15:02:58 <ad_ri3n_> #startmeeting fog-edge-massively-distributed-clouds
15:02:58 <ad_ri3n_> #chair  ad_ri3n_
15:02:59 <openstack> Meeting started Wed Mar 14 15:02:58 2018 UTC and is due to finish in 60 minutes.  The chair is ad_ri3n_. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:03:00 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:03:02 <openstack> The meeting name has been set to 'fog_edge_massively_distributed_clouds'
15:03:04 <knikolla> o/
15:03:04 <openstack> Current chairs: ad_ri3n_
15:03:04 <ad_ri3n_> #topic roll call
15:03:08 <ad_ri3n_> Hi guys
15:03:23 <ad_ri3n_> sorry for being late… I was brainstorming on OpenStack (as usual ;))
15:03:24 <parus> Hello
15:03:27 <jdandrea> o/
15:03:33 <jamemcc> o/
15:03:34 <ad_ri3n_> so it seems that there is a couple of folks
15:03:37 <ad_ri3n_> great !
15:03:58 <ad_ri3n_> #info agenda
15:03:58 <ad_ri3n_> #link https://etherpad.openstack.org/p/massivfbely_distributed_ircmeetings_2018 line 260 Agenda
15:04:24 <ad_ri3n_> Please put your (nick)name in the etherpad
15:04:37 <ad_ri3n_> arg.. wrong link
15:04:40 <knikolla> https://etherpad.openstack.org/p/massively_distributed_ircmeetings_2018
15:04:44 <ad_ri3n_> @dpertin can you please copy/paste
15:04:48 <ad_ri3n_> thanks knikolla
15:04:48 <parus> US changed to summer time last week-end.... Lots of schedule confusion this side of the Atlantic.
15:04:53 <dpertin> #link https://etherpad.openstack.org/p/massively_distributed_ircmeetings_2018
15:05:12 <ad_ri3n_> So there is a couple of new folks
15:05:21 <ad_ri3n_> maybe we can take a few minutes to introduce them
15:05:50 <ad_ri3n_> knikolla: jdandrea ?
15:05:57 <ad_ri3n_> could you please introduce briefly yourself?
15:06:01 <ad_ri3n_> thanks
15:06:31 <knikolla> o/ hi all. I'm Kristi. keystone core and leading the resource federation efforts at the Mass Open Cloud. https://massopen.cloud/blog/mixmatch-federation/
15:07:01 <ad_ri3n_> great to see you there knikolla
15:07:06 <jdandrea> Certainly. Joe D'Andrea, working on all things cloud, previously seen around Heat stuffs.
15:07:17 <ad_ri3n_> the same jdandrea
15:07:21 <jdandrea> :)
15:07:29 <ad_ri3n_> thank you guys for joining the meeting
15:07:37 <jdandrea> You're welcome! Thanks for hosting it.
15:07:53 <ad_ri3n_> so let's start
15:08:00 <ad_ri3n_> #topic announcement
15:08:17 <ad_ri3n_> so you can see the major news in the pad
15:08:35 <ad_ri3n_> regarding the PTG, we have two sessions with a rather important number of participants
15:08:51 <ad_ri3n_> during the monday afternoon we went through the expected features we identified previously
15:09:22 <ad_ri3n_> We got a couple of previous comments and we are currently trying to consolidate all the inputs in a readable/concise document
15:09:46 <ad_ri3n_> On Tuesday, the openstack foundation join the discussion and we mainly discussed concrete issues ATT is facing
15:10:10 <ad_ri3n_> This enabled us to move to prototyping discussions regarding keystone and glance
15:10:32 <ad_ri3n_> regarding keystone, the first discussion focused on the possibility to define roles/… on a region basis
15:10:50 <ad_ri3n_> i.e. define rights according to each site for each user/project…
15:11:20 <ad_ri3n_> The idea is to prepare a well described request to se whether this concept of regions/locations can be reified at the level of keystone
15:11:42 <ad_ri3n_> The second part of the discussion dealt with golden services such as keystone or glance
15:12:16 <ad_ri3n_> The main idea is to be able to create for instance a new vm image through the golden glance and push the new VMI on a set of identified sites
15:12:31 <ad_ri3n_> how we are going to implement those concepts has not been discussed yet
15:12:54 <msimonin> push = make the image available
15:13:16 <ad_ri3n_> (actually I think this is the major challenge because those new golden services would probably be stateful and thus we may have to deal with inconsistency issues)
15:13:25 <ad_ri3n_> You can find further informations on the different link I put
15:13:55 <ad_ri3n_> The second new from my side is related to the Openstack edge sesions but I would like to discuss that point later in the agenda
15:13:59 <ad_ri3n_> so that's all from my side.
15:14:05 <ad_ri3n_> Any news guys from your side?
15:14:05 <msimonin> As far as I remember the discussion it looked like an active/passive model
15:14:27 <msimonin> the golden site was active (admin task can be performed)
15:14:34 <ad_ri3n_> ansmith:  msimonin do you want to add a few words regarding the presentation in the OSLO session?
15:14:50 <msimonin> the other site was passive (usable by the users to start vm  but not for admin tasks)
15:14:56 <msimonin> other sites*
15:15:21 <ad_ri3n_> thanks msimonin
15:15:25 <msimonin> ad_ri3n_: I can share some slides
15:15:30 <ansmith> the oslo members very much appreciated the update from msimonin on bus analysis
15:15:34 <ad_ri3n_> yes please c/p the link if you are ok
15:16:32 <msimonin> https://docs.google.com/presentation/d/1LcF8WcihRDOGmOPIU1aUlkFd1XkHXEnaxIoLmRN4iXw/edit?usp=sharing
15:16:54 <ad_ri3n_> #link https://docs.google.com/presentation/d/1LcF8WcihRDOGmOPIU1aUlkFd1XkHXEnaxIoLmRN4iXw/edit?usp=sharing preliminary presentation of the bus analysis
15:17:15 <ad_ri3n_> ok
15:17:17 <ad_ri3n_> anything else?
15:17:29 * jdandrea is perusing the slides (thank you!)
15:17:48 <ad_ri3n_> maybe we can add that the framework used to perform this experiment is going to presented during the CNERT workshop soon
15:18:03 <ad_ri3n_> @avankemp_ can you please c/p the link to the workshop
15:18:04 <ad_ri3n_> ?
15:18:10 <parus> My recollection is that ONAP uses some components of Openstack. is oslo being used by ONAP? does anyone know?
15:18:26 <avankemp_> #link http://infocom2018.ieee-infocom.org/content/workshop-cnert-computer-and-networking-experimental-research-using-testbeds
15:18:40 <ad_ri3n_> thanks avankemp_
15:18:56 <ad_ri3n_> parus:  not from my side. AFAIK ONAP only use the OS API
15:19:02 <ad_ri3n_> s/only use/only uses
15:19:09 <ad_ri3n_> confirmed by dpertin ;)
15:19:15 <jdandrea> ad_ri3n_ It's possible that oslo components are being used in a few spots. I can check. Possibilities include multicloud due to plugins that interface w/OpenStack.
15:19:30 <ansmith> ONAP defines common services around messaging but not oslo specifically
15:20:09 <ad_ri3n_> should we go on on that point?
15:20:44 <ad_ri3n_> jamemcc:  could you please refresh us a bit regarding LCCO WG?
15:21:33 <ad_ri3n_> jamemcc:  still there?
15:21:58 <ad_ri3n_> ok so let's start the agenda
15:22:12 <ad_ri3n_> #topic ongoing-actions OpenStack edge sessions
15:22:21 <jamemcc> Sorry
15:22:23 <jamemcc> looked away
15:22:31 <jamemcc> I'll update the pad
15:22:44 <ad_ri3n_> (thanks jamemcc) So we (Parus and I) took part to the meeting this morning
15:23:19 <ad_ri3n_> two or three additional folks coming from Asia have joined the call.
15:23:44 <ad_ri3n_> Jonathan briefly summarized the work that has been done so far
15:24:14 <ad_ri3n_> Prakash gave a couple of new pointers regarding new activities supported by ATT and LF
15:24:27 <ad_ri3n_> in particular that one:
15:24:29 <ad_ri3n_> #link https://www.akraino.org
15:24:53 <ad_ri3n_> I don't know what is behing but unfortunately yet another piece of code to keep an eye on it
15:25:40 <ad_ri3n_> at the end of the meeting we discussed a bit about the golden service approach.
15:26:00 <ad_ri3n_> Jonathan will propose to organize a telco dedicated to discuss a bit more this architecture and try to identify possible limitations
15:26:12 <ad_ri3n_> so let's keep an eye on the edge mailing list
15:26:21 <ad_ri3n_> that's all from my side on that
15:26:32 <ad_ri3n_> if there is no question I can switch to the next point?
15:26:43 <parus> There was a post on linked in by Kandan about this.
15:26:51 <parus> #link https://www.linkedin.com/in/kandank/detail/recent-activity/
15:26:59 <ad_ri3n_> thanks parus
15:27:05 <jdandrea> Thx!
15:28:40 <ad_ri3n_> #topic ongoing-actions P2P/gap analysis
15:29:27 <ad_ri3n_> Simultaneously of the edge discussions, we are still working with Ericsson folks to better understand the pros/cons of top/down vs bottom/up approaches
15:29:42 <ad_ri3n_> to develop the resource management system for an edge infrastructure
15:30:05 <ad_ri3n_> Although the foundation seems to be in favour of a top/down approach
15:31:20 <ad_ri3n_> we would like to be sure this is the right way to go (actually, I'm even not sure that they ''push'' a top/down approach to be honest. I would rather say that we agreed that a synchronisation mechanism is mandatory to synchronise the different services between the different regions but how this will be done is still unclear as discussed previously)
15:32:16 <ad_ri3n_> So the goal of this ongoing action is to identify possible ''no go'' issues that prevent either one or the other approach
15:32:33 <ad_ri3n_> We hope to have a short document (5 pages) to be presented soon.
15:32:49 <ad_ri3n_> We also arranged an internal meeting next week at Inria to discuss this question.
15:32:59 <ad_ri3n_> So more news during the next meeting hopefully
15:33:05 <ad_ri3n_> questions?
15:33:51 <ad_ri3n_> ok so let's move to the next point
15:34:03 <ad_ri3n_> #topic ongoing-actions keystone backend
15:34:12 <ad_ri3n_> @rcherrueau
15:34:48 <ad_ri3n_> is absent for a couple of days. @marie, who is an internship in our group work with rcherrueau and will update us about the current work
15:34:59 <ad_ri3n_> @marie the floor is yours  ;)
15:35:17 <marie> hi :)
15:36:13 <ad_ri3n_> so the goal is to compare from the performance viewpoint several approaches to share state accross distinct sites
15:37:30 <marie> we are testing different type of deployments for Keystone
15:38:27 <marie> with a common authentication, either with one centralized Keystone or a Keystone on each region with a common state
15:38:57 <marie> and also later wih a federated Keystone
15:39:01 <ad_ri3n_> are you considering the federated approach
15:39:06 <ad_ri3n_> ok… too late ;)
15:39:15 <jdandrea> (hehehe)
15:39:28 <ad_ri3n_> knikolla:  did you already perform such evaluations from your side?
15:39:49 <knikolla> ad_ri3n_: not yet
15:39:52 <ad_ri3n_> @marie is your experiment protocol already available or not yet?
15:40:01 <marie> We are comparing the performance for Galera and CockroachDB when we are varying the number of nodes and latency
15:40:05 <ad_ri3n_> it would be great to share it to get feedbacks
15:40:35 <msimonin> something in the performance doc would be a good place
15:40:51 <ad_ri3n_> knikolla:  is it a possible interest point for keystone core devs?
15:40:59 <ad_ri3n_> @msimonin +1 ;)
15:41:46 <knikolla> ad_ri3n_: the performance of keystone in a multiregion setup with cockroachdb vs galera?
15:41:55 <ad_ri3n_> vs federation
15:41:56 <ad_ri3n_> yes?
15:42:14 <ad_ri3n_> (cockroach or any other relevant storage backend)
15:42:28 <knikolla> ad_ri3n_: in the case of federation, you will not have a shared database. so the defining factor will be the time it takes to exchange an assertion for a token.
15:42:48 <knikolla> (keystone to keystone federation)
15:43:12 <ad_ri3n_> the question we would like to deal with is: what is the most suited manner to get a logical global view of a keystone across distinct locations/openstack instances
15:43:29 <ad_ri3n_> knikolla:  exactly
15:44:01 <ad_ri3n_> moreover with the edge context we have to deal with possible network split brain issues
15:44:54 <knikolla> which are tricky for databases, but work better when you don't need to share state. in the case of federation.
15:45:17 <ad_ri3n_> but yes this is exactly the idea (either we replicate states everywhere: galera like manner, or we leverage a keystone to keystone federation or we use a locally aware no/new SQL db such as cockroach)
15:45:44 <parus> I have a basic question (sorry to derail)
15:46:00 <ad_ri3n_> so what is your first opinion knikolla : keystone to keystone federation should be more suited for the edge context?
15:46:10 <ad_ri3n_> (parus please go on, feel free ;))
15:46:13 <parus> Keystone is used to store user credentials, but also to generate tokens. Those are two very different use cases, When distributing the database, can we separate both types of content?
15:46:41 <knikolla> i probably need to do a lot of background reading for what's being discussed and aimed for here, as this is my first time joining the meeting.
15:47:06 <knikolla> i tend to favor the federation approach, personally. but as i just said above, I need to read more about your requirements.
15:47:30 <ad_ri3n_> ok
15:47:40 <knikolla> parus: i would say you can.
15:47:44 <parus> My question is hinting on : is federation appropriate for User credentials but maybe not for token.
15:48:23 <ad_ri3n_> I was just wondering what was the current choice for the federation efforts.
15:49:06 <ad_ri3n_> knikolla:  any pointer to share regarding current discussions on those federations' related  questions
15:49:08 <ad_ri3n_> ?
15:50:15 <knikolla> ad_ri3n_: parus: depends on how you're going to handle authorization (user access to projects)
15:50:17 <ad_ri3n_> ok maybe you can tell us more next time
15:50:46 <ad_ri3n_> knikolla:  if you have a link toward a wiki page/some minutes of your brainstorming sessions, It would be great if you can share it.
15:50:56 <ad_ri3n_> (10 minutes left, sorry we have to move to the next point)
15:51:05 <ad_ri3n_> #topic ongoing-actions AMQP evaluations
15:51:16 <ad_ri3n_> ansmith:  msimonin: please ?
15:52:51 <ad_ri3n_> msimonin:  ansmith nothing to add regarding all the crazy work you are doing :-P ?
15:52:59 <ansmith> not much to update as kgiusti and msimonin were off last week, starting up again this week
15:53:05 <ad_ri3n_> ok … next point :D
15:53:09 <msimonin> hey
15:53:11 <msimonin> yes sorry
15:53:27 <msimonin> We're scaling the framework
15:53:29 <ad_ri3n_> (I was also offline last week … winter vacations here in France for most of us ;))
15:53:32 <msimonin> This should be ready soon
15:53:32 * knikolla gotta run to another meeting. feel free to ping me on -keystone or via private message. i'd be happy to answer keystone/federation related questions or forward them to the broader keystone team.
15:53:47 <ad_ri3n_> knikolla:  thanks
15:53:48 <ad_ri3n_> sure
15:54:01 <ad_ri3n_> if you have any pointers please feel free to share it thanks
15:54:16 <ad_ri3n_> Ok thank msimonin
15:54:37 <ad_ri3n_> #topic open-discussions
15:54:43 <ad_ri3n_> 4 minutes left
15:54:56 <ad_ri3n_> so is there any point guys you would like to discuss?
15:55:30 <jdandrea> Not at this time. Absorbing "all the things."
15:55:40 <parus> Vancouver plans?
15:55:54 <parus> who is going?
15:56:02 <ad_ri3n_> jdandrea:  it would be great if you can share more informations regarding the work you are doing at ATT
15:56:17 <ad_ri3n_> I saw your emails, in particular the ones regarding the MUSIC project
15:56:46 <ad_ri3n_> but let's keep that for the next time (please feel free to add this to the agenda next time)
15:56:49 <jdandrea> ad_ri3n_ Yes! I am not working directly on music, but a colleague is. It has now been moved under ONAP.
15:56:50 <ad_ri3n_> @parus yes
15:56:51 <jdandrea> Sounds good.
15:56:58 <ad_ri3n_> vancouver…
15:57:06 <ad_ri3n_> so we are waiting to see which presentations have been selected
15:57:15 <ad_ri3n_> if everything is ok
15:57:23 <ad_ri3n_> we should come with a couple of folks from  Inria
15:57:43 <parus> one of mine has been accepted... so I will be going :-)
15:57:44 <ad_ri3n_> the other points to discuss is wether we want more than one slots
15:57:54 <ad_ri3n_> you already received your email ?
15:58:01 <ad_ri3n_> (fuck… ;) )
15:58:07 <parus> It was updated on web site.... no email yet.
15:58:10 <ad_ri3n_> (maybe this mean none from Inria was accepted)
15:58:13 <ad_ri3n_> ok
15:58:20 * jdandrea crosses fingers for Inria
15:58:25 <ad_ri3n_> so getting back to my first comment
15:58:40 <ad_ri3n_> maybe we should ask for more than one slot if we are enough people to come
15:59:03 <ad_ri3n_> acutally I really would like to dive into details about the possible architecture with the goal of preventing us to reinvent the wheel once again
15:59:12 <ad_ri3n_> so let's discuss that point next time
15:59:13 <ad_ri3n_> time to conclude
15:59:20 <ad_ri3n_> thanks guys for joining the meeting
15:59:29 <ad_ri3n_> let's keep in touch on the pad or through the different MLs.
15:59:30 <ad_ri3n_> Thanks
15:59:36 <marie> thanks
15:59:37 <ad_ri3n_> #endmeeting