[13:16:23] YJS joins the room [13:16:30] testing [13:17:51] martini joins the room [13:17:54] scott.mansfield joins the room [13:19:36] manuel.paul joins the room [13:22:21] Stewart Bryant joins the room [13:23:41] yasuo.kashimura joins the room [13:23:45] satoru.matsushima joins the room [13:23:46] giles_heron joins the room [13:23:58] meeting starts [13:24:04] agenda bashed [13:24:11] status (Stewart talking) [13:25:01] new RFCs - MIBs (shephered by Danny) [13:25:23] with the IESG - CEP MIB, mpls-transport [13:25:44] MS PW arch needs revised ID after chair review and comments - will be resent to WG [13:25:53] VCCV-BFD in AD follow-up [13:26:27] Kenichi joins the room [13:26:46] MS PW architecture - had AD discusses and a comment (OPS AD on MIB extensions), needed terminology clarifications [13:27:38] work in progress - congestion framework (Yaakov volunteered to drive) [13:27:48] dynamic ms pw - ongoing [13:28:04] fat PW is now a new WG draft [13:28:19] fc flow in chair review [13:28:27] ICCP is a new WG draft [13:28:42] Stewart Bryant leaves the room [13:28:46] rababy joins the room [13:28:57] Glenn Parsons joins the room [13:29:11] segemented and OAM msg map need proto writeup [13:29:27] MPLS-TP new RFC 5586 ! [13:29:58] MEAD team has been holding weekly WebEx sessions and 4-day meeting [13:30:05] 3 drafts will be presented here [13:30:09] Carlos Pignataro joins the room [13:30:49] Luca coming up to present ICC for L2VPN PE redundancy [13:31:03] joint work with lots of people [13:31:43] this is an update to the interchassis communications protocol accepted last time as wg doc [13:33:07] small details needed to be fixed to give an abstract way to reference protected entities [13:33:29] idea was a "redundant object ID" which is a 64 bit ID [13:34:06] some people pointed out that NAK covered an entire package rather than individual messages [13:34:35] so decided to separate 1 application message per TLV [13:35:14] another change was mLACP to carry administrative priority only [13:35:40] also separated TLVs for aggregators vs member ports [13:36:12] needed way to solicit mLACP data from peer - so defined 3 TLVs (sync req, reply start, reply end) [13:36:30] another detail - SELECTED state added [13:36:47] added communication of BW from IF-MIB [13:37:14] there has been a comment that the document is too long and complex [13:38:32] should it be split into a base ICCP protocol and then mLACP, PW redundancy, etc as applications [13:38:46] Do we want to split (Luca proposes NOT to, in order to save work) [13:38:54] Stewart asks - who wants to split [13:39:14] about half and half (slightly more for keeping as one doc) [13:39:19] will take to list [13:39:44] Luca continuing - PW status for static PWs [13:40:07] Stewart - put two presentations together and then will discuss [13:40:24] Luca - problem that has been around for a few years [13:40:35] Stewart Bryant joins the room [13:41:15] static provision, no LDP so no status message for error messages (may be BFD, but may not want to run) [13:41:41] if use BFD just for status (not CC) it is not efficient [13:41:52] wej joins the room [13:42:08] simple design (especially for someone who already has implemented LDP) [13:42:27] replace LDP transport layer with a data plane in the ACH [13:42:38] simulate session by a periodic message transmission [13:42:58] ack option added to quench the status transmitter [13:43:00] Stewart Bryant leaves the room [13:43:13] and bypass option for AC status (end to end) [13:43:13] Stewart Bryant joins the room [13:44:03] the format has the ACH header, no TLV so zero length and value [13:44:30] refresh timer (gives default values) and ack bit [13:45:09] how to we get the packet there - to reach the peer PE (S-PE not the end one) set TTL=1 [13:45:17] Stewart Bryant leaves the room [13:46:11] if no CW there is a problem, so must use the GAL label 13 (RFC 5586) [13:46:59] bypass option - S-PEs don't care about end-to-end AC status, so send directly to end PE without S-PE relay [13:47:16] Stewart Bryant joins the room [13:47:37] but there is a bug here - if there is another path (e.g. LDP) there will be a conflict [13:48:02] comments ? the MPLS-TP are interested [13:48:24] Yakov Rekhter - are you creating a reliable transport protocol ? [13:48:35] Luca : yes - repeat until OK [13:48:49] Yakov : what about windowing ? Luca - no [13:49:02] Yakov : tremendous value to use of flow control [13:49:21] Stewart Bryant leaves the room [13:49:37] Luca - this is no worse than RSVP-TE ! Yakov - make it better than RSVP-TE [13:50:00] Rahul : on first slide you said BFD is inefficient [13:50:00] martensson.jonas joins the room [13:50:06] Luca : if no CC [13:50:33] rdroms@jabber.org joins the room [13:50:40] will need to send BFD refresh only for status [13:50:57] Rahul : why don't we look at how to fix BFD ? [13:51:12] also could run TCP [13:51:26] worse of the three is to invent a new protocol [13:51:54] Matthew : there are a number of status messages, and BFD doesn't distinguish [13:52:03] Stewart Bryant joins the room [13:52:04] Luca, also BFD doesn't run through middle node [13:52:19] Rahul : I don't like a control plane in the ACH [13:53:09] Luca : depends on what your definition of control plane is , I don't consider this a control plane [13:53:25] Nurit : this answers a requirement of MPLS-TP [13:54:21] is NMS sufficient ? Luca - no [13:54:54] someone asks - is this applicable to PW without CW ? [13:55:03] Luca - like MPLS-TP LSPs ? maybe [13:55:40] Jia He is speaking [13:55:43] no, I meant the GAL that you suggested when no CW [13:55:49] (thanks) [13:56:28] she is actually asking if applicable for VCCV-BFD - Luca it isn't [13:56:53] Luca - could run VCCV without CW using the GAL [13:57:11] Stewart - would need a draft for this [13:57:41] Italo : MPLS-TP need message not for management system, but to let actions be taken (as in OAM msg mapping draft) [13:58:00] Luca agrees [13:58:31] someone asks : can this be originated anywhere along a MS PW ? [13:58:35] Luca - anywhere [13:59:38] how is destination indicated ? Luca - always the next hop (next peer LSP) if it is an S-PE it will process and forward [13:59:58] it will eventually reach T-PE which will take action [14:00:09] Nurit - why TTL=1 ? [14:00:38] Luca - if we send to the number of hops that would not be according to the MS-PW way of doing things [14:00:50] you don't want a storm of status messages [14:01:32] Jia coming up to discuss indication of client fault in MPLS-TP OAM [14:02:15] based on MPLS-TP requirements [14:03:25] section 2.2.10 of req draft requires client signal fail (CSF) to be propagated [14:04:38] shows scenarios if failure between nodes A and B, where MPLS-TP is between B and C, the MEP function in B sends CSF to C thru MPLS-TP network [14:04:51] similar for clearing the fault [14:05:21] this draft describes the function and operation, next need to describe packet format and action to be taken [14:05:51] Stewart - this has been moved to PWE instead of MPLS WG since IP-MPLS has existing mechanisms [14:05:57] so this is only applicable to PW context [14:06:27] Stewart - there are two documents with similar goals - can we merge ? [14:07:06] Luca - this design is an outline of how this could be done, which is similar to what has been worked on in my document [14:07:26] terminology differences, but not major problem [14:07:34] Jia - each need to read the other's draft [14:07:55] Stewart - can the two authors speak together this week [14:08:15] Dave Allan : missing vis a vis the OAM msg map draft which can relay info from ACs [14:08:56] Luca - don't compete since not mapping into that kind of mechanism [14:09:11] Luca - perhaps the WG needs to consider whether we can deprecate the BFD usage [14:10:03] albert Jiang - PW LDP graceful restart [14:10:28] RFC 3478 defines LSP LDP graceful restart [14:10:33] rdroms@jabber.org leaves the room: Replaced by new connection [14:11:22] needed for T-LDP since PW is an important application [14:11:57] different from GR for LSPs since there is no IP so can't do IP routing if LSP breaks [14:12:34] there are several use cases - SS, MS, VPLS, and H-VPLS [14:12:58] SS-PW - the mechanism is as in 3478 [14:13:29] need to decouple from LDP GR for MPLS as can have L2TPv3 or "dry" PWs [14:13:46] don't need new TLV, can reuse FT [14:14:09] upon failure tear down PW, or block at T-PE [14:15:19] for MS-PW perform GR separately for each segment without communication [14:16:12] inform T-PE to withdraw PW, if S-PE control plane fails then need End-to-end OAM [14:17:04] for MS-PW with mixed dynamic and static provisioning - can use as discussed in earlier presentation [14:17:21] for VPLS - each individual PW in mesh is handled as before [14:18:01] need review by WG [14:18:05] Stewart Bryant leaves the room [14:18:11] comments ? [14:18:41] sammy boutrus : how does this work with PW redundancy ? [14:18:48] do you switch over to BU PW ? [14:19:10] Albert : haven't considered this situation. [14:19:34] Stewart Bryant joins the room [14:19:54] Rahul : basic question (I haven't read the draft) why are 3478 procedures not sufficient ? [14:20:27] same procedures, but when failure need to use OAM mechanisms to inform T-PE to pull down PW [14:20:31] giles_heron leaves the room [14:20:48] Rahul : the whole idea of graceful restart is the labels should remain so service remains [14:21:09] your statement is the opposite ! What problem are you trying to sove ? [14:21:45] Albert - trying to control retention of PW labels [14:22:22] Stewart : please take offline [14:22:27] martensson.jonas leaves the room: I'm happy Miranda IM user. Get it at http://miranda-im.org/. [14:22:45] Luca : not needed, we can do without this mechanism, and OAM breaks the model [14:23:52] Ben - representing same draft that discussed in MPLS session [14:24:13] Ralph Droms joins the room [14:24:55] we are addressing T-PEs without direct IP connectivity (still have IP addresses) [14:25:31] there are security problems when T-PE are on customer premises [14:25:48] most of the draft discusses SLA monitoring [14:26:33] reqs : No IP path between T-PE and S-PE (MS-PW) or T-PE and T-PE (SS-PW) [14:26:55] 2) static provision of LSPs and PWs [14:27:18] 3) static OAM provisioning (presently use LDP) [14:28:04] also need CCs (subsecond detection - don't need 50ms), diagnostic tests (loopbacks, traceroute, ping) [14:28:16] throughput test would be nice [14:28:43] performance monitoring - bulk of draft is about this (packet drop, delay) [14:30:01] someone asked about static provisioning [14:30:18] Ben - talking about a mix of static and dynamic (e.g. core dynamic) [14:30:54] Luca : need to hear from security people what there problems are [14:31:30] is the problem the addressing ? the forwarding ? [14:31:49] Ben - if we have control packets coming in then need to ensure that no interaction with core control protocols\ [14:32:03] so can't hack the box by the protocol [14:33:31] Ben - the problem here is the impact of a hack on a the major infrastructure network (not on L3VPNs) [14:34:14] Fred: presenting changes in P2MP PWs [14:35:08] changes from 00 based on comments - editorial and some technical changes (based on comments from Luca, ...) [14:35:21] going through changes one by one [14:35:45] new terminology - P2MP PS tunnel [14:36:05] introduced P2MP bidirectional PW [14:37:17] discussing contruction of SS-PW - set up ACs at root and leaves [14:37:36] David Sinicrope joins the room [14:37:56] can add algorithm to use one LSP for several P2MP PWs [14:38:52] another change - when failure detcted at ingress PE do an OAM message mapping [14:39:12] OAM report over AC at the egress PE [14:39:33] (ingress PE needs to know about the AC leaf) [14:40:25] explaining protection schemes - egress PE protection vs ingress PE protection [14:41:53] the multiple underlaying layers can be of different technologies [14:44:23] Luca : I went through all the "trivial" edits, but not major things [14:44:53] for example - the root does not need to take action when leaf is attached - so why do we need the message on this ? [14:45:05] Fred: this is a main difference between our drafts [14:45:18] Luca - is this a req ? [14:45:31] Is this for NMS or does the root really need ? [14:45:53] Stewart - this is a question to the list [14:46:09] Yakov : please spell out what the root does when it gets the notification [14:47:49] Yakov R : in req doc OK to say need egress and ingress PE protection, but here you give a solution [14:50:03] someone is saying that the bidirectional case is more important than the unidirectional one and Fred is saying that the new draft is not limited to unidirectional [14:51:03] presently trying to merge jounay-niger draft with martini-p2mp draft [14:51:20] should be ready by Japan [14:51:52] Luca - we want to work fast and have ready before the next IETF [14:53:36] Rahul - Would recommend using LDP P2MP procedures from MPLS WG. [14:53:48] Rahul presenting P2MP encap draft [14:54:22] this is joint work with a few people [14:55:01] motivation - encap and forwarding procedures, with procedures independent of control plane [14:55:57] (static provisioning is very much in fashion nowadays) [14:56:59] VPMS defined in L2VPN, this draft discusses encaps and semantics [14:57:46] need to carry P2MP PWs in P2MP PSN tunnels - this is the building block that ensures that don't need to overburden the backbone with the leaf info [14:58:00] egress PEs must not request PHP [14:58:25] use upstream assigned labels as in 5331, and must be the bottom label [14:58:59] encap types in 4446 must be used [14:59:28] L2 MTU needs to be the same on all ACs [14:59:55] so the P2MP label is sufficient - don't need another label [15:00:44] Luca : comment - for 1-1 mode can use implicit nul, don't need upstream assigned label [15:01:33] Rahul : won't work since the egress PE doesn't get an MPLS packet [15:01:53] Luca - the PE does know since this tree is only used for this [15:02:46] Lucy : (I didn't understand the question) [15:03:06] Rahul apparently heard and is answering that the draft gives the encaps and procedures [15:03:27] Italo : didn't understand the previous conversation about how many labels are needed for 1-1 case [15:03:45] Rahul - one label and CW [15:04:00] I am up so not jabbering [15:04:31] http://tools.ietf.org/html/draft-stein-pwe3-ethpwcong-00 [15:04:36] YJS speaking now [15:05:32] Draft focuses on congestion in the attachment circuit [15:06:13] Control word is needed, and the draft defines DEI flag (drop eligibility) [15:06:24] AdrianFarrel joins the room [15:06:33] Defines the forward and backward indicators [15:07:53] Open Issues: what happens when there are not packets going in the correct direction: the mechanism in the draft was not acceptable by everyone. [15:08:39] Asking the room to determine which option should be defined. [15:09:47] Why is this important? 1) DEI is good for PSN congestion, 2) Intermediate nodes may set ECN bit [15:10:25] Need to propagate back to the source so we can slow down [15:12:28] Request to accept this draft as working draft [15:13:01] draft needed to handle Attachment circuit congestions for Etheren PW [15:13:16] And also needed to build upon [15:13:42] David Black: Grab a two bit field, might as well have enough headroom [15:14:30] Luca: How will this be transmitted? [15:15:21] One the open issue slide, this is about how to send the "fake" packet. A packet has been be "made up". There are options, like using the ACH [15:15:39] ACH cleaner than faking a packet [15:17:15] Instead of using two fields FECN & BECN, use one field. [15:17:56] There is a discussion about keeping the fields separate, because if you need 2 forward and 1 backward, all three bits will be needed. [15:18:15] The poll of the group indicated support to move this to working group draft [15:18:20] Stewart Bryant leaves the room [15:18:21] Meeting Closed [15:18:23] satoru.matsushima leaves the room [15:18:24] rababy leaves the room: Computer went to sleep [15:18:40] yasuo.kashimura leaves the room [15:18:47] AdrianFarrel leaves the room [15:18:57] scott.mansfield leaves the room [15:19:03] Kenichi leaves the room [15:19:04] YJS leaves the room [15:19:17] David Sinicrope leaves the room [15:24:36] Carlos Pignataro leaves the room [15:35:17] Ralph Droms leaves the room: Replaced by new connection [15:38:12] Kenichi joins the room [15:39:21] Kenichi leaves the room [15:40:32] Glenn Parsons leaves the room [15:40:56] Stewart Bryant joins the room [15:42:27] Stewart Bryant leaves the room [15:58:29] martini leaves the room [15:59:42] manuel.paul leaves the room [16:18:27] lllmartini joins the room [18:10:05] lllmartini leaves the room [19:15:14] lllmartini joins the room [20:12:40] lllmartini leaves the room [22:25:25] wej leaves the room [22:45:07] wej joins the room