Clone wiki

meetings / 140507_webex_6top

Minutes Webex 07 May, 6top call 6TiSCH WG

Recording: https://cisco.webex.com/ciscosales/lsr.php?RCID=9577dc19321f4c9d9fb303cdceeeee94

Note: timestamps in PDT.

Taking notes (using Etherpad)

  1. Xavi Vilajosana
  2. Pascal Thubert

Attendance (alphabetically)

  1. Pat Kinney
  2. Bryan Mc Laughlin
  3. Rouhollah Nabati
  4. Pascal Thubert
  5. Raghuram Sudhaakar
  6. Xavi Vilajosana
  7. Qin Wang
  8. Pouria Zand

Action items

  • QW and XV do an exercise off putting some cells in a block option header in a wrapper IE to see how they fit.
  • XV or QW to present block option and show how information is organized. Note that overhead needs to be studied.

Discussion

[8:05] Meeting starts

Question 1: overhead of the block option header.

Qin describes the overhead of the block option (see slides)

Raghu: is the payload of the IE of 64 bytes?

Qin: We use 15.4e packets where we carry the Wrapper IE, according to Pat K. the payload available in a 15.4e packet should be 80+ bytes?

Pat: correct, with security enabled

Qin: using wrapper IE + coap + block option of 64B we we use 77 bytes in which will fit in the 15.4e MAC layer packet

Xavi: Is only payload encoded with CBOR? or the whole wrapper IE including CoAP headers?

Qin: only payload.

Xavi: why not encoding all CoAP subheader with CBOR?

Raghu: to retain the same code base as for data only the payload is encoded in CBOR in CoAP draft.

Xavi: So it seems that the only alternative to the block option is to use an application layer fragmentation mechanism (ad-hoc)?

Raghu: what would be the size of a schedule? This should be the largest payload we may send in an IE. we can you is to evaluate

Xavi: that depends on the size of the slotframe and the number of active cells.

Raghu: can we use average traffic and occupancy?

Qin: YANG model depends how deep the resource in the YANG tree. Deep, data is smaller.

Xavi: So fragmentation at application level will require some sort of protocol to be able to break the payload in blocks if needed. What worries about block option is the overhead which is 6bytes over 64B (almost 10%).

Qin: Block option provides retransmissions, etc..

Raghu lean towards bklock option

Pascal: Discussion about how to compress identical cells in order to save more space. indicate repetitions or uses ranges instead of specific cells. To be thought.

Raghu: Is the schedule the largest set of information that 6top deals with?

Qin: statistical data may be bigger than the schedule data

Qin: the structure of teh MIB drives the result of a query. We can use filters to get what we want only. Or if we get things in the wrong order we could also retrsuture the MIB tree.

Pascal: conclusion for question 1 is to try with the block option and see is the overhead is really unacceptable, in which case we'll roll back and evaluate the app layer fragment alternative

Question 2: Need of URI host un IE

Xavi: we know who is the sender, so we can infer it and the parser can parse it.

Qin: parser does not need URI opition?

Xavi: correct, it can parse without URI-host option. It is useless for us since we know the source of the frame.

[8:40] Meeting ends

Updated