15:01:09 <rcherrueau> #startmeeting fog-edge-massively-distributed-clouds
15:01:10 <openstack> Meeting started Wed Jan 31 15:01:09 2018 UTC and is due to finish in 60 minutes.  The chair is rcherrueau. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:01:11 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:01:13 <openstack> The meeting name has been set to 'fog_edge_massively_distributed_clouds'
15:01:15 <ad_ri3n_1> #chair  ad_ri3n_
15:01:19 <ad_ri3n_1> #topic roll call
15:01:22 <ad_ri3n_1> Hi guys
15:01:43 <ad_ri3n_1> let's see who will be there today
15:01:54 <rcherrueau> #chair ad_ri3n_1
15:01:55 <openstack> Current chairs: ad_ri3n_1 rcherrueau
15:01:59 <ad_ri3n_1> parus and a few folks already apologized
15:02:06 <rcherrueau> #topic roll call
15:02:19 <kgiusti> o/
15:02:26 <ad_ri3n_1> and actually I'm also travelling so the meeting should be rather short ;)
15:02:31 <ansmith> o/
15:02:31 <ad_ri3n_1> Hi kgiusti
15:02:42 <kgiusti> hello everyone!
15:02:46 <ad_ri3n_1> hi ansmith
15:03:08 <ad_ri3n_1> msimonin has some contraints and we (Alex me,…) are travelling
15:03:24 <ad_ri3n_1> So I don't think we will have a long meeting today ;)
15:03:33 <ad_ri3n_1> anyone else?
15:03:44 <ad_ri3n_1> #info agenda
15:04:00 <ad_ri3n_1> #link https://etherpad.openstack.org/p/massivfbely_distributed_ircmeetings_2018 (line 140)
15:05:12 <ad_ri3n_1> So maybe the most important one. We started a gap analysis of the different openstack deployment scenarios according to some features end-users/developpers of edge computing infrastructures may expect
15:05:28 <ad_ri3n_1> I will come back later as I would like to ask you to go and comment it.
15:05:55 <ad_ri3n_1> Regarding the edge sessions that are done simultaneously, the meeting has been cancelled yesterday
15:05:58 <ad_ri3n_1> so nothing new
15:06:23 <ad_ri3n_1> finally regarding the presentations I listed the ones I'm aware about it
15:07:04 <ad_ri3n_1> I'm not sure there will be other ones
15:07:21 <ad_ri3n_1> if you are aware of others please can you add them.
15:07:39 <ad_ri3n_1> kgiusti:  ansmith maybe some edge presentations from redhat have been already planned to be submitted ?
15:08:18 <kgiusti> I'm not specifically aware of any, but we'll ping the right people - ansmith?
15:08:20 <ansmith> I am not aware but will look into it
15:08:39 <ad_ri3n_1> ok thanks
15:09:23 <ad_ri3n_1> any news from your side ?
15:09:32 <msimonin> Hi all :)
15:09:40 <ad_ri3n_1> #topic ongoing-action-amqp
15:09:49 <ad_ri3n_1> kgiusti:  I saw you put a lot of text.
15:10:04 <ad_ri3n_1> please
15:10:08 <kgiusti> Yeah, spent some time working on a plan for dragonflow integration
15:10:10 <ad_ri3n_1> can you give us an overwiew?
15:10:33 <kgiusti> Yes - there's interest in adding oslo.messaging support for dragonflow project
15:10:40 <oanson> Hi. Sorry I'm late. I thought this was next week.
15:10:44 <kgiusti> url: https://wiki.openstack.org/wiki/Dragonflow
15:10:46 <ad_ri3n_1> Hi oanson no worry
15:10:47 <ad_ri3n_1> thanks
15:10:53 <kgiusti> hi oanson!
15:11:02 <ad_ri3n_1> (BTW the meeting today will be a bit shorter as most of us are travelling today)
15:11:03 <kgiusti> we're talking dragonflow btw
15:11:12 <ad_ri3n_1> kgiusti:  please go ahead ;)
15:11:13 <msimonin> Good timing oanson :)
15:11:15 <oanson> Great! I'll catch up now :)
15:11:17 <kgiusti> tl;dr
15:11:33 <kgiusti> oslo.messaging's current API is not exactly a good fit
15:12:12 <kgiusti> to emulate the API used by dragonflow's pub/sub clients in o.m. would be very heavyweight (in terms of resource use - queues, connections)
15:12:41 <kgiusti> now investigating different approach - possible have a bespoke AMQP 1.0 driver for dragonflow instead
15:12:52 <oanson> +1
15:13:07 <kgiusti> or a quick 'n' dirty POC over oslo.messaging to get an accurate sense of cost
15:13:15 <kgiusti> that's about it
15:13:41 <oanson> My understanding was that rabbitmq can't handle the distributed load we need?
15:13:47 <oanson> Or did I miss something?
15:14:10 <msimonin> kgiusti: you know I'm in for the new oslo.messaging api
15:14:33 <msimonin> I really think other project can benefit from it
15:14:39 <kgiusti> oanson: amqp + router may be able to scale higher
15:14:59 <oanson> Sure. That makes sense
15:15:08 <kgiusti> oanson: for a distributed use case, but msimonin et. al. is testing that ATM
15:15:30 <kgiusti> msimonin: yeah I think it would be very handy too
15:15:35 <oanson> I see. That's the ombt project?
15:15:36 <msimonin> rabbitmq is probably not a good fit in a massively distributed environment
15:15:41 <ad_ri3n_1> ok what are the actions to conduct? I guess kgiusti you willl discuss about this point during the PTG?
15:16:18 <kgiusti> ad_ri3n_1: 2 things: propose new API at oslo PTG, and build a POC for dragonflow to kick the tires a bit
15:16:34 <msimonin> kgiusti:  +1
15:16:51 <ad_ri3n_1> ok
15:16:51 <kgiusti> sadly next week I'm stuck in meetings so I probably won't get much real work started until week after :(
15:17:13 <ad_ri3n_1> yes it seems we are all busy with meetings and project submissions :(
15:17:15 <ad_ri3n_1> ok
15:17:21 <msimonin> kgiusti thanks for having done this effort on this !
15:17:28 <ad_ri3n_1> so maybe a new update next time?
15:17:43 <kgiusti> ad_ri3n_1: yep I think that's reasonable
15:17:50 <kgiusti> msimonin: my pleasure!
15:17:54 <ad_ri3n_1> (thanks msimonin  ;) a bit stressed today especially because I have to catch my train in 10 min max)
15:18:15 <msimonin> I can backup you ad_ri3n_1 if needed
15:18:41 <ad_ri3n_1> cool
15:18:42 <ad_ri3n_1> ;)
15:18:51 <ad_ri3n_1> I  was not sure you can attend today.
15:19:16 <ad_ri3n_1> OK maybe I can give a few words regarding the gap analysis, especially because I add a todo for myself regarding dragonflow
15:19:31 <ad_ri3n_1> kgiusti:  anything else? Can i move to the next topic?
15:19:43 <kgiusti> take the floor, sir!
15:19:49 <ad_ri3n_1> Thanks
15:19:50 <ad_ri3n_1> #topic ongoing-action-gap-analysis
15:20:15 <ad_ri3n_1> so we spent a couple of hours with Inria and Ericsson folks to produce the following etherpad
15:20:23 <ad_ri3n_1> #link https://etherpad.openstack.org/p/edge-gap-analysis
15:21:08 <ad_ri3n_1> so our goal was to conduct a gap analysis of the different deployment scenarios administrators can leverage to operate a middle edge infrastructure
15:21:41 <ad_ri3n_1> so first part we defined the topology we consider
15:22:08 <ad_ri3n_1> second part we defined expected features (i.e. operations you would to perform either as a developer/user of the edge infrastructure or as an administrator)
15:22:15 <ad_ri3n_1> We defined different levels actually 7
15:22:35 <ad_ri3n_1> The first level include the simplest operations such as provisioning a VM locally
15:22:47 <ad_ri3n_1> The last level address interoperability challenges
15:23:22 <ad_ri3n_1> in the third part, we analysed different scenarios (i.e. openstack vanilla, cells, regions, federations) according to the aforementioned expected features
15:23:41 <ad_ri3n_1> It would be great if you can go through the analysis and give us comments/feedbacks
15:24:15 <ad_ri3n_1> especiall oanson
15:24:30 <oanson> Sure. Will do.
15:24:30 <ad_ri3n_1> it would be nice to integrate the dragonflow proposal
15:24:40 <ad_ri3n_1> in the different deployment scenarios
15:24:44 <oanson> Would be great for us too :)
15:24:48 <ad_ri3n_1> for instance dragonflow and regions
15:24:53 <ad_ri3n_1> how this behaves?
15:24:58 <ad_ri3n_1> so either you can do that offline
15:25:08 <ad_ri3n_1> or you can take part to our next discussion which is scheduled friday
15:25:20 <ad_ri3n_1> Feb 9th
15:25:29 <ad_ri3n_1> at 15:30 Paris time
15:25:35 <ad_ri3n_1> that's all from my side ;)
15:25:51 <oanson> Sure. I'll let you know one way or the other.
15:25:58 <msimonin> What you mean ad_ri3n_1 is to get tenant network spanning different regions ?
15:26:00 <ad_ri3n_1> great
15:26:07 <ad_ri3n_1> msimonin:  yes
15:26:09 <ad_ri3n_1> Level 3
15:26:11 <ad_ri3n_1> features
15:26:53 <ad_ri3n_1> sorry Level 4 features
15:27:00 <msimonin> this would be great actually if oanson can give us some insight on "distributed" tenant network : tenant network over several regions/sites
15:27:06 <ad_ri3n_1> oanson:  adrien.lebre@inria.fr ;)
15:27:13 <ad_ri3n_1> msimonin:  +1
15:27:22 * ad_ri3n_1 must leave :(
15:27:27 <ad_ri3n_1> @msimonin can you please chair
15:27:31 <ad_ri3n_1> #chair msimonin
15:27:32 <openstack> Current chairs: ad_ri3n_1 msimonin rcherrueau
15:27:39 <msimonin> Ok :)
15:27:41 <ad_ri3n_1> I apologize but I really have to go
15:27:46 <ad_ri3n_1> thanks @msimonin
15:27:48 <oanson> msimonin, are talking one Openstack distribution across sites, or multiple distributions with e.g. l2gw?
15:27:56 <ad_ri3n_1> ++
15:28:18 <msimonin> I would say the pros/cons of both solutions
15:28:31 <msimonin> to address the level of features of the document
15:28:49 <msimonin> btw oanson are you going to the PTG ?
15:28:57 <msimonin> I probably missed this information
15:29:04 <oanson> yes on the PTG
15:29:20 <oanson> I'll check if I registered in the etherpad after the meeting
15:29:27 <msimonin> ok
15:30:02 <oanson> From the Dragonflow PoV, a single deployment is easiest. Easier to configure and manage, and the distributed DB (e.g. etcd, redis) and pub/sub takes care of synchronizing the controllers and hosts
15:30:02 <msimonin> Any questions regarding this topic  ?
15:30:33 <msimonin> oanson:  ack
15:31:31 <msimonin> I think we can move on the next topic
15:31:36 <msimonin> # open discussions
15:31:44 <msimonin> #topic open discussions
15:32:29 <msimonin> Anything to share, kgiusti, ansmith, oanson, rcherrueau  (or other people) ?
15:32:38 <oanson> Nothing from me
15:32:47 <ansmith> nothing at this point
15:32:52 <kgiusti> I'm done
15:33:26 <rcherrueau> me neither
15:34:17 <msimonin> Ok let's end this :)
15:34:43 <msimonin> Have a nice day/evening everyone !
15:34:47 <msimonin> #end meeting
15:35:00 <rcherrueau> #endmeeting