PRELIMINARY MINUTES ********************************************************************** PWE3 - Tuesday July 27, 2010 09:00-11:30 - Brussels Room ********************************************************************** Chairs: Matthew Bocci and Andy Malis Secretary: David Sinicrope X = No presentation submitted - removed from agenda 15 min - Agenda bash, WG Agenda and Status - Andy Malis and Matthew Bocci Matthew presented: (See slides) - most presentations were submitted - those marked removed from agenda. - On the draft-ietf-pwe3-oam-msg-map - only new issues and not reopen of old issues. - draft-ietf-pwe3-iccp - need a status update - draft-ietf-pwe3-dynamic-ms-pw - editors need feedback on adding RD to MS-PW NLRI - draft-ieft-pwe3-congestion-frwk - David Black and Yaakov Stein have some issues to resolve 5 min - Encapsulation Methods for Transport of Fibre Channel frames Over MPLS Networks - Linda Dunbar http://tools.ietf.org/html/draft-ietf-pwe3-fc-encap-11.txt David Black presented - see slides - expect one more minor rev of the draft then ready to send to WG LC - two registry points for this draft, need to check during WG LC 5 min - MPLS and Ethernet OAM Interworking - Nabil Bitar http://tools.ietf.org/html/draft-ietf-pwe3-mpls-eth-oam-iwk-03.txt Nabil presented - see slides - draft ready for WG last call Yuji - in Anaheim requested to add a clear relationship between the AC and PW? Nabil - native attach circuit and PW, its is a peer relationship. Yuji - no client server relationship? Nabil - no Yuji should be clear that this draft only deals with peer relationship and that client and server relationship is out of scope Nabil - please suggest text Yuji - the reference to Y.1731 should be 2008 not 2006 Italo - why is the single loop mode is not considered in this draft? Nabil - because then the OAM is carried as user plane traffic transparent over the PW, so it is supported. This is not same as client/server relationship which is different YJS - need to beef up security section 5 min - Mandatory Features of Virtual Circuit Connectivity Verification Implementations - Nick DelRegno Note: Will also cover the associated SP survey on VCCV implementation. http://tools.ietf.org/html/draft-delregno-pwe3-vccv-mandatory-features-01.txt Andy presented Nick's slides for him: - Note the VCCV implementation survey to be conducted by Nick and Tom Nadeau. - the thread on the list has only 4 participants. Chairs encourage folks to comment on the list. 10 min - Large Flow Classification in Flow Aware Transport over PSN - Lucy Yong http://tools.ietf.org/html/draft-yong-pwe3-lfc-fat-pw-00.txt Lucy presented: Stewart: I'm concerned about the way things are encoded in the MPLS label. Need to have the discussion with the MPLS WG to ensure they are happy with the encoding technology being used Lucy: current flow label is using MPLS encoding but semantics are different from a normal LSP label. Stewart: discussion must be had with MPLS WG, they can change the semantics not PWE3. They own the definiton of MPLS. Lucy: Perhaps the encoding can change to other alternatives if needed. Wim: classification for large and small flows, but how do you avoid reordering when you move from large to small or vice versa Lucy: there are a couple of methods to handle the reordering. You ensure that large flow stays large flow and small stays small Wim: please clarify in the draft 10 min - Encapsulation Methods for Transport of Packets over an MPLS PSN - efficient for IP/MPLS - Sriganesh Kini http://tools.ietf.org/html/draft-kini-pwe3-pkt-encap-efficient-ip-mpls-00.txt Sri presented: (see slides) - discussion following Stewart's presentation 10 min - Packet Pseudowire Encapsulation over an MPLS PSN - Stewart Bryant http://tools.ietf.org/html/draft-bryant-pwe3-packet-pw-04.txt Stewart presented: - request the draft to be a WG draft Discussion on Packet PW: Rahul: asked at last IETF refers to the NLTS, what value does this bring over and above NLTS? Stewart: NLTS doesn't address how you put this together as a system. Need a drop in solution for the network. NLTS needs a mapping to make the data link work. NLTS uses the PID label approach for anything other than IP and MPLS, conclusion on hardware survey is that this is difficult for commodity hardware. The Eth PW can be handled by most commodity hardware and is supported by deployed equipment. There are basically no protocol changes, and would be document as a BCP rather than protocol changes. Rahul: Can you document this in the draft. Stewart: This has been done Rahul: will reread and comment on the list. Andy: please look at two drafts think about efficiency vs. complexity Stewart: Need to look at the definition of efficency Italo: I prefer the simplicity versus the complexity Sri: Not that much complexity, only an IP GRE header which adds only a bit of overhead. Not only just saves bits on the wire but also avoids fragmentation. Curtis: Stewart's draft doesn't require any protocol changes so no problem. Woudl be good to have the efficiency of draft kini if chips supported it and if not fell back to draft bryant Wim: much discussion of complexity vs. efficiency, and chose efficiency. The fragmentation may not be a Yaakov: would like to see a simulation Sri: have done a simulation with real traces and there is savings ??? : As a chip vendor I don't see the 14 byte issue. Andy: Should draft-bryant be a WG draft? Question will be sent to the list. 5 min - MS-PW in PON network - Hongyu Li http://tools.ietf.org/html/draft-li-pwe3-ms-pw-pon-00.txt Hongyu Li presented: Stewart: Is PWE3 the right place for this? Is there an ITU-T liaison asking PWE3 aobut this. Matthew: No liaison from ITU-T that we have seen Stewart: GPON is not an IETF specified PSN, so we wouldn't run PWE3 over GPON as it would be outside the current architecture and outside charter of the WG. There is a mechanism for doing this that we discussed when discussing dry Martini. You would use PWE3 over MPLS over GPON/Ethernet. This would be within scope of the WG. HongYu: Could do the stack as suggested. Dave Allan: OMCI is stunt doubling for the DCN, two ways to set up labels. A bit of concern. OMCI should be to define payload and manage tcont and gemport mappings and TC bits onto a GPON infrastructure. Fred: This should be addressed in the BBF vs PWE3 as it addresses broadband architecture. Not a problem getting PWE3 labels, issue is number of targetted LDP sessions. This isn't really a problem in our experience. Monique Morrow: Liasion was sent on May 5 2010 to PWE3 and MPLS working groups. Liaison 171 for information Matthew: Not in charter of working group. Not clear what are you asking the WG to do. The broadband architecture is normally handled in the BBF and other forums. 5 min - MS-PW MIB Solution - Kiran Koushik Draft is TBD. Kiran presented. - No draft posted, but if no major objections a draft will be created and posted prior to next IETF. - What are objects that the WG feel should be added? Andy: Draft submission site is already back open. Get original draft submitted as soon as possible so folks can comment Sasha: Suggest using PW indicies as the indicies for the MIB. PW indicies are used for other things that are not applicable to MS PWs. Need to resolve conflicts that the indicies are used for. Hard to comment further with no draft. Kiran: Will use two separate indicies. Should not be any conflict. Will put in draft. Sasha: Can't say any more because draft not available. After brief look at 5061 there are multiple tables where the PW index can be used. Matthew: The item is on the agenda with no draft because Chairs want people to look at the MIB issues and problem and encourage contribution to this work. 10 min - LDP Extensions for Leaf-initiated Point-to-Multipoint Pseudowire - Fred Jounay http://tools.ietf.org/html/draft-jounay-pwe3-leaf-initiated-p2mp-pw-02.txt Fred presented: Luca Martini: How do you set up LSP underneath Fred: per dynamic LSP draft. If LDP advertise element mapping Luca: so you say the LSP exists already and you choose it? Fred: following dynamic MS PW pt to pt, if LDP based LSP and no existing LSP then take LSP to loopback address Luca: extremely inefficient because no multicast replication. Fred: Yes no multicast replication Luca: than this doesn't do multicast at all Fred: right, optimizing content is the next step. More and more star topologies so optimization may not be that important. Luca:distribution in network should be multicast based or this won't work very well especially in a ring topology. Andy: need to cut discussion, please take to the list. 10 min - LDP Graceful Restart for Pseudowire - Wu Bo http://tools.ietf.org/html/draft-jiang-pwe3-ldp-gr-01.txt Wu Bo presented: Authors request comments from the WG on whether this is a real problem. Luca: already implemented and deployed. Do we need a specific capability for this. It is really and LDP issue not PWE3. Don't need to specify capability. Wu: in the common case Luca: there is already a graceful restart capability. Wu: yes but PWE3 and LSP FECs are separated. May want to restart PWE3s independently from LSPs ???: how do you distingish between restarting PE and dead PE. More an issue with graceful restart. Target PE might be dead and result may be blackholing traffic. Wu: yes need to negotiate capability ???:yes but how do you differentiate Luca: need another mechanism like BFD to tell if box is dead. 10 min - LDP extensions for Explicit Pseudowire to transport LSP mapping - Mach Chen http://tools.ietf.org/html/draft-cao-pwe3-mpls-tp-pw-over-bidir-lsp-00.txt Mach presented: Authors would like feedback and comments on the draft. Luca: up to today mapping of PW to PSN tunnel is an internal matter. Do you want to tie the PSN tunnel to the PW? THere are different ways to do this and has been considered in the past. What application requires this? Mach: MPLS-TP networks Luca: but those networks are statically configured Mach: yes but this is a burden Luca: so then how is the static config tied to the PW Matthew: can use LDP in MPLS-TP. Haven't really worked on tying PWs to LSPs. For FEC 128 there doesn't seem to be any relevance to MPLS-TP Nabil: should probably address active-active case in the MPLS-TP control plane framework. In active active mode can't do tie break on the Node ID, may need to use bandwidth and other factors as well Italo: discussed a lot about PW role in MPLS-TP and binding to LSP. Need to address this. Matthew: Need to move on. Comments should be posted to the list. 5 min - Multiprotocol Label Switching Transport Profile (MPLS-TP) MIB-based Management Overview - Scott Mansfield http://tools.ietf.org/html/draft-farrel-mpls-tp-mib-management-overview-00.txt Scott presented: This draft does not define MIBs, defines the gaps to determine what must be put into MIBs and extensions Nurit Sprecher: Should this be limited to MPLS TP or applicable to MPLS Scott: is applicable to MPLS Kam Lam: MIB list is control plane related, but control plane in MPLS-TP is optional so needs to be clarified. Scott: It will be noted in the draft 10 min - Pseudowire freeze mechanism - Lizhong Jin http://tools.ietf.org/html/draft-jin-pwe3-pw-freeze-00.txt Lizhong presented: Jeff Tantsura: is there a live customer that requested this function? Lizhong: yes in particular for mobile backhaul Jeff: they prefer to blackhole their traffic because there is noone to receive Marc Lassere: If there is a control plane there are already ways to do this. Even if there is no control plane there are ways to do this without such an outtage. Luca: the outtage listed is 49 days... IP is much more resilient than this, so usually the case is the transport path fails and the control plane recovers. This results in stale state. Lizhong: don't have stale state Luca: while you are in this state the AC goes up and down and remote AC has no notification that there is failure in the network causing a larger outage. Matthew: please take more comment to the list 10 min - LDP Extensions for Proactive OAM Configuration of Dynamic MPLS-TP PW - Fei Zhang http://tools.ietf.org/html/draft-zhang-mpls-tp-pw-oam-config-02 Fei presented: Andy: take comments to the list 5 min - Pseudowire Stitching Procedures in MPLS-TP Networks - Sami Boutros http://tools.ietf.org/html/draft-boutros-pwe3-ms-pw-tp-00.txt Sami presented: Andy: questions and comments should be taken to the list 20 min - Open mike session on PWE3 control word and GAL in MPLS-TP The slides are included in the Chairs opening slide deck. In the 6 remaining minutes Andy presented a couple of slides to start the discussion that will be continued on the list regarding CW and GAL on PWs. Luca: On Preferred CC-type slide - More modes than Ethernet don't have control word. The current restriction requiring CW on MPLS-TP has some issues. If SP wants to deploy going from MPLS to MPLS-TP need a special function to add control word between the two. Why do we have this requirement and can we remove it? We already have a WG document for PW status over MPLS incase no control plane, why can't we do for MPLS-TP as well? It would be nice to have same label stack for MPLS, MPLS-TP and PW. If the GAL is put in, it could be normally put in as BoS with or without CW and would be consistent. Today is more complex. If we use GAL we could consolidate to one type. Where did the text for 5586 come on mandatory CW from and can we remove it? Sasha: This may be the wrong WG to ask that question. The RFC was generated by the MPLS WG. Luca: Yes but the PW is done here George: MPLS WG Chairs lead MPLS-TP, but it is joint, so to make such a change both WGs need to be involved. It can start here. Matthew: fate sharing issue Luca: fate sharing issue doesn't exist in MPLS-TP. Can we come up with a short document that says we can use the GAL in MPLS-TP. Matthew: The control word RFC it is preferred that PW encap should use the CW. Yaakov:5586 says the CW must be used in any PW. The easiest way to do this is to use the CW and not use TTL expiry. Luca: 99% of Ethernet implementations don't use control word Yaakov: yes we want that to be changed Luca: but then we add 32 bits for nothing Jia: maybe some implementations can make CW mandatory, but doesn't support backward compatibility. Matthew: backward compatibility need a common exception mechanism and older implementations would use different exception mechanism. Marc: at beginning we said we would have a survey to determine preferred VCCV method. Need to wait for survey to determine how we should proceed here. Luca: Do we want to update RFC5586 or write a smaller document? Matthew: as Chair simpler process to write small RFC to update RFC 5586. Andy: several places in 5586 that need change Luca: yes small document, not specifying paragraphs, but just addressing this topic Additional comments should be taken to the list. ** If time remains we will address the following presentations ** X - 5 min - Definition of Time-to-Live TLV for LSP-Ping Mechanisms - Sam Boutros http://tools.ietf.org/html/draft-boutros-mpls-lsp-ping-ttl-tlv-01 Not covered due to time. X - 5 min - LDP Typed Wildcard PW FEC Element - Sami Boutros http://tools.ietf.org/html/draft-raza-l2vpn-pw-typed-wc-fec-01 Not covered due to time. - Closing - Andy Malis and Matthew Bocci ********************************************************************** WEBEX INFORMATION FOR THE PWE3 SESSION(S) ********************************************************************** WebEx Link: https://workgreen.webex.com/workgreen/j.php?ED=135092597&UID=1130226237&RT=MiM0 Meeting Number: 963 851 042