can be found in the ML archives at
11:24:45 <martial_> #http://lists.openstack.org/pipermail/openstack-sigs/2018-March/000309.html
11:24:49 <martial_> # link http://lists.openstack.org/pipermail/openstack-sigs/2018-March/000309.html
11:24:53 <martial_> #link http://lists.openstack.org/pipermail/openstack-sigs/2018-March/000309.html
11:24:56 <martial_> (silly fingers)
11:25:03 <verdurin-alt> Our users would probably be similar to daveholland
11:25:53 <martial_> thank you verdurin-alt
11:27:25 <martial_> thank you all for sharing your comments, this is not going to be solved this morning, but as we are trying to bring awareness to the problem, we invite all to contribute to the forum
11:27:44 <martial_> any additional comment on this topic?
11:28:30 <martial_> moving on to the next topic then
11:28:46 <martial_> #topic upcoming Federated Cloud workshop
11:29:05 <martial_> #link https://federatedcloud.eventbrite.com
11:29:47 <martial_> Some of you might be familair with the "Open Research Cloud" effort that was started at the Boston summit last yea
11:29:50 <martial_> +r
11:30:44 <martial_> in September 2017, the National Institute of Standards and Technology (NIST) stated a Public Working Group to bring people together to discuss cloud interoperability
11:30:57 <martial_> this was an effort in collaboration with the IEEE P2302
11:31:55 <martial_> next week in Gaithersburg, MD, USA we are continuing this conversation with a meeting hosted by the NIST PWG to bring together people and continue our conversation on this effort
11:32:37 <martial_> more details are at the link posted above
11:33:21 <martial_> any question on this effort?
11:33:44 <martial_> we will have remote participation for people who are interested
11:33:59 <verdurin-alt> martial_ - that was going to be my question
11:34:10 <martial_> and some of the members of our community are joining us
11:35:06 <martial_> note that it is not openstack specific (IEEE effort)
11:35:13 <daveholland> there was some interest in ORC here but I've not heard of this event, I'll pass it on
11:35:40 <martial_> daveholland: my apologies, I have posted about it in the forum but we have not followed up
11:35:51 <daveholland> NP!
11:35:53 <martial_> I mention it once in a while :)
11:36:46 <martial_> but we truly welcome our community and participation, so please feel free to share the information
11:37:12 <martial_> the plan is so far to host the follow up meeting during the Vancouver Summit
11:37:33 <martial_> Ildiko has been instrumental to help us with the effort
11:38:22 <martial_> any questions that I can answer?
11:39:53 <martial_> cool :)
11:39:59 <PeteC> I've not heard anything regarding spot price equivalents in OpenStack for a while. Has anyone heard anything ?
11:40:23 <martial_> bringing us to
11:40:26 <martial_> #topic AOB
11:41:25 <martial_> Pete: you are talking about pricing comparison with AWS and such?
11:42:32 <PeteC> The ability to use pre-emptive instances, which may be disposable if higher priority instances are required.
11:43:55 <martial_> this sounds like it could be a side conversation to the nested quota we had earlier
11:44:31 <verdurin-alt> PeteC there's a summary of the Dublin PTG discussion about this here https://etherpad.openstack.org/p/scientific-sig-ptg-rocky
11:44:39 <PeteC> So this ( I think ) is slightly skewed to quotas.
11:44:42 <PeteC> and ty
11:47:15 <martial_> PeteC: Stig might be able to provide additional information on spot instances during a follow up meeting
11:47:28 <martial_> (beyond the SIG's PTG notes)
11:48:47 <martial_> PeteC: thank you for brining this up, I will follow up with Stig and have this topic added to a future conversation
11:49:06 <martial_> anything else?
11:51:43 <martial_> okay then, with that, I will call the meeting to an end, thank you for joining us Today
11:51:52 <daveholland> thanks :)
11:52:14 <verdurin-alt> Thanks.
11:52:26 <martial_> and if you have quota needs, join the mailing list/forum conversation :)
11:52:56 <martial_> #endmeeting