Monday, 25 July 2011< ^ >
Room Configuration

[17:01:21] Ilya Varlashkin joins the room
[17:03:19] adrianfarrel joins the room
[17:03:33] <Ilya Varlashkin> good
[17:03:45] <adrianfarrel> good?
[17:03:58] <Ilya Varlashkin> good <whatever time of the day is there>
[17:04:12] <Ilya Varlashkin> and good that somebody from the site has joined
[17:04:33] <adrianfarrel> Ah, but I am in the LISP WG meeting :-(
[17:04:40] <Ilya Varlashkin> ouch
[17:05:36] <adrianfarrel> let me jabber someone direct?
[17:05:42] <Ilya Varlashkin> thank you
[17:05:54] <Ilya Varlashkin> audio is fine but only one-way
[17:06:59] YJS joins the room
[17:07:18] <Ilya Varlashkin> YJS, are you onsite?
[17:07:59] <YJS> yes, I am
[17:08:10] <Ilya Varlashkin> great
[17:08:12] <adrianfarrel> In Trill?
[17:08:22] <YJS> Stewart just asked the chair for someone to jabber
[17:08:27] Stewart Bryant joins the room
[17:08:41] <adrianfarrel> Ahah! Thanks Yaakov
[17:08:58] <Ilya Varlashkin> thank you
[17:09:00] <YJS> I didn't volunteer as I am multitasking this sesion
[17:09:25] <YJS> but, they are going over the document status
[17:09:57] <Ilya Varlashkin> I hear audio just fine, just for questions
[17:10:32] <YJS> OK - ask any questions here and one of us will relay to the mike
[17:12:40] mrichardson joins the room
[17:12:43] <mrichardson> hi.
[17:12:45] <mrichardson> sorry.
[17:12:52] <mrichardson> first draft: find-grained VLAN labelling.
[17:13:06] <Ilya Varlashkin> to "existing selicon" - is it hypothetically or confirmed by major vendors that only SW upgrade would do the trick?
[17:13:08] <mrichardson> (i had to do a new DHCP to get into jabber. weird...)
[17:13:14] <mrichardson> radia: Goals.
[17:14:31] <mrichardson> Basic Idea: use two VLAN tags.
[17:14:34] <mrichardson> slide: Double VLAN tag
[17:15:36] <mrichardson> slide: Notation.
[17:15:46] <Ilya Varlashkin> Q2: is daisy-stacking ok per specs?
[17:16:55] <Ilya Varlashkin> sorry, chain-stacking
[17:17:31] <mrichardson> Ilya, do you want me go to the mic with that?
[17:17:42] <Ilya Varlashkin> at the end of the presentation, if you could
[17:17:43] <mrichardson> slide: some issues with mixing ST and DT.
[17:18:10] <Ilya Varlashkin> you understand what I mean by chain-stacking, right?
[17:18:22] <Ilya Varlashkin> sorry for terse phrases
[17:18:32] <mrichardson> I don't.
[17:19:05] <Ilya Varlashkin> If DR RBridge double-tags, can another RBridge in the middle add its own, like with Q-in-Q?
[17:19:22] <mrichardson> okay.
[17:19:34] <mrichardson> I think it would be valid, but I'll ask Radia at the end.
[17:19:47] <Ilya Varlashkin> and the first Q before you joined: "existing selicon" - is it hypothetically or confirmed by major vendors that only SW upgrade would do the trick?
[17:20:14] <mrichardson> slide: illegal configuration.
[17:21:27] <mrichardson> Ilya, your RBridge in the middle, is it some kind of service provider, which is connecting two sites via ethernet, for instance?
[17:21:51] <YJS> the question of whether you can take 2 VLAN tags and create a single 24-bit identifier has been asked before (look up Provider Ethernet VLAN Cross Connect)
[17:22:02] <YJS> The IEEE didn't like it then ...
[17:22:04] <Ilya Varlashkin> one rbridge domain connects via another rbridge domain, sort of Carrier-supporting-Carrier
[17:22:07] Simon Leinen joins the room
[17:22:23] <mrichardson> Ilya, yes, I experience that situation today....
[17:22:36] <Ilya Varlashkin> so it already works?
[17:22:45] <mrichardson> not with trill!
[17:22:51] <mrichardson> slide: ST in core.
[17:23:23] <Ilya Varlashkin> ah, so the question was can one RBridge be transit for another RBridge provider
[17:23:34] <mrichardson> issue: avoiding false filtering with ST R1 in core.
[17:23:44] <mrichardson> er, slide: Unicast
[17:24:08] <mrichardson> slide: possible variant
[17:26:53] weiyinxing joins the room
[17:27:54] <Ilya Varlashkin> I like double-tag more than single but large tag
[17:28:22] <YJS> ISID is only for MAC-in-MAC
[17:30:53] koduck joins the room
[17:32:44] Stewart Bryant leaves the room
[17:35:11] <Ilya Varlashkin> good
[17:35:45] <Ilya Varlashkin> and provider also runs TRILL
[17:36:24] <Ilya Varlashkin> thanks to both Q&A
[17:37:44] <mrichardson> if provider also runs TRILL, then it will be encapsulated, I think.
[17:37:46] <Ilya Varlashkin> it was about RBridge over RBridge
[17:37:59] <Ilya Varlashkin> not RBridge over plain ethernet
[17:38:11] <mrichardson> the packet would like like <DST><SRC><TRILL><VLAN><TRILL><VLAN>
[17:38:38] <Ilya Varlashkin> ok, I'll have to ask authors offline
[17:38:54] <mrichardson> I think. sorry if I missed the Provider runs TRILL.
[17:39:30] <Ilya Varlashkin> no problem, not last day of TRILL :-)
[17:40:37] <Ilya Varlashkin> oh I don't think networks who want trill will settle for legacy MTU
[17:45:11] <mrichardson> what is the legacy MTU, 1500 or 9K?
[17:47:44] <mrichardson> Linda Dunbar. Directory Assisted RBridge Edge.
[17:49:20] <mrichardson> slide: characteristics of internet/cloud data centre.
[17:49:35] <mrichardson> slide: When TRILL is used in Data Centre
[17:50:24] <Ilya Varlashkin> legacy is 1500, while 9K is jumbo to me
[17:52:15] <mrichardson> slide: TRILL Boundary - Scenario #1
[17:54:48] <Ilya Varlashkin> did she say "1000 ToR switches is too much" for IS-IS? On modern platforms? surprises me
[17:55:09] <mrichardson> she did.
[17:55:18] <mrichardson> slide: TRILL Boundary - Scenario #2
[17:57:24] <mrichardson> slide: DR Assisted Non-RBridge for TRILL Encapsulation
[17:57:30] <Ilya Varlashkin> I'm curious how directory assisted approach compares to leaving it all in big ISIS in terms of performance
[17:58:37] <mrichardson> is that a question? (I suspect that they haven't implemented yet)
[18:00:14] <Ilya Varlashkin> at least in terms of theoretical estimates
[18:00:54] adrianfarrel leaves the room
[18:02:01] <Ilya Varlashkin> that got to be slow
[18:02:46] <Ilya Varlashkin> looks like control-plane software lookup vs. hardware/silicon lookup
[18:07:41] <Ilya Varlashkin> still puzzling - if ToR still needs to run IS-IS then we still have N (e.g. 1000) nodes in IS-IS and ToR need to cope with that
[18:08:17] <Ilya Varlashkin> I believe discovery of leave nodes (MAC) is not as expensive as SPF calculation itself
[18:12:04] adrianfarrel joins the room
[18:13:02] <mrichardson> slide: Phantom Nickname.
[18:17:40] <mrichardson> new speaker.
[18:17:47] <mrichardson> slide: Problem Statement.
[18:19:05] <mrichardson> slide: Pseudonode Nickname.
[18:20:14] <mrichardson> slide: control plan of pseudonode nickname
[18:22:19] <mrichardson> slide: Data Plane of Psueonode nickname
[18:24:55] hiroki joins the room
[18:26:03] <mrichardson> slide: TLV Extension for pseudonone nickname.
[18:26:50] <mrichardson> questions?
[18:26:58] <Ilya Varlashkin> not yet
[18:30:30] <mrichardson> slide: current header extensions, first 32 bits.
[18:32:54] YJS leaves the room
[18:35:55] <mrichardson> slide: proposed actions
[18:36:34] hiroki leaves the room
[18:37:18] weiyinxing leaves the room
[18:39:19] satoru.matsushima joins the room
[18:41:07] satoru.matsushima leaves the room
[18:42:17] adrianfarrel leaves the room
[18:44:03] <mrichardson> Mingui Zhang, slide.
[18:44:08] <mrichardson> slide 2: Draft updates.
[18:45:34] <mrichardson> slide: link aggregation.
[18:45:50] <Ilya Varlashkin> 3?
[18:45:58] <mrichardson> I think I might have missed a slide.
[18:46:03] <Ilya Varlashkin> looks like 6
[18:46:05] <mrichardson> it says "6" at the bottom.
[18:46:12] <Ilya Varlashkin> ok
[18:46:15] <mrichardson> slide: load balancing through hashing.
[18:46:40] <mrichardson> slide: frame processing.
[18:47:05] <Ilya Varlashkin> need better entropy for hashing
[18:48:18] <Ilya Varlashkin> deep inspection like in MPLS seems like The Must to me
[18:49:04] <Ilya Varlashkin> indeed
[18:50:08] <mrichardson> next....
[18:51:43] <mrichardson> slide: motivate
[18:51:51] <mrichardson> TRILL ESADI at bottom.
[18:52:01] <mrichardson> #3.
[18:55:38] <mrichardson> slide: TRILL ESADI Strcuture.
[18:55:51] <mrichardson> (sorry for being slow. my wife is asking me important question in other chat)
[18:56:05] <mrichardson> slide: ESADI DRB Election. (6)
[18:57:03] koduck leaves the room
[18:58:20] <mrichardson> slide: Summary (9)
[18:58:24] <mrichardson> questions.
[18:59:50] <mrichardson> we are done.
[18:59:52] <Ilya Varlashkin> time over?
[19:00:38] <Ilya Varlashkin> ah, cool it's made it
[19:00:44] <Ilya Varlashkin> I do have Q to it
[19:01:04] <mrichardson> oops, not done.
[19:01:22] <mrichardson> slide: FCRB - FCF over RBridge.
[19:02:05] <Ilya Varlashkin> personally I'm not involved in storage networking, but our colleagues say FCoE is already at the limit of their performance desires. Proposed FCoE over TRILL looks like will introduce more delay during encapsulation. So the Q is: why not to move TRILL encapsulation into FCoE cards on the hosts?
[19:02:53] <Ilya Varlashkin> I do appreciate need for FC…over…TRILL though
[19:03:46] <Ilya Varlashkin> can you ask it?
[19:05:10] <Ilya Varlashkin> cant see slides obviously
[19:10:01] <Ilya Varlashkin> michael thanks for serving as bridge to the WG! :-)
[19:10:07] <Ilya Varlashkin> cheers
[19:10:13] Ilya Varlashkin leaves the room
[19:10:22] Stewart Bryant joins the room
[19:38:11] mrichardson leaves the room
[19:38:14] mrichardson joins the room
[19:44:15] mrichardson leaves the room
[19:55:25] mrichardson joins the room
[20:10:08] Stewart Bryant leaves the room
[21:42:52] Simon Leinen leaves the room
[22:30:06] Stewart Bryant joins the room
[22:32:00] Stewart Bryant leaves the room
[22:41:43] Stewart Bryant joins the room
[22:45:56] Stewart Bryant leaves the room
[22:51:22] Stewart Bryant joins the room
[22:55:07] Stewart Bryant leaves the room
[22:56:20] Stewart Bryant joins the room
[23:21:10] Stewart Bryant leaves the room
[23:23:07] Stewart Bryant joins the room
[23:33:34] Stewart Bryant leaves the room
[23:41:55] mrichardson leaves the room
Powered by ejabberd Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!