-
Notifications
You must be signed in to change notification settings - Fork 46
Meeting Minutes: March 15, 2018
10 am to 11:30 am PDT (GMT - 7), Dell EMC, 5450 Great America Parkway, Santa Clara, CA 95054
Alibaba: Jianwen Pi
AT&T: Shyam Parekh
Barefoot Networks: Arkadiy Shapiro, Jeongkeun Lee, Mickey Spiegel, Milad Sharif, Roberto Mari
Cisco Systems: Mario Baldi
Dell : Anoop Ghanwani, Raja Jayakumar, Senthil Ganesan
IXIA: Chris Sommers
Marvell: Tal Mizrahi
Surfnet: Ronald Vanderpol
UFRGS: Jonatas Marques
VMware - Mukesh Hira
Following are the main open issues as we approach our target timeline of March end for releasing version 1.0 of the network telemetry specifications
Use of DSCP bits and probe marker to indicate presence of INT metadata has been documented in the specifications, all feedback from the working group has been incorporated. Adding support for use of IPv4 Header Options to indicate INT metadata is still outstanding.
Action Item: Mario/Rajesh from Cisco Systems to send out proposed edits for this over the next couple of days.
Tal’s edits to support checksum neutral updates via a checksum complement metadata were reviewed and agreed upon. Comments on the pull request have been incorporated and this is ready to merge.
Action Item: Mukesh to make some more edits to the checksum update behavior to allow for scenarios where checksum update may not be required at all (e.g. INT source and sink are also L4 tunnel endpoints and do not care about outer L4 checksum)
This will be addressed post-1.0. However, Mickey Spiegel presented a proposal to add opaque data to telemetry reports. This opaque data may be used to report INT metadata units and semantics among other potential use-cases. This made sense to the working group members. One comment was to rename this to “extension” header if we decide to use this for reporting metadata units as this would not really be considered “opaque” data. Another concern was that if we decide to standardize certain schema identifiers for this opaque/extension data, then we should agree on that before releasing this feature. If we were to release this in 1.0 without carving out space for reserved schema identifiers, implementations may start to use the entire space for proprietary extensions. One possibility would be to reserve the most significant bit of schema ID to indicate standard schema vs proprietary schema.
Action item: Mickey to follow up on above comments and concerns.
Milad Sharif from Barefoot Networks presented an overview of IPv6 Segment Routing, applicability of Programmable Dataplanes and P4. While the technology and interoperability is being standardized in IETF, the P4 applications working group may be a good forum to work on reference P4 implementations of IPv6 Segment Routing Source/Endpoints.
Milad's slides can be found here
- Next bi-weekly meeting will be on Thursday March 29, 2018, 10 am to 11:30 am. Meeting location and detailed agenda will be announced closer to the meeting date. Version 1.0 of the Network Telemetry specifications is to be released at the next meeting.
- Github/mailing list will be used for discussion over the next few days to close on above issues. A close to final draft of version 1.0 of the specifications will be released to the working group by March 23 with a last call for any editorial changes. March 15th meeting was the last call for bringing up any new issues. No new issues will be considered for version 1.0 unless deemed critical.