Minutes Webex 20 December 2012, 6TSCH group
Note: timestamps in PST.
- Alfredo Grieco, Politecnico di Bari
- Gennaro Boggia, Politecnico di Bari
- Kris Pister, UC Berkeley/Dust Networks
- Maria-Rita Palattella, U. Luxembourg
- Pascal Thubert, Cisco Systems
- Robert Assimiti, Nivis
- Thomas Watteyne, UC Berkeley/Dust Networks
- Tom Phinney
Overview of interests and resources.
- round-table, for each person:
- expertise: theory? simulation? implementation?
- implementation details: platforms? hardware? frequency bands?
- type of implementation: open source?
- application: industrial? others?
- use cases
- problem statement: what's missing?
- ideal schedule: work items and deliverables
- meet up in Orlando?
- [Robert, Kris, Thomas] (08.30) TSCH on 4g: 4g radio is not really different from any 2.4GHz IEEE802.15.4 radio: switch on/off, TX, RX, set channel. One exception is the timing. It should hence be possible to port OpenWSN to a 4g radio by just rewriting the radio drivers, and modifying the timeslot template. Robert believes a 25ms template is doable.
- [Pascal] (08.45) Let's start meeting, following Kris' agenda.
- Bio. Research Associate at the University of Luxembourg, working on IoT, MAC protocols for WSNs and embedded systems, QoS. Context: EU project. PhD in Electronic Engineering from Polytechnic of Bari.
- Interests. Mainly theory, but wants to work more on simulation/implementation. Lab has started buying some hardware; TelosB for now, looking for other platforms.
- Goals. Previous work on centralized scheduling for 15.4e TSCH together with CTTC. Main goal is to implement this scheduling algorithm (TASA), and improve on it, e.g. by combining centralized and distributed approaches. Wants to study the integration between RPL and 15.4e.
- [Gennaro, Alfredo]
- Bio. Both work at the telematics lab a Politecnico di Bari. Previous joint work on 15.4e/6LoWPAN/RPL/CoAP stack with Xavi Vilajosana, Mischa Dohler and Thomas Watteyne. Context: 3 Italian projects. Main research: adoption of TSCH, RPL, 6LoWPAN, CoAP stack.
- Interests. Set up a small testbed to run the stack in a real environment. Theoretical work on 15.4e/upper later interaction. Interest in both theory and implementation.
- Goal. Effectiveness of a centralized vs. distributed scheduling approach for 15.4e.
- Bio. PhD in France (Orange Labs), moved to Berkeley in 2008. Postdoc in Kris Pister's lab. OpenWSN project: open source, 15.4s-TSCH, 6LoWPAN, RPL, CoAP stack on variety of platforms. Joined Dust Networks 2 years ago. Still close relationship with UC Berkeley and OpenWSN project.
- Interests. Agrees with Robert that RPL over 4e can have many applications above and beyond industrial. Open-source implementation (OpenWSN). Identify range of applications: industrial, home, urban, building
- Goal. Fill the gap between 15.4e and RPL:
- define the mechanisms (not policy) to build TSCH schedule
- upper stack instruct TSCH to add/delete links
- Mechanism for RPL to tap in TSCH (Pascal: reservation mechanism within or external to RPL?)
- define the mechanisms (not policy) for RPL to access TSCH info
- ideal schedule:
- define mechanism to build schedule -> white paper?
- interoperability event
- define the mechanisms (not policy) to build TSCH schedule
- [Pascal] We should start the IETF process ASAP, rather than after interop.
- [Thomas] Agreed.
- Bio. Pioneered those technologies. Professor at UC Berkeley. Long experience with WSN technology. Early development efforts, collaboration with Dave Culler. Founded Dust Networks in 2002. Lots of work within standardization bodies to promote TSCH (WirelessHART, ISA100.11a, 15.4e)
- Interests. Hook RPL into TSCH.
- Goal. Sees the need to separate parallel tracks:
- Absolute simplest "IP over TSCH" possible to produce it rapidly and avoid debates. Interop to show really easy. Possibly IETF RFC.
- RPL over TSCH as a working group, which focuses on different ideas and optimization. Plenty of opportunity for theory and simulation.
- [Pascal] While separate tracks are possible, we should keep the track under a same management.
- [Kris] Simplest scheme as an RFC so we can run IP on top of TSCH. Put hooks which enable other stuff. This is one document.
- Bio. Director of Technology at Nivis. Company started on WSNs, moving towards smart object technologies. Nivis has had its own technology, targeted highly at scalable systems (e.g. municipal): 1000+ nodes in a network. Nivis is demand driven, customers want scalability. Applications: ARM, street lightting, tank level monitoring. About 4-5 years ago, Nivis started in the wireless industrial space. Was exposed to something similar to TSCH (TDMA with path and frequency diversity). Nivis then developed WirelessHART and ISA100.11a stacks, systems which are very deterministic, with strict latency guarantees and a centralized PCE. Having a single PCE limits the scalability and hop count. Currently rule of thumb: 100 devices per subnet. While networks can have multiple subnets, that's a limit. 2.5 years ago, Nivis tried to align this proprietary technology to standards. Looked at the different 4e MACs for the most applicable to highly scalable systems (incl. simulations). TSCH turned out to be the best at scalability. Possible modification to 4e: mixture of centralized and distributed. Chairs interop WG at IPSO, filling Kris' shoes.
- Interests and Goals. Find other implementations to interop with. Found out aspects of 4e and RPL that needed to be tuned. Interaction between TSCH and RPL, and RPL and other 15.4 MAC. E.g. how does discovery at MAC translate to discovery at routing. Concept of Information Elements (IEs) in 4e in interesting. Nivis has been relying on defining new IEs. Would like to look at 6LoWPAN ND vs. RPL discovery. Yet, all these aspects will fall into place through interop events. White papers and IETF RFCs are needed.
- [Kris] Aligned exactly with his view. All the details brought up by Robert highlight the need to have parallel tracks because the interop part is a lot of work.
- [Pascal] How many IE in 4e are visible for L3? Is this done "under the cover"?
- **[Robert](( Those IEs fully under the cover, i.e. completely transparent to the routing layer. RPL runs on top of TSCH.
- Bio. Mathematician by training. Retired from GE and Honeywell. Strong background in industrial communication and system design. Involved in major standardization bodies: ISA100.11a, WirelessHART. Final editor of draft version of 802.11. Lived for some time in Brussels. Back in Phoenix, AZ.
- Interests and Goals. System architecture for a world of objects. Stresses the importance of an open-source stack solution which can handle a wide range of application, beyond industrial. Believes that 15.4 might be a temporary PHY, that might evolve through e.g. low-power technology in the cell phones. We need to be able to accommodate other PHY layers, without changes in MAC. Stresses the important of a clean layer separation; a messy separation is one of the only things that causes problems in technology. Concurs with remarks by Kris/Robert/Thomas. Doesn't really care about implementation specifics, but has to be open-source. Highlights the need for both centralized (efficiency and latency guarantee), and distributed scheduling (RPL and distributed computation) could serve to recover from some catastrophic case. Many real world application don't need tight close loop control for wireless.
- Bio. Works at Cisco System in Nice. Strong background in standardization efforts, especially in IETF IoT related groups (ROLL, 6LoWPAN).
- Interests and Goals. Centralized and distributed computation routing in particular. Help RPL become generalized to more applications; in particular, can be applied to more than industrial. We could develop new usages for RPL, even voice over TSCH. Believes that we need to go to the IETF to leverage RFCs and byproducts to promote our work. While OpenWSN exists, we need to have at least 2 open-source implementations for interop tests. Interested in helping with the routing algorithms and steering the work within IETF. Believes from the discussions in the webEx that we have a good consensus: same idea on distributed/centralized, parallel tracks. We have enough material to do a bar BoF in mid-March, maybe even a BoF. We need to put together a good problem statement and reach area chairs who need to agree on the creation of the WG. Proposes to do some write-up very soon in January, will need input and help. Encourages all communication to go through the long list of people n the mailing list.
- [All] A poll concluded that a BoF in Orlando would be perfect.
While we're setting up a mailing list, these are the email addresses of the people involved in this discussion, at the time of the webEx:
- Kris Pister firstname.lastname@example.org,
- Robert Assimiti email@example.com,
- Alfredo Grieco firstname.lastname@example.org,
- Tom Phinney email@example.com,
- Laurent Toutain Laurent.Toutain@telecom-bretagne.eu,
- Andrzej Duda firstname.lastname@example.org,
- Gennaro Boggia email@example.com,
- Domenico Striccoli firstname.lastname@example.org,
- Xavi Vilajosana email@example.com,
- Maria-Rita Palattella firstname.lastname@example.org,
- Mischa Dohler email@example.com,
- Paul Didier firstname.lastname@example.org,
- Thomas Watteyne email@example.com