Skip to content

Meeting Minutes: January 18, 2019

Jump to bottom
Mukesh Hira edited this page Jan 22, 2019 · 3 revisions

Meeting Time and Location

10 am to 11:30 am PDT (GMT - 8), Barefoot Networks, 4750 Patrick Henry Dr, Santa Clara, CA 95054

Attendees

Barefoot Networks: Jeongkeun Lee, Mickey Spiegel
Cisco Systems: Ramesh Sivakolundu
Intel: Michael Orr
VMware: Mukesh Hira

VXLAN to VXLAN-GPE Conversion for INT

Since the destination port for VXLAN and VXLAN-GPE is different, converting the encapsulation of a flow from VXLAN to VXLAN-GPE for INT purposes will potentially change the path of traffic in an ECMP environment. Moreover, enabling this in version 1.1 adds complexity in a mixed environment with v1.0, as v1.0 devices will not recognize the new bit proposed, and need to be statically configured. Given these open questions, this PR was deferred to v2.0, at least to eliminate the complexity with respect to a mixed environment.

More encapsulation types for INT

There is a need for INT to support more encapsulation formats. Following encapsulation types that have been proposed in the past would be targeted for v2.0

INT in GRE for IPv4

For IPv4 environments, use of DSCP bits and/or IPv4 options doesn't work in many environments. Support for GRE encapsulation is a good alternative, but again with the caveat that using GRE encapsulation only for INT may change the path for the flow compared to the native traffic without GRE encapsulation.

INT for IPv6 traffic

With the restriction that IPv6 extension header can only be added by the traffic source, IPv6-in-IPv6 was discussed as the only compliant option for using IPv6 extension headers for INT.

YANG model

The current version of the YANG model was an excellent start by Mickey Spiegel, but it needs to be refined to precisely define format and units for different metadata. A 90-minute working group meeting wouldn't suffice to do a deep dive and discuss necessary refinements to the YANG model. A longer dedicated meeting 9 am to 1 pm was proposed for this purpose. We are targeting having this meeting on Thursday February 21. Ramesh and Mickey have volunteered to go through the YANG model in detail and come up with a list of items that need to be addressed.

Next Steps

  • Next meeting would be held on Thursday February 7, 2019 at the same time 10 to 11:30 am Pacific Time.
  • Tentatively a longer meeting on Thursday February 21, 2019 to discuss details of Yang model
Clone this wiki locally