RNTWG Packet Marking WG Meeting

Marian Babik (CERN), Shawn Mc Kee (University of Michigan (US))

Packet marking working group meeting to review status and plans.  

Join Zoom Meeting

Meeting ID: 913 4320 9239
Passcode: 382133


   Packet Marking WG

   SciTags https://www.scitags.org/ 

Live meeting notes https://docs.google.com/document/d/1kWoc9Smi3nB6cajuomt-kYmNUgoyRE1C1_REkEdCPQo/edit?usp=sharing

    • 6:00 PM 7:00 PM
      Packet Marking Meeting: Packet Marking WG
      Conveners: Marian Babik (CERN), Shawn Mc Kee (University of Michigan (US))
      • 6:00 PM
        News 10m
        Speakers: Marian Babik (CERN), Shawn Mc Kee (University of Michigan (US))
      • 6:10 PM
        Packet Marking 30m
        • dCache Meeting Summary 5m
          Speakers: Marian Babik (CERN), Shawn Mc Kee (University of Michigan (US))

          Discussed different options for flow and packet marking implementation in dCache. Agreed that easiest would be to establish interface to flowd and let it handle both flow and packet marking. 

          dCache will need to implement a way how to detect start of the transfer as this is currently missing (end of transfers and bytes transferred is already done, src/dst ports can also be extracted). Flowd will need to implement a local interface and a communication protocol (flow marking is already done; Tristan is looking into packet marking). We have also agreed to align the developments with the WLCG monitoring task force which is also interested in this data.

          We discussed proposal for HTTP-TPC protocol extension in the current technical specification. Following up on this few minor changes were made in the tech. spec. 



        • eBPF-TC Updates 10m
          Speaker: Tristan Sullivan (University of Victoria)
        • RFC Updates 5m
          Speakers: Marian Babik (CERN), Tim Chown

          There’s three 6man WG drafts that are relevant, but in practice with not much impact on our WG.

          The Alt Marking draft is still in the RFC Editor queue.
          Described passive performance monitoring method. Proposes AltMark Option, which is encoded in the EHs (HbH and/or Destination). Provides very good discussion on the IPv6 flow label trade-offs. 
          There is also a draft on the SRv6 AltMark (https://datatracker.ietf.org/doc/draft-fz-spring-srv6-alt-mark/)

          The newly adopted 6man WG draft on processing EHs is interesting:
          It would only affect us if we used EHs and the WLCG packets had a lot of other EHs.

          Likewise there is now a 6man WG draft on Hop by Hop option processing:
          In principle this draft might mean fewer packets with HbH options get dropped.

        • XRootD Flow Label Implementation 15m
          Speaker: Andrew Hanushevsky (STANFORD LINEAR ACCELERATOR CENTER)
      • 6:50 PM
        Plans and next steps 10m
        • TBA 5m