Clone wiki

meetings / 131122_webex

Minutes Webex 22 November 2013, 6TiSCH WG

Note: timestamps in PST.

Taking notes (using Etherpad)

  1. Xavi Vilajosana
  2. Pascal Thubert
  3. Thomas Watteyne

Present (alphabetically)

  1. Thomas Watteyne
  2. Pascal Thubert
  3. Diego Dujovne
  4. Geraldine Texier
  5. Giuseppe Piro
  6. Guillaume Gaillard
  7. Herman Storey
  8. Kuor Hsin Chang
  9. Maria Rita Palattella
  10. Nicola Accettura
  11. Pouria Zand
  12. Qin Wang
  13. Xavi Vilajosana
  14. Rouhollah Nabati

Recording

Slides

Agenda

  • Administrivia [3min]
    • Approval agenda
    • Approval minutes
  • On-The-Fly Scheduling [40min]
    • update OTF draft and next steps
    • Slot Allocation Principles
  • Mapping 6top Commands to IEEE802.15.4e TSCH [10min]
  • AOB [2min]

Minutes

  • [08.05] Meeting starts
    • Meeting starts.
    • Goal: Discuss OTF and at the end Discuss mapping of 6top commands to 15.4e TSCH.
  • [08.07] Administrivia
    • Approval agenda

      No issues raised. Agenda approved.

    • Approval minutes

      No issues raised. Minutes approved.

    • Administrative Updates:
  • On-The-Fly Scheduling
    • [08.13] Slot Allocation Principles [Pascal]
      • Considering the allocator point of view, it can be centralized, fully distributed, etc.
      • Recovering discussion before, allocation of blocks of slots:
        • two-step process:
          • delegation of group of cells by a centralized resource:
            • grouped in "chunks" of slots
            • various sizes
          • Redistribution by parents nodes
            • parent nodes allocate individual slots from its chunk
      • Possibility to have chunks which contain cells present in other chunks (i.e. shared chunks)
      • Parent when needs more bandwidth, would allocate a full chunks.
      • [Thomas] OTF has a principle of allowing two nodes negotiate about a cell. Are you suggesting to enable negotiation of chunks between two nodes?
      • [Pascal] No, a node is allocated a chunk. It can then use the cells from that chunk.
      • [Pascal] A centralized entity cuts the cell matrix into chunks. Similar to a "malloc". Instead of negotiating individual cells, nodes negotiate for a chunk, then allocate cells from that chunk. Used to define the cells that a node owns.
      • [Thomas] Chunk is an optional step where you limit the cells you can use to allocate in a particular node.
      • [Pascal] Chunk organization can be done in such a way that two-hop neighbors can be assigned different chunks to avoid collision and hidden terminal problems.
      • [Pouria] I have interesting simulation results.
      • [Pascal] Let's do allocation by chunk instead of by cell.
      • [Thomas] Pouria, what simulation results?
      • [Pouria] Yes, hybrid topology with "router" nodes around which a star topology forms. A centralized entity assigns chunks to each router. Chunks can be allocated according to an offset and reserve the cells to that node.
      • [Thomas] Are those results ready? Could you send more information to the ML?
      • [Pascal] Discussing with the neighbors to try to understand the hidden collision domains. From a centralized manager it becomes more difficult as it is dependent on the information sent by nodes (including hidden ones).
      • [Thomas] If adopted the concept of chunks would be added to the 6top API. Qin?
      • [Qin] Yes, makes sense.
      • [Thomas] Make this optional, and have a fully distributed random allocation of soft cells to keep a first solution as simple as possible.
      • [Pascal] Yes, chunks could be part of a profile. We could have a simpler profile which functions without chunks.
      • [Thomas] In low-traffic setting, a simple slot allocation scheme will work, see Xavi's early simulation results.
      • [Thomas] OTF is an algorithm that actuates in reaction to some change and soft cells are allocated dynamically.
      • [Qin] Chunks is not part of OTF, should be in a separate proposal/draft.
      • More bit: if traffic is not deterministic, use a more bit to indicate that more packets are coming.
      • Have a specific ACK IE to indicate there is an unused slot from a given chunk which can be reused.
      • In summary, 3 proposals:
        • chunks
        • more bit
        • ACK IE
      • [Kuor Hsin] When will these ideas go into a draft?
      • [Pascal] Need to discuss further. Architecture draft could a first step.
    • [08.52] Update about OTF draft and next steps [Diego]
      • update pushed to OTF draft repo.

        Not yet published.

      • 6top allocation methods: what does OTF do?
      • update to include only the softcell and bundles.
      • Algorithm selection: defining some OTF algorithms to be used in different conditions.
      • Next steps:
        • Use the 6top allocation methods
        • Define profiles
        • Define interface for OTF with 6top
        • Define management entities
      • [Thomas] Any tentative date for a -00 draft?
      • [Diego] Possibly 2 weeks from now.
  • [09.00] Mapping 6top Commands to IEEE802.15.4e TSCH [Qin]
    • Link to the standard documents.
    • MLME of 15.4e TSCH are used to manage the MIB of TSCH
    • TSCH adds a new primitive to 15.4-2011
    • TSCH defines PIB
    • TSCH adds more primitives
    • mapping 6top command to TSCH
      • Cell Create -- MLME-SET-Link(ADD)
      • TSCH does not provide a method to read a schedule
    • [Qin] Do we allow 6top to manage 15.4 MIB?
    • [Qin] Do we enable higher layers to access 15.4 PIB directly? This would reduce API on 6top.

      Action item: Qin to continue discussion about 6top to TSCH MLME mapping on ML.

  • [09.12] AOB

    No other business raised.

  • [09.13] Meeting ends

Updated