IETF
grow
grow@jabber.ietf.org
Tuesday, July 30, 2013< ^ >
Room Configuration
Room Occupants

GMT+0
[02:36:21] Humberto Galiza leaves the room
[14:58:39] joel jaeggli joins the room
[14:59:13] Wes George joins the room
[14:59:48] joel jaeggli has set the subject to: GROW IETF 87
[14:59:59] <joel jaeggli> fyi I have to cover mboned
[15:02:25] <Wes George> ok. I'll be on/jabbering
[15:03:30] <Wes George> Camilo Cardona reviewing http://www.ietf.org/proceedings/87/slides/slides-87-grow-0.pdf
[15:03:34] <Wes George> slide 2
[15:05:33] <Wes George> WG restarts after 20 new people enter
[15:06:10] <Wes George> slide 3
[15:06:32] <Wes George> slide 4
[15:06:32] <Wes George> slide 5
[15:06:52] <Wes George> slide 6
[15:07:25] <Wes George> slide 7
[15:08:28] <Wes George> slide 8
[15:08:58] <Wes George> slide 9
[15:10:26] <Wes George> slide 10
[15:10:29] danyork joins the room
[15:11:30] <Wes George> ron bonica: slide 9 - no one is rec'ving traffic the ydidn't advertise a prefix for
[15:11:34] <Wes George> how is this undesired flow
[15:12:07] <Wes George> morrow - so you're describing a route leak?
[15:12:46] <Wes George> looks like in this picture, as 0's route got leaked by AS2 to As1
[15:13:09] <Wes George> peter - issue is one of conflicting policy - creates situations that aren't beneficial to some of the ases
[15:13:46] <Wes George> michael - orange - good observation, sounds like economic free ride in that example
[15:13:54] <Wes George> cdn using a peer than traffic from peer goes to provider
[15:14:04] <Wes George> question - do you think it can create loops, combining two examples
[15:14:08] <Wes George> interdomain loops?
[15:14:14] <Wes George> a: can't think of one
[15:14:55] <Wes George> jon mitchell - useful to document this, going to struggle with language that doesn't relate to business policies
[15:15:11] <Wes George> in this scenario, you're calling this undesired, but it may not actually be that way depending on which AS you're talking about
[15:15:34] <Wes George> ruediger volk - this is about business policies. i know i have seen situations wher eI saw things liek this happening
[15:15:57] <Wes George> figured some customer was collaborating with other asns and taking economic advantage, could happen accidentally, cna be a nasty thing, nice to have documented
[15:16:06] <Wes George> hum on document adoption
[15:16:45] <Wes George> now discussing chen steering use case http://www.ietf.org/proceedings/87/slides/slides-87-grow-1.pdf
[15:17:06] <Wes George> mach chen presenting
[15:17:12] <Wes George> slide 2
[15:18:37] <Wes George> slide 3
[15:20:04] <Wes George> slide 4
[15:21:17] <Wes George> slide 5
[15:23:04] <Wes George> slide 7
[15:24:26] <Wes George> slide 8
[15:28:12] <Wes George> peter: what's the aim wiith this draft? do you want to standarize, document?
[15:28:28] <Wes George> wes - caution on routing loops, route-reflector stretch
[15:29:08] <Wes George> a: most other solutions will require upgrading other devices. difficult for operator
[15:29:37] <Wes George> diffuclt to upgrade
[15:30:05] <Wes George> jeff haas - why not TE, add path, ORR, multiplane route selection
[15:30:15] <Wes George> what's the motification for yet another mechanism
[15:30:53] <Wes George> a: if the sp doesn't have mpls and doesn't plan to deploy
[15:31:15] <Wes George> haas: not sure that's a motivation for most backbones
[15:31:43] <Wes George> robert razuk - how do you acheive consistency with at least 2 rrs to announce the same routes at the same time to the clients?
[15:31:52] <Wes George> a: need sync between two RRs
[15:31:56] <Wes George> not currently in draft
[15:32:37] <Wes George> jon mitchell - agree with comments - maybe people were thinking about this in traditional terms (existing vendor products) rather than some sort of software computational product
[15:32:46] <Wes George> draft is interesting, but probably not someting we'd want to standardize
[15:33:11] <Wes George> rob shakir - if we're doing RR sync and getting ordered announcements, shouldn't we just call it PCE?
[15:34:49] <Wes George> wes - small network = comparitively easy to roll mpls and upgrade code. large network, maybe not, but this gets hard to do on a large network
[15:35:16] <Wes George> Robert Raszuk now presenting http://www.ietf.org/proceedings/87/slides/slides-87-grow-2.pdf
[15:35:26] <Wes George> IXP route Server stuff
[15:36:36] <Wes George> split Ixp drafts in two - one to cover what changes need to happen in BGP, in IDR, then the other is covering BCP
[15:36:44] <Wes George> slide 2
[15:37:05] <Wes George> slide 3
[15:37:36] <Wes George> slide 4
[15:37:58] <Wes George> skude 5
[15:38:02] <Wes George> slide 5 even
[15:38:51] <Wes George> slide 6
[15:39:33] <Wes George> slide 7
[15:41:04] rjs joins the room
[15:41:27] <Wes George> slide 8
[15:42:05] Benno Overeinder joins the room
[15:42:14] <Wes George> jeff haas- haven't read draft
[15:42:33] <Wes George> all things on slide are all common operational issues from mae-e/w RADB route servers years ago
[15:42:41] <Wes George> seems like you need the same answers - are we just missing the history?
[15:42:52] <Wes George> a: view was that this was somehow missing
[15:43:05] <Wes George> attempt to document the operational component
[15:43:29] <Wes George> jeff - nsfnet, isi archive - how do we give them different slices of the routing table
[15:43:39] <Wes George> used to be route server views, now vrfs, etc
[15:43:51] <Wes George> sounds like the doc could benefit from pulling in nsfnet
[15:44:03] <Wes George> layer 2 issues still hasn't been resolved
[15:44:13] <Wes George> proposal - bfd to BGP next hops?
[15:44:19] <Wes George> incremental deployment will take a while
[15:44:48] <Wes George> problem you run into is two clients that have reachability on to the route server, but clients can't talk to one another
[15:45:23] <Wes George> make bfd part of the resolvability for next-hop, but make the timer more sedate the further away
[15:45:42] <Wes George> various tricks for tracking
[15:45:49] <Wes George> most better than nothingg
[15:46:23] <Wes George> randy bush - anytime that layer N is not congruent with layer n-1, you're going to be in bug hell  - this has the routing plane not congruent with the data plane
[15:46:35] <Wes George> a) this has the smae issues as RR in as
[15:46:43] <Wes George> randy - only for the fools who put RRs out of the data plane
[15:47:01] <Wes George> peter - route servers deployed in production - draft is to cover those inconsistencies between deployment and spec
[15:47:10] <Wes George> jeff - BIRD supports route server views
[15:47:41] <Wes George> morrow - is the goal to document the operational reality of RS
[15:47:52] <Wes George> randy - we weren't saying "don't document" we were saying "don't use"
[15:48:03] <Wes George> robert - what's the alternative
[15:48:50] <Wes George> randy : NTT - all configs are generated automatically
[15:49:23] <Wes George> robert - to answer question, yes this is to document, not whether they're good or bad
[15:49:56] <Wes George> jon- this is only for IXP operator - do we have something for IXP participants?
[15:50:24] <Wes George> rob shakir - work that went into RS area , massively improved stability so that it's almost usable
[15:50:28] <Wes George> they used to break all the time
[15:50:37] <Wes George> certain networks won't use them due to other still existing issues
[15:53:30] <Wes George> Gunter van de Velde presenting http://www.ietf.org/proceedings/87/slides/slides-87-grow-3.pdf
[15:53:38] <Wes George> common bgp filtering
[15:53:58] <Wes George> slide 2
[15:55:16] <Wes George> slide 3
[15:56:04] <Wes George> slide 4
[15:56:51] <Wes George> slide 5
[15:57:52] <Wes George> slide 6
[15:58:57] <Wes George> slide 7
[15:59:00] <Wes George> slide 8
[15:59:06] <Wes George> slide 9
[15:59:50] <Wes George> randy - like much of BGP, the RPKI stuff is delicate and complex
[16:00:12] <Wes George> might be easiest to just point to the origin auth docs instead of goingg into much depth
[16:00:25] <Wes George> session ends
[16:01:07] <Wes George> wes - question about route leaks draft
[16:01:16] <Wes George> danny - update with feedback
[16:01:33] <Wes George> rob shakir- bgp error handling is being rev'd again, fairly detailed rewrite
[16:01:42] <Wes George> it has signifcantly fewer words
[16:01:45] <Wes George> now session ends
[16:01:47] Wes George leaves the room
[16:02:10] <joel jaeggli> thanks
[16:02:21] Benno Overeinder leaves the room
[16:02:27] rjs leaves the room
[16:27:06] Humberto Galiza joins the room
[16:47:36] joel jaeggli leaves the room
[17:16:20] danyork leaves the room
[18:43:16] Humberto Galiza leaves the room
[20:58:17] Humberto Galiza joins the room
[22:56:03] Humberto Galiza leaves the room
[23:19:49] Humberto Galiza joins the room
[23:29:18] Benno Overeinder joins the room
[23:29:37] Benno Overeinder leaves the room
Powered by ejabberd Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!