PWE3 - Monday, July 27, 2009 15:20-17:20 --------------------------------------------------- CO-CHAIRS: Stewart Bryant and Matthew Bocci SECRETARY: David Sinicrope 15 min - WG Status - Stewart Bryant and Matthew Bocci - Stewart opened the meeting - no changes to the agenda, accepted as posted - see Status slides RFC Status - Thanks to Danny for sheparding With the IESG - pwe3-mpls-transport-04 with Ralph waiting - pwe3 ms arc - some comments Matthew and Stewart addressed MS-PW Architecture Status - authors addresed AD comments and revised document. WG will have chance to read before progressed Work in Progress - draft-ietf-pwe3-congestion-frmwk-02 - YS agreed to drive - fc-flow and fat will both need a respin Stewart has comments. - segmented and msg map need proto statements - Stewart mpls-et-oam-iwk Tom Nadeau will take over as editor because stalled MPLS-TP - pwe3 part of MEAD team - further MPLS-TP discussion will take place Wed during MPLS WG session No questions 10 min - Inter-Chassis Communication Protocol for L2VPN PE Redundancy - Luca Martini http://tools.ietf.org/html/draft-ietf-pwe3-iccp-00 See Luca's slides Intoduce ROs and ROID Limit RG Application DAta Message to 1 Application per Message mLACP:PortConfig TLV Next - some comments that document is long and complicated will only get worse. Could split into multiple documents, e.g., into protocol and one for each application of mLACP, but not author's preference. Slightly more support in the room for keeping as one document. Will take to list. 10 min - Pseudowire Status for Static Pseudowires - Luca Martini http://tools.ietf.org/html/draft-martini-pwe3-static-pw-status-00 - Luca will present this document, Jia will present CSF and then we will have discussion of what WG wants to do. Slight differences in documents. - problem for several years, how do you do PW status when PW is static (i.e., no signaling) Static provisioned PW and Status could do with BFD, but if you don't want to use BFD, no status. Protocol Design - simple Message Transport Design Message Transport Encapsulation If there is a CW it is set to 1 and after label is good, if there is no CW, need to support, then use GAL End to End Status Message Bypass Option For AC status only, doesn't reflect PSN status Review with Matthew found bug, so it needs updating. Will address bug and keep it backward compatable. Q&A - Yaakov Rehkter - Are you recreating reliable transport again? - Luca - yes, you repeat and repeat and repeat - YR - would you consider flow control? - Luca - No - YR - you should consider flow control for control traffic like this. - Rahul - Why is BFD inefficient? - Luca - To bring up BFD just to send status occasionally is inefficient - Rahul - 3 ways to go about 1. use within BFD, 2. this proposal, 3 start running TCP in ACH - Luca to make BFD efficient would need to look at disabling the polling - Matthew could be injected by intermediate nodes via ACH - Luca this is not a full control plane over AIS or other mechanisms - Nurit - need to handle status of AC, converge this and CSF into one document - Jia - using a GAL if no CW. Is this applicable for PW in general or PWs with only a control word. - Luca could it be extended for MPLS-TP. Yes, but right now only PWs, but we should look at extending. - Matthew misunderstand question - could use use this with the GAL on other protocols? - Luca, yes you could use this with VCCV without the control word - Jia so it is useful not only for this draft but in other cases - Italo - from MPLS-TP we need to have message not for informing mgmt, but for informing about the AC and for taking action. - Albert John - Is this only for end of the PW or intermediate nodes. - Luca - at segment points in a MS PW. Destination is always TPE. SPEs will see it. - Nurit - why set TTL to 1? and not number of hops? - Luca - don't need to send directly to end, because every SPE will process on way to TPE 10 min - Indication of Client Failure in MPLS-TP - Jia He http://tools.ietf.org/html/draft-he-mpls-tp-csf-00 Jia presented slides Clarify to why this is moved hear vs. MPLS. MEAD team discussed problem. In the case of IP and MPLS we had mechanisms that addressed this, so would only need to address in PW context. No questions. - Stewart - two pieces of work for same goal, some way to merge might be best - Luca - no problem because only and outline for how to send packet and how to solve issue described in my document. No problem merging. Some terminology differences. - Jia - look at each others draft and then figure how to merge - Stewart - two authors should speak prior to end of IETF to determine merging - Dave Allan - missing something related to OAM message mapping draft which addresses passing of AC status. - Luca - that draft tells how to map, this tells how to transmit without LDP status signaling and BFD may not be best way of doing it. Should we deprecate BFD usage and just use this. Not sure yet. 10 min - LDP Graceful Restart for Pseudowire - Albert John http://tools.ietf.org/html/draft-jiang-pwe3-ldp-gr-00 Albert presented slides Questions - Sami Boutros - question how will your mechanism work with PW redundancy with GR. Would you recommend switching over to back up PW? - Albert - Don't consider that situation yet. Can take to list or offline. - Rahul - Why are the procedures of LDP GR RFC 3478, not sufficient for PW graceful restart - Albert we use exact same mechansims but use OAM mechanism for notifying the remote PEs - Rahul The procedures state that PE1 will hang onto labels in the event of PE2 failure. This proposal does the opposite. What problem are you trying to solve? control plane failure or data plane failure - Albert - all mechanisms are the same, except using the OAM mechanism -Rahul - will take offline - Luca - I don't think this is necessary. Already doing GR for PWs without this. You can't send an OAM message because the idea is NOT to take down the data plane. -Stewart - take to the list 10 min - BT Requirements for MPLS-TP features - Ben Niven-Jenkins http://tools.ietf.org/html/draft-jenkins-mpls-tp-bt-requirements-00 Ben presented - presented in MPLS this morning - these are the most pressing MPLS-TP requirements that BT has. - not looking to deploy new equipment, have MPLS already - in the draft not having an IP path doesn't mean devices won't understand IP, TPEs and SPEs won't talk IP to one another - covers SS and PW Questions - Ben Talking about provisioning, not necessaryily provisioning end-end. Some LSPs would be LDP based. Other end of MS PW might be static or dynamic depending on use case. - Luca - What specifically is the security issue? - Ben- if we have control packets coming into an SPE from a TPE we must ensure there is no way to hack the box and gain access to the control protocols of the network. - Luca - what do you do with L3VPN today? - Ben - different use case, L3VPN is not same as 21C which is used for critical infrastructure - Luca - but still customer has access - Ben - but not as critical More discussion to be done on list 10 min - Requirements for Point-to-Multipoint Pseudowire - Fred Jounay http://tools.ietf.org/html/draft-ietf-pwe3-p2mp-pw-requirements-01 10 min - Point to Multipoint Solutions Discussion - Fred Jounay & Luca Martini Fred presented slides Questions: - Luca (to WG) - went through the document and picked out some things that could be picked out without major changes. Didn't want to make changes without asking WG. Do you want root or NMS notified when leaf is attached? Root really won't do anything different when leaf attaches. Do we need it? What is the requirement? If root, what is action it will take? - no opinion from group - Yakov - If you have this requirement, then you need to spell out what root is supposed to do with notification. - Fred - need to differentiate some leafs for some protection schemes. - Luca we need to look because probably leaft that will take action, but need to spell out - Rahul - if use is resiliency then draft should spell that out. - Yakov - OK in req document to spell out that PW must be protected, this goes into how it is protected which shouldn't be in a requirments draft. - Fung Huan ??? - draft focuses on uni directional p2MP, but in some cases bidirection p2MP is more important. Any plans to work on BiDir P2MP? - Fred - I removed all bidir P2MP, it is optional - Fung Huan ??? -perhaps you should handle bidir P2MP - Luca - he wants to make it mandatory - Luca - plan is that we will work together and come to next IETF with joint draft that hopefully we can make a WG draft before the next IETF. - Rahul - one of the drafts talks about using the LDP upstream procedures, while the other draft does something different. Would recommend using LDP P2MP heirarchy procedures from MPLS WG. 10 min - Point-to-Multipoint Pseudo-Wire Encapsulation - Rahul Aggarwal http://tools.ietf.org/html/draft-raggarwa-pwe3-p2mp-pw-encaps-00 Rahul presented slides - expect this to evolve work with a bunch of people Questions - Luca - the 1:1 mode really only need to use implicit null. - Rahul - to be clear you're saying there is no label on the stack? - Luca - there will be one label on the stack because you specify implicit null. Assumption is that tree only used for PW. - Lucy - A clarification on this draft is to state how P2P PW encapsulation can apply to P2MP PW or propose a new encapsulation method for P2MP PW. - Rahul - whether you have CW or not is defined by PWE3. The second is there are upstream assigned labels and that is where this comes in. - Italo - confused by one or two labels discussion - Rahul - In the 1:1 case there would be one label. We will revise the draft and take it from there. 10 min - Ethernet PW Congestion Handling Mechanisms - Yaakov Stein http://tools.ietf.org/html/draft-stein-pwe3-ethpwcong-00 Yaakov (YJS) presented slides - Draft addresses how to solve problem of marking packets when no traffic flowing in that direction Questions - David Black - make sure that when done grab a 2 bit field rather than 2 bits to handle relation to some other work. - YJS - OK if forced. - Luca - how do you propose to transmit this? - YJS - this is for the "fake" packet - I would inject a packet if there is no packet going in that direction. I indicate injected packet by doing something illegal. - Luca - why not ACH? - YJS - OK can do that. Cleaner - Stewart - CW or Label 13... - YJS - Oh no, not label 13. :-) - Albert - you use 1 bit in the CW to notify ??? - YJS - you use 3 bits, don't need to separate AC from MPLS location, enough to say congestion, but can look at extension. - Luca - there is a case to separate two. - YJS - not sure separate managment systems, but could be. - Stewart - who thinks this is a work group item we should progress Support in the room to make it a WG draft. Will also ask on the list. 5 min - Closing - Stewart Bryant and Matthew Bocci ************************************