15:01:39 <serverascode> #startmeeting operators_telco_nfv
15:01:40 <openstack> Meeting started Wed Feb  8 15:01:39 2017 UTC and is due to finish in 60 minutes.  The chair is serverascode. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:01:41 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:01:44 <openstack> The meeting name has been set to 'operators_telco_nfv'
15:01:52 <serverascode> hi all, sorry was a bit late
15:01:53 <jamemcc> Hi
15:02:01 <serverascode> hi jamemcc :)
15:02:06 <jamemcc> No worries
15:02:25 <serverascode> anyone else here for the telecom/nfv meeting?
15:02:29 * PerfectChaos waves
15:02:38 <serverascode> hi PerfecChaos :)
15:03:02 <serverascode> we'll give it another min or so to see who shows up
15:04:13 <serverascode> ok so maybe just the 3 of us :) might be a nice short meeting
15:04:31 <serverascode> #topic Boston summit session
15:04:46 <serverascode> I don't have much to add here, I put in a session submission but won't hear back for a while
15:04:59 <serverascode> I know adrien did as well for the massively distributed group
15:05:07 <serverascode> jamemcc did you as well for LCOO?
15:05:17 <uli-k> hi!
15:05:26 <serverascode> hi uli-k :)
15:05:45 <jamemcc> Yes
15:05:48 <fengxia41103> hello
15:06:05 <serverascode> hi gengxia41103
15:06:14 <serverascode> hi fengxia41103 that is
15:06:15 <jamemcc> 2 one for general working group and another as a Birds of a Feather to try to introduce and make LCOO more approachable
15:06:17 <fengxia41103> hi.. being first time here
15:06:39 <fengxia41103> I'm a Lenovo engineer, just want to sit in
15:06:51 <serverascode> ok cool thanks jamemcc, will be good to keep an eye on everyones sessions and maybe work together a bit on it
15:06:54 <fengxia41103> and learn of this project :)
15:07:09 <jamemcc> agreed
15:07:26 <jamemcc> welcome fengxia41103
15:07:45 <fengxia41103> thank you jamemcc
15:07:58 <serverascode> I don't have anything else for summit session info right now
15:08:27 <serverascode> though I should mention they extended the call for presentations, so if you still want to put one in there is time
15:08:36 <serverascode> also they are managing the lightening talks insttead of the vbrown bag
15:08:45 <serverascode> so there is a lot of room to present
15:09:04 <ad_rien_> o/ (sorry for being late)
15:09:11 <serverascode> hi ad_rien_ :)
15:09:15 <jamemcc> Hi
15:09:20 <jamemcc> THanks for joining
15:09:33 <serverascode> ad_rien_ any thoughts on Boston summit sessions?
15:09:34 <ad_rien_> Hi (sorry last meeting was a bit longer than expected)
15:09:51 <serverascode> LCOO put in a request for a session, and a birds of feather
15:09:54 <jamemcc> This to some extent (at least for me) is where I get together with you both
15:10:15 <ad_rien_> ?
15:10:23 <ad_rien_> what do you mean?
15:10:26 <serverascode> #link https://etherpad.openstack.org/p/ops-telco-nfv-meeting-agenda
15:10:32 <ad_rien_> serverascode:  did you already ask for rooms?
15:10:41 <serverascode> fengxia41103 that is the agenda etherpad
15:10:49 <serverascode> I did already put in a submission yes
15:10:57 <fengxia41103> Got it, thank you serverascode
15:11:16 <ad_rien_> From my side, I sent an email to Erin (but didn't get any answer yet)
15:11:30 <jamemcc> I meant that for me this is a good touch base point between LCOO and Ops-Telco-NFV and Massive Scalable
15:11:37 <ad_rien_> It was unclear how to perform such a request (i.e. room for face-to-face meetings)
15:11:52 <serverascode> ad_rien_ my impression was you actually put in a request through the CFP system
15:12:14 <serverascode> but email might work too
15:12:26 <ad_rien_> I did both so wait and see.
15:12:32 <jamemcc> Adrien, I can walk you through it if desired- you do it just as if you were requesting a real presentation at the Summit
15:12:32 <serverascode> ok cool
15:12:40 <jamemcc> ok
15:13:06 <ad_rien_> jamemcc:  sorry your point is that we can take part to the BOF of LCOO? so we can discuss possible collaborations between our WG?
15:13:11 <ad_rien_> s/WG/WGs/
15:13:22 <jamemcc> Yes - that's right
15:13:39 <jamemcc> Though I think we are also doing that here and we'll get to that today
15:13:48 <ad_rien_> I think it is valuable definitely
15:14:09 <serverascode> ok that is a good segue :)
15:14:16 <serverascode> #topic Discuss LCOO a bit
15:14:20 <jamemcc> just a sec
15:14:26 <jamemcc> ok - related and a little more near term - the Ops Meetup - Did we cover that before
15:14:31 <ad_rien_> Actually it seems that several (important) members from the community request/agree that having a shared sessions for all WG chair makes sense.
15:14:45 <ad_rien_> I'm wondering who is going to arrange/animate such a session
15:15:03 <serverascode> I'll put the ops meetup on the agenda
15:15:07 <serverascode> we haven't discussed it much recently
15:15:39 <ad_rien_> BTW jamemcc I'm in touch with Abdelhadi Chari that is involved in the Massively Distributed WG (i.e. a few engineers from Orange actually but I'm used to work with Abdelhadi)
15:15:52 <serverascode> let's just chat about LCOO briefly what with all the recent emails :)
15:16:00 <ad_rien_> :)
15:16:07 <jamemcc> Great - thanks for connecting the thread - I didnt' know that
15:16:20 <jamemcc> Ok - brief is ok
15:16:49 <serverascode> I just see some added sections in the agenda, presumably you added those jamemcc?
15:17:02 <jamemcc> Yes - just a few minutes before the meeting
15:17:07 <serverascode> if you want to discuss any of those points just go ahead
15:17:10 <ad_rien_> (Actually we are working with Orange since three years on revising OpenStack to operate Fog/Edge infrastructres, i.e. http://beyondtheclouds.github.io)
15:17:38 <jamemcc> So the first thing and maybe this is the brief part - Any advice/insight as to the questions that have come up?
15:18:05 <ad_rien_> Which ones (IOW, which are the most important from your point of view)?
15:18:41 <jamemcc> I shoudl ahve had the links ready, let me know and I can find them in a minute or 2
15:18:42 <serverascode> It is an interesting situation, I think it's kind of a first for openstack
15:19:04 <serverascode> most of the groups have been based on individuals
15:19:10 <jamemcc> Good questions asked about what is this new WG and why does it seem so somilar to Telco NFV and Massive Scalable
15:19:27 <ad_rien_> Jay is still challenging every change
15:19:44 <serverascode> he hasn't emailed for a while
15:19:49 <serverascode> but his points are all good
15:19:51 <ad_rien_> I think this is the way he uses to make things progressing
15:20:15 <ad_rien_> There are overlaps between NFV and Massively distributed definitely
15:20:16 <serverascode> I think his major thing is to make sure that the core openstack components are considered
15:20:16 <jamemcc> exchanges taht did seem to push forward
15:20:39 <ad_rien_> serverascode:  you mean?
15:20:42 <serverascode> eg the gluon project vs neutron
15:20:48 <ad_rien_> ok
15:21:01 <serverascode> improving core systems as opposed to writing something new
15:21:19 <ad_rien_> sometimes he made some tweets on the fact that OpenStack will dissapear in favor of OpenNFV ;)
15:21:32 <ad_rien_> but actually this is our position in the massively distributed WG
15:21:36 <serverascode> yeah I read those as well
15:21:47 <ad_rien_> and he made also some remaks a few months ago
15:22:07 <ad_rien_> and it is clear in the definition of the WG
15:22:12 <ad_rien_> OpenSTack and Upstream first
15:22:20 <ad_rien_> the issue is that it takes times
15:22:36 <jamemcc> I guess the fer is that a group like LCOO might be creating new things in a vacumn and seems to me the actionable aspect of that for us here is to make sure we aren't adding to that dynamic with multiple WG and in fact helping to bring more parties together
15:22:36 <ad_rien_> OpenStack has been historically designed/developed to operate single site cloud infrastructure
15:23:07 <serverascode> jamemcc: right
15:23:21 <ad_rien_> Now academics and industry experts agree that clouds should move to a more distributed model. The question is that should we do it with OpenStack or not
15:23:36 <ad_rien_> Doing without (i.e. developping from sratch) looks to be a non sense
15:23:54 <serverascode> ad_rien_ good point, the massively distributed concept is contentious as well :)
15:24:02 <ad_rien_> but at the same time we need to have some folks from the core developers that can hear our requests/whishes
15:24:16 <ad_rien_> otherwise, there will be a fork for sure
15:24:36 <serverascode> yes that is a difficult issue
15:24:49 <ad_rien_> We spoke/exchanged a lot with ttx
15:25:42 <ad_rien_> but we need to find
15:25:46 <ad_rien_> additional persons
15:26:01 <ad_rien_> that argue for such a change in the design
15:26:15 <ad_rien_> i.e. ensuring that OpenStack core services can cope with WANwide deployments
15:26:33 <ad_rien_> IT seems that now more and more persons are interested by that
15:26:44 <ad_rien_> and not only for NFV/SDN use-cases
15:26:57 <ad_rien_> (don't get me wrong, NFV/SDN use-cases are valuable too ;))
15:27:29 <ad_rien_> I just say that we need additional use-cases that can justify the needs of having clouds operated by the same provider but geographically disrtibuted
15:27:30 <serverascode> ok, there is a lot of change afoot :)
15:27:47 <serverascode> but we should stick to advice for LCOO if any for this topic
15:27:53 <serverascode> anyone have any advice to share?
15:28:24 <serverascode> I will just say 1) work upstream 2) work with core components as much as reasonable 3) use openstack community tools as much as possible
15:28:36 <uli-k> I think it is not only about "operated by the same provider".
15:29:03 <ad_rien_> I  think a valuable goal is to identify what are the missing components or at least the core-services that does not cope with requirements.
15:29:16 <ad_rien_> and to have strong arguments to say why (for instance performance experiments).
15:29:20 <uli-k> It is about sharing the resources of different sites, so somebody can assign workloads flexibly between sites or tie them to a specific site.
15:29:35 <ad_rien_> uli-k: I agree but this makes the big picture more complex
15:29:46 <uli-k> I know :(
15:30:03 <ad_rien_> my idea/vision (that can be wrong) is something like TPC/IP two decades ago
15:30:20 <ad_rien_> so if OpenStack can natively cooperate like TCP/IP does
15:30:38 <ad_rien_> then each time you plug two openstack deployment they will be able to colloborate in a native manner
15:30:50 <jamemcc> Thanks Curtis
15:31:00 <ad_rien_> then we can be sure that peering agreement between operatos etc…. will come to the surface
15:31:17 <jamemcc> Adrien, That's a nice simple explanation/start of a user story
15:31:28 <ad_rien_> AFAIK right now, this is rather not evident to operate a significant number of small DCs
15:31:57 <serverascode> that sounds more like federation than distributed
15:32:12 <ad_rien_> you can use the regions to operate a few DCs but if you consider to deploy a small DC (let's say two racks) in each Network of Point of Presence, then it will become a nightmare to operate that with the region concept
15:32:39 <ad_rien_> maybe the cells concept can be better (especially with the revision of cell V2) but this is not clear yet… in particular because the cell is just for Nova :(
15:33:06 <matrohon> ad_rien_, if you refer to AS interconnection, eBGP would be a better model I think :)
15:33:13 <ad_rien_> it depends.. do you want to have a glance repository in every site.
15:33:26 <ad_rien_> matrohon:  yes
15:33:35 <serverascode> ok, unfortunately we should probably move on in our agenda
15:33:43 <serverascode> very interesting topic though
15:33:49 <jamemcc> IN terms of LCOO next steps we are trying to form functional specialty teams (SME teams) in the areas/groups I listed in the Etherpad.  We'll try to be bringing user stories out of thise teams and in time for the Boston Forum and where applicable (cross project and cross release) through PWG User Story process.
15:34:25 <ad_rien_> jamemcc:  what kind of user stories are you expecting for intance for the container item .
15:34:26 <jamemcc> As User Stories that seem related to Telco/NF or assive Scalable - I'll bring here as well
15:34:26 <ad_rien_> ?
15:34:33 <serverascode> are you planning on the user stories first, then to start actual coding projects?
15:34:55 <jamemcc> The main one would be to spin up and manage the control plane
15:35:51 <jamemcc> The details of the User Story would be what personas need what out of all that
15:36:27 <jamemcc> Seems very reaalted to scalable too
15:36:59 <serverascode> what's LMA stand for?
15:37:00 <jamemcc> Just an example - I'm not the one per se to submit the User Story
15:37:54 <jamemcc> LMA is this https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0ahUKEwjbqs2k6oDSAhXkjFQKHW7DDNAQFggoMAA&url=https%3A%2F%2Fwww.mirantis.com%2Fblog%2Fmos-7-0-logging-monitoring-alerting-lma-enhancements%2F&usg=AFQjCNFVNx6V_sBxfiXqiXhUc2C3PI-6jQ&sig2=A_5hHwUIO7pTCDZTLYpxEQ
15:37:57 <jamemcc> oops
15:38:13 <serverascode> ah ok
15:38:19 <jamemcc> https://www.mirantis.com/blog/mos-7-0-logging-monitoring-alerting-lma-enhancements/
15:39:01 <jamemcc> But the team is more about that subject area - kind of a bad example as it seems to suggest the team has alredy chose
15:39:06 <serverascode> and is the plan to work on user stories first, then to start to understand what devleopment work needs to be done?
15:39:10 <jamemcc> n the project
15:39:53 <jamemcc> Yes
15:40:21 <serverascode> ok, sounds good to me :)
15:40:34 <serverascode> I think you might get some questions around the sub groups
15:41:05 <serverascode> as there are some existing working groups in those areas, but not all, but I could forsee questions on that
15:41:49 <serverascode> any other thoughts on this topic? should perhaps move onto the milan ops meetup
15:42:11 <jamemcc> ok to move on - thanks for feedback
15:42:24 <serverascode> #topic Milan Operators Meetup
15:42:25 <jamemcc> Adrien - I'll pass on the notes you put in to those teams as they get started
15:42:37 <serverascode> I should first say that I'm not sure I will be attending
15:43:01 <ad_rien_> jamemcc:  not sure I understood, are you expecting some actions from my side?
15:43:06 <serverascode> jamemcc are you going? are others from the LCOO going?
15:43:13 <jamemcc> Yeah - not clear for me either to attend in Milan
15:43:20 <jamemcc> Just wanted to clear the air
15:43:22 <serverascode> I know shintaro from NTT will be there, or I think he will
15:43:22 <ad_rien_> serverascode:  I will not attend the Milan meeting
15:43:48 <serverascode> other than that I'm not sure how many ppl working in NFV related fields will be there
15:44:15 <serverascode> ok so perhaps there is not much to discuss here :)
15:44:19 <jamemcc> ok - Adrien - no action needed form you - until the team (assumign we can get started) wants to collaborate - then we'd reach out to you
15:44:33 <ad_rien_> ok thanks for the clarification
15:45:09 <serverascode> ok if no other thoughts on that then next topic...
15:45:23 <serverascode> #topic Generic NFV Platform Definition document
15:45:36 <serverascode> #link https://etherpad.openstack.org/p/generic-nfv-platform
15:45:58 <serverascode> I am still just entering sections and occasionally writing in that document, and I will just keep doing so over the next while
15:46:11 <serverascode> please feel free to add or edit or suggest or critique :)
15:47:54 <serverascode> any thoughts on that at all?
15:47:55 <ad_rien_> Is CloudRan a NFV use-case?
15:48:05 <ad_rien_> If yes I think we should invite some Orange folks
15:48:33 <ad_rien_> Because we had some discussions and they have a use-case that looks really interesting to illustrate the interest of a disitrbuted infrastructure
15:48:40 <serverascode> I'm not super familar with cloudran, what exactly is that?
15:48:43 <ad_rien_> with all advanced features you can envision
15:48:49 <ad_rien_> matrohon:  still there ?
15:49:01 <matrohon> yep
15:49:05 <ad_rien_> you want to give a few words on that or should I?
15:49:22 <matrohon> cloudran is the same as vRAN :)
15:49:26 <ad_rien_> ok
15:49:28 <ad_rien_> thanks:D
15:49:50 <matrohon> the ability to manage compnent of 4G/5G design in VM
15:49:57 <ad_rien_> briefly the idea is to virtualize as much features as possible
15:50:02 <matrohon> from antenna component to backends
15:50:12 <serverascode> what kind of special features does that typically require? real time kernel?
15:50:21 <ad_rien_> some features cannot be deployed in centralized clouds due to latency constraints
15:50:29 <ad_rien_> whereas other can be put higher in the infrastructure
15:50:43 <matrohon> some of those VM will need to be as close as possible to the antenna (law latency) whereas some other might move to a central DC
15:50:56 <ad_rien_> (lower = edge, ie. at the bottom of the radio base station, higher, means the traditional DC)
15:51:11 <ad_rien_> so ou have placement issues
15:51:14 <ad_rien_> network issues
15:51:16 <ad_rien_> etc..
15:51:36 <ad_rien_> because some VMs should be interconnected (i.e. they exchange messages) while not being located on the same site
15:51:43 <serverascode> so cloudran would require a distributed hypervisor type setup?
15:51:48 <ad_rien_> so you have wanwide-related issues
15:52:02 <ad_rien_> if by hypervisor you mean openstack yes
15:52:07 <ad_rien_> if by hypervisor your mean KVM
15:52:11 <ad_rien_> the answer is no
15:52:24 <serverascode> so a separate cloud in the radio base station?
15:52:33 <ad_rien_> there is different scenarios
15:52:51 <ad_rien_> you can have only compute nodes deployed at the bottom of radio base stations
15:53:02 <ad_rien_> and the control services in one centralized cloud
15:53:05 <serverascode> I think for this "generic NFV platform" it would probably just start with one single openstack deployment
15:53:09 <ad_rien_> so only one openstack to operate this infrastructure
15:53:14 <ad_rien_> exactly
15:53:22 <ad_rien_> here you have only one openstack
15:53:34 <serverascode> ok yeah then that is a potential use case sure
15:53:51 <ad_rien_> (please keep in mind that in our vision, i.e. the massively distribetud WG, there is/will be always one OpenStack)
15:53:53 <serverascode> is there one of those use-case diagrams for this model? I know you had a bunch of diagrams
15:53:59 <ad_rien_> yes
15:54:06 <ad_rien_> let me find the url
15:54:32 <serverascode> the only problem is that the massively distributed doesn't really exist yet, in my mind, right?
15:54:46 <matrohon> a use case that needs to scale to potentially thousand hundreds of compute...
15:55:03 <serverascode> there's no code in openstack yet to do that
15:55:08 <ad_rien_> #link https://docs.google.com/presentation/d/1jJFZejZqgYDxu5FX4K8g3I5zQ87afnjYI4VSRSuCQ6U/edit#slide=id.p different manner to operate distinct DCs
15:55:22 <serverascode> (5 min letf)
15:55:24 <ad_rien_> all the architecture deployment that are presented in these google slides
15:55:27 <ad_rien_> are valid
15:55:32 <ad_rien_> i.e. you can deploy them now
15:55:36 <ad_rien_> and it will run
15:55:47 <serverascode> ok, will take a look :)
15:55:49 <ad_rien_> the question is that as matrohon mentioned, the scalability
15:55:51 <ad_rien_> issue
15:55:53 <ad_rien_> the latency issue
15:56:06 <ad_rien_> some features that do not run well in a WAN context
15:56:08 <ad_rien_> etc…
15:56:23 <ad_rien_> the main issue is that when we discuss with core developer of Openstack
15:56:38 <ad_rien_> they argue that you can already deal with the use-case we are discussing
15:56:50 <ad_rien_> by selecting one of the archtecture illustrated in those slides.
15:57:08 <ad_rien_> what we would like to do is to identify the issues/barriers and show them that they are wrong (or right ;))
15:57:46 <serverascode> ok, I'd like to leave the last couple of min for open discussion
15:57:50 <serverascode> #topic open discussion
15:57:56 <serverascode> anyone have anything they would like to add?
15:58:01 <serverascode> PerfectChaos?
15:58:15 <serverascode> uli-k?
15:58:19 <ad_rien_> serverascode:  may I ask you whether you can clarify the difference between distributed and federated (i.e. not now but maybe we can discuss it next time)
15:58:42 <serverascode> yeah we will have to discuss that at some point
15:58:49 <ad_rien_> thanks
15:58:52 <serverascode> especially if our presentation is selected :)
15:58:58 <ad_rien_> :-P
15:59:07 <serverascode> fengxia41103?
15:59:30 * uli-k happy if we discuss CloudRan, but also no expert.....
15:59:45 <serverascode> 20 seconds left :)
15:59:57 <PerfectChaos> I don't really have anything to discuss this time...
15:59:57 <serverascode> thanks everyone for this week, lots going on in the NFV and OpenStack world
16:00:04 <serverascode> #endmeeting