IETF 86 - L3VPN WG Meeting session agenda
Thursday, Afternoon Session II, 15:10-17:10 (85/120 min)
WG Status and Administrivia
Co-chairs
- One notable point is that the WG milestones have been updated.

mVPN MIB
Jeffrey
- Co-chair highlighted that since this work affects L2VPN WG, maybe present and discuss there as well. Presenter agreed. 
- Co-chair also mentioned that presenting (target Berlin) in the L2VPN WG is a prerequisite for adoption and that a joint adoption of the common piece seems the right way to proceed.

Global table multicast
Jeffrey
- Co-chair mentioned that this work does not seem to be in charter ; therefore we need to figure out which WG this work needs to reside in. 
- A request for chairs of MBONE or PIM comment. This was followed by MBONED chair highlighting the document used L3VPN protocols. PIM chair, also suggested expertise to progress document resides in L3VPN.
- Authors think the document could be Informational

VPN scaling considerations
Rob
- Agree that experience in multicast scaling is definitely something we could use help with.
- Comment (by Lucy Yong) that multiple protocols exist, do they all have scaling issues? Presenter responded that some would benefit from analysis, others agreed.   
- Concern raised (by Thomas Morin) that scope may need to be reduced to make document meaningful. Support of this idea by Jon Mitchell.
- Request for design considerations as well. 
- Suggest that any recommendation be underlined with empirical evidence and reference architecture. Operators suggested this may be problematic and they are not willing to disclose too much network information. 
- Who has read the document ? => approx. 10 hands raised

Virtual PE
Luyuan 
- Clarification requested if this (see slide) was one-to-one mapping. Presenter mentioned control plane may be centralized and distributed.
- Clarification requested on if entire data center represented as single virtual device, or each box vPE. Additional observation is that the instances should be clarified. 
- Comment (by Wes George) to indicate the need to consider interactions with the provisioning system and have a way to decide what PE to put a said endpoint on and for a feedback loop.
- Suggestion to consider various scenarios, (avoid?) feedback loops between provision system, to nodes, and ports. 
- Question on push vs. push by Lucy Yong ; observation by Rob Shakir to avoid “religious war” in the use VMs and functions. 
- Comment (by Kireeti Kompella) supporting the idea but questioning the usefulness of a document.
- Positive comments on concepts and document intention (deployment guide).
- Comment from presenter and co-author (Nabil Bitar) that work could drive protocol extensions, and solutions. Suggestion to make an Informational document and see what extension would be needed.
- Who has read the document ? => approx. 20 hands raised

Virtual subnet
Lucy
- Clarification requested regarding the use of L3VPN to interconnect subnet within data center, instead of CE-to-CE? Presenter mentioned this was a typo (missing line), confirms the applicability to intra-DC.
- Who has read the document ? => approx. 10 hands raised

NVGRE and VXLAN encapsulation for L3VPN
Lucy
- Suggestion (by Andy Malis) that function could be achieved with existing technology (MPLS over GRE). Presenter mentioned a need for L2 & L3 overlay edge internetworking.  
- Concern from commentators (including Dino Farinacci) that WG already has various existing encapsulation methods.
- Answer by Ali Sajassi that having VXLAN is motivated by having a global id.
- Comment (by Kireeti Kompella) that the encapsulation specified would by incompatible with VXLAN hardware.
- Comment (by Dhananjaya Rao) that he presented the idea of a VXLAN encapsulation for L3VPN, but based on VXLAN as it is, not modified.
- Presenter highlighted that this is a NVO3 (WG) requirement. Suggestion that NVO3 is wrong (Dino Farinacci). 

Virtual CE
Luyuan
- Clarifications requested (by Lucy Yong) on the position of the PE in the architecture, on the relation with the virtual PE architecture, and questioning the purpose of showing the document in L3VPN.
- Questioning (by Ali Sajassi) on the kind of work in L3VPN that this architecture may lead to.
- Observation by the chairs that more work is needed on the draft to let it appear how it relates to L3VPN WG.

L3VPN Data centre Interconnect
Luyuan
- Comment (by Jeff Tantsura) on a possible security issue and suggesting the use of carrier's carriers techniques. Answer by presenter: unsure if this would be the right way, and more complexity.
- Comment that this “is one way you can do it.”
- One positive comment on the draft.

[End of Session]