14:00:37 <mkoderer> #startmeeting telcowg
14:00:37 <openstack> Meeting started Wed Jan 28 14:00:37 2015 UTC and is due to finish in 60 minutes.  The chair is mkoderer. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:38 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:41 <openstack> The meeting name has been set to 'telcowg'
14:00:53 <mkoderer> hello everybody
14:00:57 <aveiga> hello
14:01:00 <vks> hi
14:01:00 <smazziotta> hi
14:01:02 <matrohon> hi
14:01:02 <mkoderer> #topic roll call
14:01:03 <adrian-hoban> Hi
14:01:05 <cloudon> hi
14:01:14 <mkoderer> ok so whos around?
14:01:19 <mkoderer> #link https://etherpad.openstack.org/p/nfv-meeting-agenda
14:01:24 <smazziotta> hi
14:01:26 <mkoderer> ^the agenda of today
14:01:52 <mkoderer> all right, let's get started
14:01:57 <mkoderer> #topic action items from last week
14:02:13 <beagles> hi
14:02:18 <ndipanov> hey
14:02:25 <mkoderer> I took the action item to write down requirments about openstack HA
14:02:42 <mkoderer> I didn't had the time but I created a etherpad ;)
14:02:47 <mkoderer> #link https://etherpad.openstack.org/p/telcowg-usecase-openstack-ha
14:03:02 <mkoderer> anyone can add stuff there
14:03:18 <mkoderer> I will fill it with some details during the week
14:03:55 <vks> mkoderer, how did this HA will handle the breakage of service chain?
14:04:33 <mkoderer> vks: ok so the scope that I had in mind was OpenStack service HA
14:04:46 <mkoderer> vks: but your right we possible need to think about that too
14:05:41 <vks> mkoderer, yeah thats true its ok for computes but i guess here focus should be what service will effect if that compute goes down
14:06:14 <cloudon> by "OpenStack service HA" do you mean ensuring the services provided by OpenStack (nova, neutron etc) are themselves HA, or that OpenStack provides HA assistance to network functions running as VMs in OpenStack e.g. VM monitoring & restart?
14:06:20 <mkoderer> vks: we have a draft for service chaning here
14:06:22 <mkoderer> #link https://etherpad.openstack.org/p/kKIqu2ipN6
14:07:25 <mkoderer> cloudon: your first point
14:07:36 <mkoderer> cloudon: but we can enhance the scope
14:07:40 <cloudon> thanks
14:07:52 <cloudon> think keeping scope restricted to 1st point will help
14:08:10 <vks> mkoderer, so focus is on Openstackservices ???
14:08:22 <vks> like nova, neutron
14:08:25 <vks> ???
14:08:27 <mkoderer> vks: yep
14:08:42 <mkoderer> vks: we have already many things to describe there
14:09:00 <mkoderer> and HA for service chaning is a bit fuzzy
14:09:12 <beagles> cloudon: concur... functionality that supports HA in VMs is really a different bag of tricks.
14:09:37 <mkoderer> correct
14:09:38 <vks> mkoderer, i agree but can we have one doc where the goals are enumerated?
14:09:44 <vks> for HA
14:09:57 <aveiga> I think it's important to remember that we have to crawl before we can get to the run part.  Going straight for service chain HA is a quite big bite to chew on
14:10:02 <mkoderer> vks: i will put it into that etherpad and make it clear
14:10:08 <vks> mkoderer, ok
14:10:24 <cloudon> aveiga: +1
14:10:35 <vks> aveiga, i agree
14:10:40 <matrohon> mkoderer : I undesrtood that the focus of HA was : what kind of services need to be span across multiple openstack DC...
14:10:42 <mkoderer> #action mkoderer define scope of OpenStack HA use case clearly
14:10:47 <matrohon> mkoderer : +1
14:11:02 <vks> mkoderer, +1
14:11:22 <vks> i think the sooner the finer detail gets enumerated better
14:11:52 <cloudon> matrohon: agree that's important but think that's different: NFV is pretty explicit about VNFs potentially spanning multiple VIMs (i.e. cloud instances) with the orchestrator managing all that; this is about making an individual OpenStack instance as available as possible (IMHO)
14:11:52 <mkoderer> matrohon: like supporting OpenStack multiregion and so on?
14:12:47 <mkoderer> let me bring my thought into the etherpad and then let's discuss more deeply
14:12:55 <matrohon> mkoderer : yes
14:13:02 <matrohon> mkoderer : fine
14:13:06 <mkoderer> all right
14:13:20 <mkoderer> #topic Use Cases
14:13:34 <vks> cloudon, but each instance will have complete n/w functionality???
14:13:44 <adrian-hoban> cloudon: +1, I think we agreed previously to focus on single DC deployments
14:14:12 <vks> adrian-hoban, yes even i agreed to focus on single DC deployment
14:14:17 <dalgaaf> Is HA about only about inter DC OpenStack services or also about HA within a DC or even fire compartment?
14:14:32 <mkoderer> adrian-hoban: yep, we write all use cases for single DC deployment and add the needed requirments for HA into the HA use case
14:14:35 <dalgaaf> I would say both is important
14:14:56 <vks> dalgaaf, its both but we should focus first on single site
14:15:06 <dalgaaf> ok
14:15:18 <mkoderer> ok back to the topic... "use cases"
14:15:36 <mkoderer> sgordon told me that we didn't have much progress in reviewing the existing ones
14:15:55 <aveiga> mkoderer: that's correct
14:16:22 <mkoderer> so would it help to have a usecase repo and use gerrit to review it?
14:16:28 <vks> mkoderer, i think if we focus on single case first, may be we will have gud idea
14:17:05 <aveiga> vks: we attempted that two weeks back
14:17:12 <aveiga> it didn't go so well
14:17:13 <cloudon> the concern about gerrit is barrier to entry for non-devs, particularly SPs
14:17:27 <mkoderer> I have the felling it's a tool issue... since reviewing in etherpad or inside IRC isn't very efficent
14:17:31 <aveiga> cloudon: we also don't have a place for a cross-project repo at the moment
14:17:40 <ybabenko> +1
14:18:04 <ybabenko> the problem is that we have not reached any serious progress so far on discussion of the use-cases
14:18:15 <mkoderer> cloudon: yep this is a valid concern...but anyway I suggest that we simply try it
14:18:24 <vks> aveiga, so if single use case is problem, how multiple use case will get studied
14:18:32 <matrohon> aveiga : what about openstack-specs?
14:18:44 <ybabenko> for example, on service chaining etherpad there are not comments: either the topic is not interesting (i can not imagine this) or our model of working in telcowg does not work
14:18:46 <matrohon> aveiga : for cross-project specs
14:18:48 <aveiga> vks: not saying we shouldn't, just saying doing it over IRC didn't work too well
14:19:02 <vks> aveiga, i agree
14:19:06 <aveiga> matrohon: we can try that, but like cloudon said none of the Telcos will go there
14:19:26 <mkoderer> aveiga: we will go there :P
14:19:28 <aveiga> well, I shouldn't make a blanket statement, since I work for a telco and we contribute code, but in most cases...
14:19:51 <mkoderer> so at least there are some telco that are already active in reviews
14:19:56 <matrohon> aviga : I agree, it is managed by the TC
14:20:32 <mkoderer> does somebody has a better idea to make the use cases review more efficent?
14:21:14 <aveiga> I think whatever method we use needs to get advertised on the operators mailing list
14:21:39 <mkoderer> aveiga: +1
14:22:08 <vks> aveiga, can we pick one general use case and advertise on operators mailing list?
14:22:25 <matrohon> I think this might be the role ETSI NFV to review use cases
14:22:52 <mkoderer> vks: this would meen we do the review in the ML
14:23:14 <aveiga> vks: I think we need to send the whole list of them to the operators, because not all of them will be interested in just the one use case
14:23:20 <matrohon> we consume them, and try to implement them in Openstack
14:23:20 <matrohon> that is the way I see it :)
14:23:28 <aveiga> mkoderer: the idea is to advertise the method, not do the actual work there
14:23:49 <vks> mkoderer, but gving too many things, response will be sparse
14:24:03 <vks> probably not enough
14:24:10 <mkoderer> matrohon: yep I would like to find the fastest way to work on the implementation :)
14:24:55 <ybabenko> mkoderer: maybe we need a telcowg agenda for the next summit
14:25:08 <ybabenko> to draw attention to our issues and make community aware about telcos :)
14:25:15 <aveiga> ybabenko: already working on it
14:25:22 <matrohon> mkoderer : of course, this is the hard part
14:25:52 <mkoderer> ybabenko: sure but the summit is some months away.. I really would like to see progess :)
14:25:55 <vks> but thats far from now
14:26:15 <matrohon> mkoderer : but I think we should translate ETSI NFV uses cases in openstack specs if needed
14:26:43 <mkoderer> matrohon: sure if there are useful
14:26:49 <matrohon> for instances, cloudon implemented a use case and didn't need any modification in openstack
14:26:51 <vks> we need to start
14:27:11 <vks> matrohon, which use case?
14:27:28 <cloudon> ETSI NFV use cases are quite vague - the thinking is picking specific concrete examples and identifying what gaps exist is much more helpful for/convincing to OpenStack devs
14:27:39 <aveiga> cloudon: +1
14:27:49 <aveiga> this is what sgordon was attempting with the use cases in the wiki
14:28:07 <matrohon> cloudon, vks : I think it was SBC
14:28:26 <vks> cloudon, can i get the details?
14:28:32 <cloudon> it was the vIMS one that didn't need changes - SBC did!
14:28:48 <mkoderer> cloudon: this one https://etherpad.openstack.org/p/telcowg-usecase-Virtual_IMS_Core?
14:28:55 <cloudon> yup
14:28:57 <mkoderer> #link https://etherpad.openstack.org/p/telcowg-usecase-Virtual_IMS_Core
14:29:28 <mkoderer> the security use case also have many aspects that are somehow solves
14:29:30 <mkoderer> #https://etherpad.openstack.org/p/telcowg-usecase-Security_Segregation
14:29:33 <mkoderer> #link https://etherpad.openstack.org/p/telcowg-usecase-Security_Segregation
14:29:37 <cloudon> that was the one we started discussing a while back, identifying a possible gap in scheduling ("group affinity"), but that falls into the category of "make it work better" rather than "doesn't work at all in OpenStack"
14:29:53 <mkoderer> but anyway it's good to collect them and give guidance how to implement them
14:30:06 <matrohon> cloudon : cool is there a gap analysis for SBC use cases?
14:30:42 <cloudon> yes - the use case mentions lots of gaps with links to bps, many of which are now done
14:30:45 <cloudon> (strugles to find link)
14:31:02 <mkoderer> ok folks, anyway the review topic isn't really solved :)
14:31:33 <mkoderer> I will upload a simple use case to the sandbox repo and send it around
14:31:41 <mkoderer> and we will see how many ppl participate
14:31:51 <cloudon> #link https://wiki.openstack.org/wiki/TelcoWorkingGroup/UseCases#Session_Border_Controller
14:31:59 <mkoderer> if it works I will create a repo or we use the cross project one
14:32:05 <cloudon> the requirements section has lots of instances of gaps
14:32:39 <mkoderer> do we have any other use realted to Use Cases?
14:33:00 <matrohon> FYI : we are currently working on IPVPN interconnection use case and trying to find consensus on API in neutron
14:33:30 <mkoderer> #info matrohon working on IPVPN interconnection use case
14:33:43 <matrohon> which is related to VPN instantiation use case
14:33:52 <vks> matrohon, single site or multisite?
14:35:15 <mkoderer> ok I think we can move on
14:35:26 <mkoderer> #topic Intel NFV CI Update
14:35:27 <matrohon> vks : both, being able to attach tenant networks to IPVPN, and being able to interconnect DC through an IPVPN
14:36:08 <mkoderer> I don't know whos added that... is someone from Intel around?
14:36:35 <adrian-hoban_> Yep
14:36:59 <adrian-hoban_> I've had a few requests for info on the status of our CI for NFV
14:37:00 <mkoderer> adrian-hoban_: ok
14:37:14 <adrian-hoban_> Thought you guys might be interested too
14:37:21 <mkoderer> adrian-hoban_: sure
14:37:56 <adrian-hoban_> Intel is working on 3 CI environments for NFV in development. One CI for PCIe & SR-IOV, one CI for NFV items in Nova such as NUMA, CPU pinning, I/O aware scheduling, and one CI for Neutron related items such as the new mechanism driver/agent for ovs+dpdk-netdev
14:38:15 <adrian-hoban_> The logs should be visible for the SR-IOV CI in ~2 weeks
14:38:59 <adrian-hoban_> The other two CIs will be seen as one "intel-networking-ci account on review.openstack.org
14:39:13 <adrian-hoban_> Expect to see logs by Kilo-3
14:39:22 <mkoderer> adrian-hoban_: ok nice
14:39:23 <aveiga> adrian-hoban_: are these interconnected?
14:39:35 <aveiga> i.e. can you run dpdk and sr-iov?
14:39:58 <adrian-hoban_> aveiga: Different environments to start with
14:40:02 <aveiga> ok
14:40:53 <mkoderer> adrian-hoban_: and what's the current state? your building the envs?
14:41:22 <adrian-hoban_> We'll let the dev list know as we make the logs available.
14:42:22 <mkoderer> adrian-hoban_: ok, let me know if you have any issues with the integration into the infra chain
14:42:49 <adrian-hoban_> mkoderer: Ok, will do.
14:43:07 <mkoderer> great, so let's move on
14:43:16 <mkoderer> #topic Open discussion
14:43:26 <mkoderer> do we have anything else?
14:43:48 <margaret__> I came in late - did we discuss the philadelphia meeting?
14:44:02 <mkoderer> margaret__: nope
14:44:28 <aveiga> margaret__: what did you want to discusss about it?
14:44:36 <margaret__> I haven't looked at the etherpad - but do we have an agenda or work activity?
14:44:50 <margaret__> Curious if I should show up or not based on the agenda
14:45:01 <aveiga> we don't yet, since it's a limited attendance and we were waiting to see if enough of us would be there to warrant a session
14:45:24 <aveiga> signups only opened last week
14:45:38 <mkoderer> aveiga: could you paste a link
14:45:58 <aveiga> https://etherpad.openstack.org/p/PHL-ops-meetup
14:46:02 <margaret__> how will we know if enough folks are interested in the telcowg from the sign up?
14:46:48 <aveiga> margaret__: I'm not sure which metric sgordon was planning to use
14:47:59 <margaret__> actually if you look at the meetup - there is an agenda and folks vote on topics they are interested in.
14:48:18 <margaret__> We could put a topic like - telco WG session (I think)
14:48:57 <mkoderer> it's sometimes hard to get travel approvals within telcos ;)
14:49:24 <mkoderer> #link https://etherpad.openstack.org/p/PHL-ops-meetup
14:49:30 <margaret__> oh - so you don't think we will get enough folks. just AT&T, Comcast and maybe VZ...
14:50:32 <mkoderer> that could be an issue
14:50:47 <mkoderer> we might think about a virtual meetup for the next time
14:50:52 <margaret__> ok was just curious.
14:51:24 <mkoderer> ok folks so please have a look to https://etherpad.openstack.org/p/PHL-ops-meetup and add your name :)
14:51:42 <vks> mkoderer, so its done u will be sending the mail on  use case?
14:52:23 <mkoderer> vks: which topic are you referring to?
14:52:32 <margaret__> actually someone started the telco WG and there are 4 folks who are interested...
14:52:48 <mkoderer> vks: ha? yep I will inform the least if it reaches a draft state
14:52:59 <aveiga> margaret__: yes, and sgordon also volunteered to moderate a session if you scroll down
14:53:44 <mkoderer> s/least/list/
14:54:45 <mkoderer> ok folks.. I think we are done so far
14:54:51 <vks> mkoderer, did i missed something? :) I was abut nfv use case to advertise
14:56:11 <mkoderer> vks: yep, we can try to discuss them in the ML
14:56:29 <mkoderer> vks: I will add a review and bring it to the ML
14:56:42 <mkoderer> and we can discuss there how to process
14:56:45 <vks> ok that would be grt
14:57:06 <mkoderer> vks: all right
14:57:22 <mkoderer> so thanks everybody
14:57:32 <mkoderer> #endmeeting