Clone wiki

meetings / 130607_webex

Minutes Webex 07 June 2013, 6TSCH group

Note: timestamps in PDT.

Taking notes (using Etherpad)

  1. Xavi Vilajosana
  2. Thomas Watteyne

Present (alphabetically)

  1. Alfredo Grieco
  2. Elvis Vogli
  3. Pascal Thubert
  4. Pouria Zand
  5. Raghuram Sudhaakar
  6. Robert Assmiti
  7. Penny Chen
  8. Thomas Watteyne
  9. Tina Tsou
  10. Xavi Vilajosana

Recording

Slides

Agenda

  • Logo Challenge [5min]
  • Opportunistic Track reuse [15min]
  • 6LoWPAN Fragment Forwarding [10min]
  • Simplest TSCH [30min]

Minutes

  • [08.05] Meeting starts
  • Pascal reviews agenda for the call.
  • Logo Challenge [5min]
  • [08.07] Opportunistic track reuse
    • Pascal presents slides.
    • we decided to avoid tracks being merged/split. Instead, use L3 capabilities.
    • reservation is a waste of resources if they are not used
    • idea: opportunistically reuse a one-hop time slot, but we need intermediate node to distinguish between a packet being switched on track, or opportunistically routed on that cell.
    • proposal:
      • packets being switched have L2 MAC destination set to 0xffff (broadcast). This bypasses L2 MAC address check, but indicates to 6TUS that this packet has to be switched.
      • packets being routed, and using this cell opportunistically set the L2 MAC destination address to the next hop.
    • Destination MAC restored at ultimate or penultimate hop?
      • Discussion indicates ultimate hop makes the most sense.
  • PCE installing routes on top of schedule
    • allow PCE to install (RPL) routes on top of tracks.
    • Use different RPL instance IDs.
    • Ends up adding information in the nodes' routing table.

      '''TODO''': start thread on ML about allows PCE to install routes.

    • [Thomas]
      • This only applies when we want to do p2p inside the network.
      • If you don't install anything the routes will go to root
      • Without this feature, P2P works, but suffers from the "RPL stretch" whereby a P2P packet has to go "upstream" before going "downstream"
  • [08.21] Fragment Forwarding
    • Pascal presents slides illustrating the following draft:

      http://tools.ietf.org/html/draft-thubert-6lowpan-simple-fragment-recovery-07

    • This forwarding mode comes in addition to L2 switching and L3 routing.
    • Uses datagram tag as (explicit) switchable label.
    • Enable one not to reconstruct the packet at each hop, better latency and memory use
    • Problem: there is no IP header in anything but the 1st fragment.
    • 1st fragment sets up the state in each node along the way, to allow the switching to happen.
    • end-to-end bitmap to retry fragments.
    • Discussion
      • [Thomas] all for this method. But what is the price for end-to-end ACKs?
      • [Pascal] Needed when bundle breaks.
      • [Thomas] We should not modulate the number of times you retransmit a packet depending on whether it is a fragment or data
    • Proposal: change classical fragment with recoverable fragments (new 6LoWPAN dispatch patterns)
    • Discussion
      • [Thomas] where does this draft go? Do we want to address this withing 6TSCH?
      • [Pascal] No. If 6lo becomes WG, that's a good location to do this.
  • [08.32] Simplest TSCH [Xavi]
    • Intro Thomas
      • Started by e-mail from Kris: what the absolutely simplest way of using a TSCH network
      • Problem statement: if we meet in Berlin and we want to interoperate without any scheduler, what is the simplest way.
      • Input from Alfredo, Elvis, Maria Rita, Pouria, Xavi.
    • Xavi presents slides.
    • How to implement the minimum configuration/code to build a TSCH network?
    • Idea:
      • single slotframe, 101 slots, 16 channel offset
      • hard-coded schedule. All slots are shared.
      • First slot is used for ADV. Next 5 for "slotted Aloha" behavior.
      • Each slot marked as TX, RX and shared.
      • All other slots are off.
      • 3 TX attempts
    • Motes start running, receive EB to join the network.
    • EB are generated periodically. Put in the TX queue for tranmsmission. No ACK.
    • Slotted Aloha behavior comes for "free" in TSCH if the slot is marked TX, RX and shared.
    • Neighbor table maintained, with statistics. RPL can use that information.
    • Time parentL: best connected parent (based on neighbor table stats). Option: stability counter.
    • Xavi presents the format of the packets:
      • IEEE802.15.4e general format
      • IE (either header or payload)
      • EB contains the following IEs:
        • sync IE
        • Frame and Link IE
          • Xavi proposes to have all links appear in the IE.
          • Thomas Lets see whether it fits
          • Pouria Whether it all fits in a EB depends on the EB
        • timeslot template IE
        • channel hopping IE
    • Pouria presents the ACK frame format
      • sequence number repeats the DSN of the data
      • IE is used to provide time synchronization between node (ACK/NACK time correction IE)
    • Thomas how do you pick time source neighbors when you don't have any routing state yet
    • Xavi RSSI?
    • Thomas Let's take this on the ML.
    • Thomas where do DIO/DAOs go?
    • Xavi On the data cells?
    • Thomas How does TSCH know ADV needs to go on slotoffset 0 and not the others?
    • Pouria we can use the broadcast cell.
  • AOB
    • Xavi Next steps for logo?
    • Thomas this logo is pretty much done. Can you send a high resolution version?
    • Xavi Will do.
  • [09.05] Meeting ends.

Updated