15:01:38 <ad_rien_> #startmeeting massively_distributed_clouds
15:01:38 <ad_rien_> #chair ad_rien_
15:01:38 <ad_rien_> #info agenda
15:01:38 <ad_rien_> #link https://etherpad.openstack.org/p/massively_distributed_ircmeetings_2017 (agenda: line 378 )
15:01:39 <openstack> Meeting started Wed Mar 29 15:01:38 2017 UTC and is due to finish in 60 minutes.  The chair is ad_rien_. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:01:40 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:01:44 <openstack> The meeting name has been set to 'massively_distributed_clouds'
15:01:45 <openstack> Current chairs: ad_rien_
15:01:46 <ad_rien_> ok let's start
15:02:01 <ad_rien_> as usual please add you name/nick in the pad
15:02:02 <ad_rien_> thanks
15:02:03 <ad_rien_> o/
15:02:18 <jfpeltier> Hello all
15:02:27 <kgiusti> o/
15:02:28 <ad_rien_> Hi jfpeltier
15:02:41 <ansmith> o/
15:02:43 <rcherrueau> o/
15:02:49 <mabderrahim> Hi o/
15:03:40 <ad_rien_> #topic announcements
15:03:53 <ad_rien_> So from my side two important info:
15:04:00 <pabelanger> o/
15:04:03 <ad_rien_> First we have a slot for our face to face meeting in boston
15:04:24 <ad_rien_> #link  https://www.openstack.org/summit/boston-2017/summit-schedule/events/18671/fogedgemassively-distributed-clouds-working-group face to face meeting in Boston (please book the slo)
15:04:59 <jamemcc> Hi - Jamey here
15:05:10 <ad_rien_> I asked this slot because it will be come after the different presentations that will deal with multi sites/massively distributed / Fog /Edge …. stuffs
15:05:17 <ad_rien_> so this is I think rather good
15:05:20 <ad_rien_> Second announce:
15:05:37 <ad_rien_> We should agree on submitting a topic to the forum call
15:05:57 <ad_rien_> the link is in the pad.
15:06:24 <ad_rien_> Curtis, chairman of the NFV working group has already submitted something this morning and contacted us to propose to join his proposal
15:06:29 <ad_rien_> so the question is rather clear
15:06:37 <ad_rien_> Do you want to submit another topic ?
15:07:18 <ad_rien_> #link https://etherpad.openstack.org/p/BOS-UC-brainstorming-MassivelyDistributed-Fog-Edge (Massively Distributed WG, braimstorming on possible topics for the forum)
15:07:30 <ad_rien_> so I think we can spend 5 minutes (no more) on that question ?
15:08:16 <ad_rien_> If I'm correct this pad is rather ''empty'' in the sense that unless I am mistaken we did not get so many inputs ?
15:08:24 <ad_rien_> so I think we have two questions :
15:08:38 <jamemcc> My thought in general on Forum topics - and spcific I think to our group here - is you shoudl have a Forum topic to match any presenation you are giving
15:08:56 <ad_rien_> 1./ Does anyone know what is expected from the forum view point?  and 2./ Does it make sense to submit something?
15:09:30 <ad_rien_> So I think we can join the Curtis proposal
15:09:48 <ad_rien_> #link  http://forumtopics.openstack.org/cfp/details/58  (NFV WG forum proposal)
15:09:50 <jamemcc> So one or more on Fog/Edge is appropraite and even necessary to explain the work going on here
15:10:09 <jamemcc> Perhaps one on each of the Fog/Edge cases of the group
15:10:36 <jamemcc> Sorry if I'm going off in a strange direction - but I can't stop
15:10:38 <ad_rien_> It may overlap the Curtis proposal/
15:11:44 <jamemcc> In essence the Forum topic will be like a way to debate that specific area - think of it as a room and a time to have face to face
15:11:45 <ad_rien_> FYI I just copied/pasted the proposal of curtis in the pad
15:11:55 <rcherrueau> Can someone explain me, what a forum is exactly?
15:11:59 <ad_rien_> yes but how many people are expected ?
15:12:15 <ad_rien_> I mean having an informal discussion with 20/25 people is something which is already tedious
15:12:25 <ad_rien_> so if you get 100 persons, I don't know what we can expect ?
15:12:48 <kgiusti> rcherrueau: https://wiki.openstack.org/wiki/Forum
15:12:55 <rcherrueau> kgiusti: thanks
15:13:01 <paul-andreraymon> If you get 100 person, maye the outcome will be to create multiple subgroups.
15:13:05 <ad_rien_> Keeping in mind the proposal of Curtis, maybe we can make a proposal on the AMQP/oslo.messaging
15:13:06 <ad_rien_> ?
15:13:31 <jamemcc> Yes - from the Curtis run Telco NFV sessions in Barcelon - will get 25+ on this topics with majority being "interested"
15:13:34 <ad_rien_> paul-andreraymon:  good point. Who is going to propose to make subgroups, I guess you would need moderators
15:13:52 <ad_rien_> ok
15:13:58 <ad_rien_> So let's try to move forward
15:14:17 <ad_rien_> 1./ we have the curtis proposal (that explicitly mentions the Massively Distributed WG)
15:14:42 <ad_rien_> 2./ we can make another proposal related to the massively distributed but in that case we should position ourself w-r-t the Curtis proposal
15:14:51 <jamemcc> But still - being scared the discussion will be slow/often interrupted is probably not a good reason to abandon it entirely
15:14:56 <ad_rien_> 3./ we can make something more specifics like the AMQP discussion?
15:15:06 <jamemcc> @adrien #2 - Agreed
15:16:01 <ad_rien_> Guys, who will be interested by moderating a discussion around the AMQP debate (that has been launched last week)?
15:16:12 <ad_rien_> I think this can be valuable
15:16:30 <ad_rien_> more precisely this can make sense to try to clarify why we have folks that deal with RMQ
15:16:40 <ad_rien_> why other try to push ZMQ
15:16:50 <ad_rien_> and why Monasca has been built on top of Kafka
15:17:01 <ad_rien_> and finally what can be the place of Qpid
15:17:31 <ad_rien_> From my point of view this makes sense but it requires times to prepare the discussion (with appropriate content)
15:17:37 <rcherrueau> ad_rien_: I am OK to be part of that, but we should push agreed on the topic first.
15:17:48 <rcherrueau> we should be*
15:17:58 <jamemcc> @ rcherrueau - To try to answer in 1 line: Forum is the replacement for previous Design Summit - more emphasis on getting both User/Operators gettign their input and collaboration going but also together with the Project Teams.  Need to have the Operators/Users be specifying the important topics.
15:17:59 <kgiusti> ad_rien_: we'd need participation from the other oslo.messaging cores I would think.
15:18:00 <ad_rien_> I'm unfortunately already involved in different presentations for Boston, So I would rather prefer to not be to overwhelmed.
15:18:32 <ad_rien_> kgiusti:  agree but proposing something is the first step.
15:18:34 <ad_rien_> I think
15:18:43 <jamemcc> Logistically it is a block of rooms for 3 days straight, you look up the schedule and go to the next one you are most interested in.
15:18:55 <ad_rien_> wo can make a draft (deadline is next sunday)?
15:19:11 <rcherrueau> jamemcc: OK thanks
15:19:26 <ad_rien_> But AMQP what can be discussed?
15:19:50 <ad_rien_> No proposal?
15:20:07 <ad_rien_> ok seems not
15:20:10 <kgiusti> ad_rien_: In my opinion there's not a consistent story regarding the use of backends
15:20:31 <ad_rien_> kgiusti:  can you elaborate a bit more ?
15:20:33 <ad_rien_> please
15:20:49 <kgiusti> ad_rien_: e.g. - when asked "which backend should I use" there's no official guidelines
15:20:55 <ad_rien_> exactly
15:21:20 <ad_rien_> but there are efforts through the architecture WG to try to identify common tools
15:21:21 <kgiusti> ad_rien_: we talked about this at the last summit, but there were not enough interested parties to come to a conclusion
15:21:52 <ad_rien_> Maybe if we suceed to get sufficient content we can animate a discussion and finally make progress
15:22:06 <kgiusti> ad_rien_: simply getting a consensus regarding the purpose of the different backends would be ideal
15:22:15 <paul-andreraymon> is there a discussion to be had on a reference AMQP architecture for Massively distributed?
15:22:28 <ad_rien_> yes paul-andreraymon
15:22:33 <ad_rien_> at least we want to identify requiremetns
15:22:47 <kgiusti> ad_rien_: so I can propose to the oslo folks that we're planning some discussion around that in particular
15:22:51 <ad_rien_> in order to be able to investigate whether this brick ou this one satisfies the requirements
15:22:57 <ad_rien_> kgiusti:  great
15:23:04 <ad_rien_> can you come back to us shortly
15:23:20 <ad_rien_> so if it makes sense to submit something we can do it by sunday
15:23:46 <ad_rien_> (even better, if you can come with a proposal like the curtis one but that focus on the oslo.messaging question, it would be perfect)
15:24:00 <ad_rien_> Then we need to identify moderators that can help you on this action.
15:24:06 <ad_rien_> kgiusti:  what do you think?
15:24:08 <kgiusti> ok
15:24:11 <ad_rien_> Great thanks
15:24:20 <ad_rien_> ok so briefly two more minutes
15:24:26 <paul-andreraymon> kgiusti: let me know if you need help
15:24:31 <ad_rien_> are there any other news from your side?
15:24:44 <ad_rien_> or we can move to the next topic
15:24:45 <ad_rien_> ?
15:25:09 <ad_rien_> ?
15:25:17 <ad_rien_> ok so let's move to the next topic
15:25:21 <ad_rien_> #topic AMQP
15:25:46 <ad_rien_> Can we have an overview of the current discussion? I saw several emails (thanks BTW ;))
15:25:50 <ad_rien_> rcherrueau:  kgiusti ?
15:25:56 <rcherrueau> Let me try to sum up discussions we started around this topic
15:27:06 <rcherrueau> So we start an informal discussion with the aim to get a good understanding of the specification of oslo_messaging
15:28:14 <ad_rien_> so much works :)?
15:28:55 <kgiusti> most of the conversation is to define the aspects of the high level API
15:28:57 <rcherrueau> So we try to categorise the different oslo_messaging feature (ie. rpc.call, rpc.cast, notify ...) regarding to three properties that commonly arise when you wanna compare messaging paradims
15:29:09 <kgiusti> rcherrueau: +1
15:29:18 <ad_rien_> ok cool
15:29:52 <rcherrueau> Lemme put the three properties into the pad
15:30:06 <ad_rien_> (let's consider that we have five minutes for getting a summary: what has been discussed and what are plans for the next two weeks)
15:30:28 <rcherrueau> see line 410
15:32:02 <rcherrueau> The aim of this work is the following: we need to correctly understand what oslo_messaging is supposed to do, before discussing/understanding the different backend (RMQ/ZMQ/QPid ...)
15:32:40 <ad_rien_> ok
15:33:54 <ad_rien_> (mabderrahim: do you know whether monasca is using oslo.messaging ? probably a naive question but in case…., I know it leverages kafka but I would like to know whether it is using kafka through oslo.messaging or any other API?)
15:33:59 <rcherrueau> ad_rien_: I could write a sum up of the discussion for the next meeting if you want.
15:34:05 <ad_rien_> yes please
15:34:29 <paul-andreraymon> We have had a good start on functional overview. It would be good if we could get some data on which features are most used. That discussion did not happen yet.
15:34:31 <ad_rien_> #action rcherrueau prepare a summary for the next meeting (i.e. put it in the pad thanks)
15:34:31 <mabderrahim> Monasca uses Kafka directly
15:34:42 <ad_rien_> mabderrahim:  thanks
15:34:46 <mabderrahim> it does not rely on an other api
15:34:52 <mabderrahim> welcome
15:35:14 <ad_rien_> paul-andreraymon:  good point
15:35:33 <paul-andreraymon> How could we move forward on that? any suggestions?
15:35:34 <ad_rien_> Maybe ENOS (that I hope will be discussed later on) can help us to get such information
15:35:43 <rcherrueau> paul-andreraymon: yes you are. I wanna run some static analysis of the code for that.
15:35:51 <ad_rien_> rcherrueau:  do you agree?
15:36:07 <rcherrueau> ad_rien_: yep
15:36:15 <paul-andreraymon> +1
15:36:37 <ad_rien_> ok so,  leveraging the overview rcherrueau is going to do, it would be great to invite the oslo.messaging core developers and get their feedbacks
15:36:46 <ad_rien_> ok anything else on that point ?
15:37:07 <kgiusti> I'll reach out and invite them
15:37:13 <ad_rien_> ok thanks
15:37:25 <ad_rien_> so maybe we can have a brief overview regarding the AMQP stresser
15:37:37 <ad_rien_> jfpeltier:  can you give us a few elements regarding the stresser ?
15:38:08 <jfpeltier> yes tests are still in progress, I plan to do a presentation tomorrow with Abdelhadi
15:38:44 <jfpeltier> For the time being we don't manage to separate stress to the Rabbimq controller from stress on the newtworking part
15:38:45 <ad_rien_> ok is the code on github ?
15:38:59 <jfpeltier> yes I send link
15:39:09 <paul-andreraymon> jfpeltier: in what forum are you making this presentation?
15:39:20 <jfpeltier> https://github.com/Carroll/osnoise
15:39:24 <ad_rien_> sorry paul-andreraymon
15:39:29 <jfpeltier> Presentation is slides/discussion
15:39:30 <ad_rien_> an internal meeting we have tomorrow
15:39:48 <ad_rien_> we will push the links toward slides in the pad
15:39:56 <ad_rien_> so everyone can see them by the next meeting
15:40:03 <paul-andreraymon> Thanks!
15:40:16 <ad_rien_> (sorry for the missunderstandings)
15:41:08 <ad_rien_> #link https://github.com/Carroll/osnoise AMQP stresser tools (developed by OL)
15:41:22 <jfpeltier> the good news is that we manage to put a lot of stress on amqp contrller
15:41:49 <ad_rien_> ok
15:42:13 <jfpeltier> the bad news are that it may not be very representative of real massively distributed solution where stress is not put on a single component
15:42:33 <jfpeltier> (nor link)
15:42:40 <ad_rien_> I hope we wil be able to discuss a bit more this tool next time (and see how it can be integrated in a framework link Enos)
15:42:49 <jfpeltier> we want to discuss this in the presentation
15:42:53 <ad_rien_> ok
15:43:02 <ad_rien_> so I propose to move forward
15:43:09 <ad_rien_> if there is no more question on that point.
15:43:10 <ad_rien_> ?
15:43:29 <ad_rien_> seems not
15:43:31 <ad_rien_> ok
15:43:32 <ad_rien_> #topic ECOMP
15:43:47 <ad_rien_> Jamey put the link in the etherpad for the doodle
15:44:22 <ad_rien_> Please all if you are interested by the presentation
15:44:28 <ad_rien_> fit the doodle by
15:44:42 <jamemcc> Thanks Adrien
15:44:48 <ad_rien_> let's say by friday ?
15:45:03 <jamemcc> Good so I can schedule by Monday Morning
15:45:14 <ad_rien_> jamemcc:  do you think it would be possible for you to extend the doodle now so that every one can fill it asap
15:45:15 <ad_rien_> ?
15:45:15 <jamemcc> UST
15:45:59 <jamemcc> YEah - I did - shoudl show through APril 5 now
15:46:23 <ad_rien_> ok thanks
15:46:40 <ad_rien_> ok so I propose we move forward
15:46:47 <ad_rien_> unless there is particular question?
15:46:52 <ad_rien_> regarding ECOMP?
15:47:16 <paul-andreraymon> Can we get a copy of the slides ahead of the presentation?
15:47:30 <jamemcc> Makes sense - will try
15:47:35 <ad_rien_> thanks
15:47:59 <ad_rien_> #jamemcc will send slides before the presentation
15:48:08 <ad_rien_> #action irc://irc.freenode.net:6667/#jamemcc will send slides before the presentation
15:48:36 <ad_rien_> #all please fill the doodle if you are interested by taking part in the discussion on ECOMP (please see the pad line 427)
15:48:48 <ad_rien_> #topic ENOS
15:49:01 <ad_rien_> @msimonin @rcherrueau ?
15:49:04 <ad_rien_> pleas go aehad
15:49:23 <rcherrueau> OK, lemme start with the link of the procjet
15:49:31 <rcherrueau> #link https://github.com/BeyondTheClouds/enos
15:49:56 <rcherrueau> So Enos is a tool for Performance analysis of OpenStack
15:50:08 <ad_rien_> #link  https://www.openstack.org/summit/boston-2017/summit-schedule/events/17952/toward-fog-edge-and-nfv-deployments-evaluating-openstack-wanwide (Enos Boston presentation)
15:50:44 <rcherrueau> it lets you deploy an OpenStack, then run some bench and finally get Network/CPU/RAM usage
15:51:33 <rcherrueau> Enos comes with a DSL, so that you can easily describe the topology of your OpenStack
15:52:31 <rcherrueau> Actually we support 3 providers to deploy OpenStack
15:53:12 <rcherrueau> g5k and chameleon that are research testbed
15:53:29 <ad_rien_> #link https://www.chameleoncloud.org Chameleon
15:53:31 <rcherrueau> and vbox (for development)
15:53:50 <ad_rien_> #link https://www.grid5000.fr/mediawiki/index.php/Grid5000:Home Grid'500 testbed
15:54:22 <rcherrueau> The fun part is that chameleon is operate through an OpenStack so with Enos you can deploy, more generally, OpenStack over OpenStack
15:54:36 <rcherrueau> and then run bench, ...
15:55:11 <rcherrueau> FYI, chameleon provides both baremetal and vm resources
15:55:45 <rcherrueau> We recently integrate osprofiler that produces traces of your OpenStack
15:56:37 <rcherrueau> So, I hope we could use this to understand better interaction that use the message bus.
15:57:06 <rcherrueau> ad_rien_: I don't know if we have time for question?
15:57:19 <ad_rien_> I think we should conclude
15:57:27 <ad_rien_> so please add content directly in the pad
15:57:34 <ad_rien_> #topic opendiscussion
15:57:47 <ad_rien_> (sorry for being restrictive but we run out of time)
15:58:06 <ad_rien_> so I just added a new etherpad for our face to face meeting in Boston and more generally to try to apply a divide and conquer strategy
15:58:13 <ad_rien_> so you will see for instance in
15:58:26 <ad_rien_> #link https://etherpad.openstack.org/p/Massively_distributed_wg_boston_summit face to face meeting in Boston
15:58:47 <ad_rien_> that I propose to list all rpesentations/sessions that make sense for our WG
15:59:23 <ad_rien_> by this way we will be able to try to cover/attend most of the presentations and then exchanges during our face to face meeting that is scheduled on Wednedsay afternoon
15:59:31 <ad_rien_> I will give your more details next time
15:59:38 <ad_rien_> Thanks all for joining the meeting
15:59:44 <ad_rien_> let's continue by mail or by the pad
15:59:44 <ad_rien_> thanks
15:59:52 <jamemcc> thanks
15:59:53 <ad_rien_> #endmeeting