Hello Kris: I certainly buy your arguments, which fit the intuition of the network operation.
There are two additional angles that may influence this proposition, though, both relating to uses cases that minimal enables and that are a lot less engineered than the art TSCH: - One is a case where more than one minimal 6TiSCH network are deployed in one interference domain. The networks may not be coordinated and they may not be synchronized. Say they are dimensioned for their own traffic but not for necessarily twice that traffic. If they both operate around the default of 100 slots per slotframe, the chances of contact are very low. Yet there’s this roughly 1/1000 chance that the end up on a same channel at a shifted time, and if the slot duration is the same, the collision becomes systematic. In that case both networks may experience degradation. CCA may still help that case just like it may help coexistence with other types of networks. If we do CCA, only the network which is slightly lagging will experience bad performances, which is half of the way. It may make sense to provide recommendations for that sort of case, like, if there is a minimum of coordinated configuration available, then at least do not use slotframes of equal length and if possible make them prime to one another to avoid systematic collisions. - The other is longer sleep and higher drifts. Would it be possible that because the minimal devices experience more asynchronous / statistical traffic, there may be a wider distribution of interval between frames and thus of time drifts? If so, wouldn’t it be possible that 2 devices are desynchronized enough that one starts emitting while the other is still sensing the channel, in which case CCA would protect the first come? Also, CCA simplifies the compliance with regulation in Europe and countries that follow the same regulations. I’m hearing though that 200328 2.0 will be more open to channel hoppers. I’ll try to get the exact wording but I understand that you can be a channel hopper if you hop between 5 channels as opposed to 15 as before. So CCA may be less critical there as it was before. All in all, I do not see a strong case for CCA, but incremental benefits. I also sense some push back against it but I have less understanding of where that comes from. Is that an battery drain problem? Is that a matter of complexity? Thanks for all this, Pascal From: Kris Pister [mailto:[email protected]] Sent: vendredi 18 décembre 2015 19:36 To: Pascal Thubert (pthubert) <[email protected]> Cc: 6tisch <[email protected]>; [email protected] Subject: Re: [6tisch] #40 (minimal): Ralph's INT AREA review on minimal Pascal - CCA can be a useful feature in a TSCH network. It is certainly useful if the EU standards prevent you from selling your devices without it. It doesn't provide much benefit for the purpose that you describe below however. Most motes in a TSCH network will the synchronized to a small fraction of the guard time (+/- 1 ms default for minimal). That's just simple statistics - if six sigma of the synchronization error distribution is approach the guard time, then you're already in deep trouble with motes falling out of the network. Published academic networks routinely run with a few tens of microseconds of synchronization error. Commercial networks can be below one microsecond RMS. In general, synchronization improves with increasing network traffic, as the motes get updates more frequently. Compare these numbers to the RX-to-TX turnaround time specified in the 15.4 standard, which is 192 microseconds. If I'm supposed to start transmitting at time T0, I need to *end* my CCA RX of the channel something like 192 microseconds before T0. But everyone who might collide with my transmission is also planning to transmit at time TO. As described above, they will probably succeed with an error of a few microseconds, to possibly tens of microseconds. Result: I won't hear them. 192 microseconds is the upper bound given in the standard, and most radios are faster. But even so CCA is just not going to help you very often. It may be useful to detect WiFi or other interferers, but it won't detect other motes in your network contending for the same slot. ksjp On 12/18/2015 8:55 AM, Pascal Thubert (pthubert) wrote: Hello Xavi: Point is, we are doing statistical multiplexing in the minimal time slots. If a node starts talking, the others should yield. The draft says that CCA is optional; this is not a very strong language. Shouldn’t we encourage it a bit more loudly? Cheers, Pascal From: [email protected]<mailto:[email protected]> [mailto:[email protected]] On Behalf Of Xavier Vilajosana Sent: vendredi 18 décembre 2015 17:47 To: [email protected]<mailto:[email protected]> Cc: Pascal Thubert (pthubert) <[email protected]><mailto:[email protected]>; Tero Kivinen <[email protected]><mailto:[email protected]>; Ralph Droms (rdroms) <[email protected]><mailto:[email protected]>; Kris Pister <[email protected]><mailto:[email protected]>; 6tisch <[email protected]><mailto:[email protected]>; Simon Jonathan <[email protected]><mailto:[email protected]>; [email protected]<mailto:[email protected]> Subject: Re: [6tisch] #40 (minimal): Ralph's INT AREA review on minimal As Jonathan said in a previous thread. CCA is optional and does not bring a big advantage as nodes are synchronized and communication occurs at the same time and hence CCA does not bring a lot. Maybe to get some benefit in duty cycle regulation due to LBT and maybe to avoid some external interference. But at the end the behaviour in shared slots is slotted aloha if this CCA option is not used. regards, Xavi 2015-12-18 17:00 GMT+01:00 [email protected]<mailto:[email protected]> <[email protected]<mailto:[email protected]>>: You’re correct, they are very similar so no need to change. Pat On 18, Dec2015, at 8:57, Pascal Thubert (pthubert) <[email protected]<mailto:[email protected]>> wrote: Hello Pat: Actually, I provisionally wrote the following in last call text for the abstract: > This document describes a minimal mode of operation for a 6TiSCH > Network, to provide IPv6 connectivity over a Non-Broadcast > Multi-Access (NBMA) mesh that is formed of IEEE 802.15.4 Time slotted Channel > Hopping (TSCH) links. > This minimal mode uses a collection of protocols including the 6LoWPAN > framework and RPL to enable shared access operations over a static > TSCH schedule. Note that Ralph had an issue with the beginning of the sentence which now mentions a “collection of protocols”. Otherwise, the last call text seems to be very similar to your proposal. Would you suggest to change the above abstract or is it OK as is? Pascal From: [email protected]<mailto:[email protected]> [mailto:[email protected]] Sent: vendredi 18 décembre 2015 15:41 To: Pascal Thubert (pthubert) <[email protected]<mailto:[email protected]>> Cc: Simon Jonathan <[email protected]<mailto:[email protected]>>; Qin Wang <[email protected]<mailto:[email protected]>>; Xavier Vilajosana <[email protected]<mailto:[email protected]>>; Tero Kivinen <[email protected]<mailto:[email protected]>>; Ralph Droms (rdroms) <[email protected]<mailto:[email protected]>>; Kris Pister <[email protected]<mailto:[email protected]>>; [email protected]<mailto:[email protected]>; 6tisch <[email protected]<mailto:[email protected]>> Subject: Re: [6tisch] #40 (minimal): Ralph's INT AREA review on minimal We could change the sentence to : This minimal mode leverages 6LoWPAN and RPL to enable communication links over a static TSCH schedule via shared time-slots. Pat On 18, Dec2015, at 8:25, Pascal Thubert (pthubert) <[email protected]<mailto:[email protected]>> wrote: We need to reach consensus on this; With minimal, we are using a time slotted medium with shared access. And we want to do cca to avoid collisions, don’t we? If that’s so, then even if the MAC has that optional, minimal needs it. Should we call this TDMA? For some people (including Wikipedia and yours truly), TDMA is about exclusive access to time slots. Which will be the case of the slots that are assigned by the 6P protocol between parent and child, but is not the case of the minimal draft. I agree that because we do cca, we are not aloha stricto sensu either. What we are doing extends slotted-aloha to make it “polite”. In a way that makes minimal compliant with ETSI, since politeness is what the regulation is all about. Will we agree if we replace “slotted-aloha” by “polite slotted-aloha” or “a polite form of slotted-aloha”? Cheers, Pascal From: Jonathan Simon [mailto:[email protected]] Sent: mardi 15 décembre 2015 01:28 To: [email protected]<mailto:[email protected]> Cc: Qin Wang <[email protected]<mailto:[email protected]>>; Pascal Thubert (pthubert) <[email protected]<mailto:[email protected]>>; Xavier Vilajosana <[email protected]<mailto:[email protected]>>; Tero Kivinen <[email protected]<mailto:[email protected]>>; Ralph Droms (rdroms) <[email protected]<mailto:[email protected]>>; Kris Pister <[email protected]<mailto:[email protected]>>; [email protected]<mailto:[email protected]>; 6tisch <[email protected]<mailto:[email protected]>> Subject: Re: [6tisch] #40 (minimal): Ralph's INT AREA review on minimal Pat - Carrier sense (via CCA) is an option in TSCH, so it can be used where appropriate (e.g. for coexistence), but isn't required in general as part of the media access scheme, again because it may not be useable in a tightly synchronized network. Jonathan On Mon, Dec 14, 2015 at 4:18 PM, [email protected]<mailto:[email protected]> <[email protected]<mailto:[email protected]>> wrote: I made an error in my earlier email, although shared slots do not require carrier sense, it is really recommended. In most 15.4 modes (but not TSCH) when a device wishes to use a shared medium, the devices use CSMA to avoid collisions. Also, devices compliant to ETSI 300-328 must use carrier sense for LBT (802.15.4’s CSMA is cited in that regulation) Pat On 12, Dec2015, at 16:29, Jonathan Simon <[email protected]<mailto:[email protected]>> wrote: Pat - unless something was changed in 802.15.4-2015, that was not how the original TSCH shared slots worked. Devices don't do carrier sense, since transmissions are synchronized and talk at the same time (within sync tolerances) - they do however back off using a similar backoff mechanism, but counted in shared slots as opposed to time, to avoid persistent collision. I think that's what "slotted Aloha" is supposed to mean here - a slotted shared medium without carrier sense. Jonathan On Sat, Dec 12, 2015 at 4:57 AM, Pat Kinney <[email protected]<mailto:[email protected]>> wrote: Hi Qin; A shared slot is open for all devices. To transmit on this timeslot a device shall sense the medium for activity, if active it shall wait for the next available time slot. Hence a shared slot is a contention access period for CSMA-CA. This isn't slotted aloha, since it senses the medium first. Pat Patrick Kinney Kinney Consulting +1.847.960.3715<tel:%2B1.847.960.3715> [email protected]<mailto:[email protected]> On Dec 11, 2015, at 5:06 PM, Qin Wang <[email protected]<mailto:[email protected]>> wrote: Hi Pat, According to my understanding, in the TSCH mode of 802.15.4, if the attribute of a slot is Shared, slotted- aloha access should be allowed in the slot. Right? Thanks Qin On Friday, December 11, 2015 2:29 PM, "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> wrote: Xavi; As I understand slotted-aloha, TSCH is really Time Division Multiple Access (TDMA), not slotted-aloha. Slotted-aloha access to the medium is used in the 802.15.4 CSMA algorithms for some modes but not TSCH. Pat On 11, Dec2015, at 11:24, Xavier Vilajosana <[email protected]<mailto:[email protected]>> wrote: Dear all, I wrapped up the proposed changes and integrated them to the version in bitbucket. https://bitbucket.org/6tisch/draft-ietf-6tisch-minimal/commits/28cb63fde078a0aec8307d416e82cdf482c0608a For simplicity, see here a summary of the changes. Abstract: [OLD] This document describes the minimal set of rules to operate an IEEE 802.15.4 Timeslotted Channel Hopping (TSCH) network. This minimal mode of operation can be used during network bootstrap, as a fall- back mode of operation when no dynamic scheduling solution is available or functioning, or during early interoperability testing and development. [NEW] This document describes a minimal mode of operation for a 6TiSCH Network, to provide IPv6 connectivity over a Non-Broadcast Multi- Access (NBMA) mesh that is formed of IEEE 802.15.4 Timeslotted Channel Hopping (TSCH) links. This minimal mode leverages 6LoWPAN and RPL to enable slotted-aloha operations over a static TSCH schedule. Introduction: [OLD] The nodes in a IEEE 802.15.4 TSCH network follow a communication schedule. The entity (centralized or decentralized) responsible for building and maintaining that schedule has precise control over the trade-off between the network's latency, bandwidth, reliability and power consumption. During early interoperability testing and development, however, simplicity is more important than efficiency. One goal of this document is to define the simplest set of rules for building a TSCH-compliant network, at the necessary price of lesser efficiency. Yet, this minimal mode of operation MAY also be used during network bootstrap before any schedule is installed into the network so nodes can self-organize and the management and configuration information be distributed. In addition, the minimal configuration MAY be used as a fall-back mode of operation, ensuring connectivity of nodes in case that dynamic scheduling mechanisms fail or are not available. The IEEE 802.15.4 specification provides a mechanism whereby the details of slotframe length, timeslot timing, and channel hopping pattern are communicated when a node time synchronizes to the network [IEEE802154]. This document describes specific settings for these parameters. [NEW] A 6TiSCH Network provides IPv6 connectivity over a Non-Broadcast Multi-Access (NBMA) mesh that is formed of IEEE 802.15.4 Timeslotted Channel Hopping (TSCH) links. The 6TiSCH [I-D.ietf-6tisch-architecture] architecture requires the use of both RPL and the 6LoWPAN adaptation layer framework ([RFC4944], [RFC6282]) as defined over IEEE 802.14.5. 6LoWPAN Neighbor Discovery [RFC6775] (ND) is also required to exchange Compression Contexts, form IPv6 addresses and register them for the purpose of Duplicate Address Detection, Address Resolution and Neighbor Unreachability detection over one TSCH link. In order to reduce the header overhead of the RPL artifacts in data packets, the Routing header [RFC6554], the RPL Option [RFC6553] and the related IP in IP encapsulation MUST be encoded as prescribed in [I-D.ietf-6lo-routing-dispatch] Nodes in a IEEE 802.15.4 TSCH network follow a communication schedule. A network using the simple mode of operation uses a static schedule. This specification defines a Minimal Configuration to build a 6TiSCH Network, using the Routing Protocol for LLNs (RPL) and a static TSCH Schedule. The 802.15.4 TSCH mode, RPL [RFC6550], and its Objective Function 0 (OF0) [RFC6552], are used unmodified, but parameters and particular operations are specified to guarantee interoperability between nodes in a 6TiSCH Network. More advanced work is expected in the future to complement the Minimal Configuration with dynamic operations that can adapt the Schedule to the needs of the traffic in run time. Section 11.2 [OLD] In addition to the Objective Function (OF), nodes in a multihop network using RPL MUST indicate the preferred mode of operation using the MOP field in DIO. Nodes not being able to operate in the specified mode of operation MUST only join as leaf nodes. RPL information and hop-by-hop extension headers MUST follow [RFC6553] and [RFC6554] specification. In the case that the packets formed at the LLN need to cross through intermediate routers, these MUST follow the IP in IP encapsulation requirement specified by the [RFC6282] and [RFC2460]. RPI and RH3 extension headers and inner IP headers MUST be compressed according to [RFC6282]. [NEW] In addition to the Objective Function (OF), nodes in a multihop network using RPL MUST indicate the preferred mode of operation using the MOP field in DIO. Nodes not being able to operate in the specified mode of operation MUST only join as leaf nodes. RPL information and hop-by-hop extension headers MUST follow [RFC6553] and [RFC6554] specification. In the case that the packets formed at the LLN need to cross through intermediate routers, these MUST follow the IP in IP encapsulation requirement specified by the [RFC6282] and [RFC2460]. RPI and RH3 extension headers and inner IP headers MUST be compressed according to [RFC6282] and [I-D.ietf-6lo-routing-dispatch]. have a nice weekend, Xavi 2015-12-11 17:49 GMT+01:00 Kris Pister <[email protected]<mailto:[email protected]>>: Ralph - to my knowledge no one has deployed the specific time-parent selection scheme described in 802.15.4-*. The basic scheme will likely work, but the devil will be in the real-world details. We've had about 8 years of successful deployments of industrial tsch mesh networks using a time-parent selection scheme similar to what is proposed in minimal. 6TiSCH present a rich design space at many levels. The goal of minimal was to do something simple, based as closely as possible on things that are known to work in deployed networks. The hope and belief is that new and better ideas will emerge, but it is certain that many of the proposed "good ideas" will fail. By defining minimal we provide a reliable interoperable platform on which papers like "Comparing time-parent selection in 15.4-* and foo" can be written. ksjp On 12/10/2015 5:43 AM, Ralph Droms (rdroms) wrote: Is there an analysis published somewhere that demonstrates how time synchronization in 802.15.4-* is inadequate for 6TiSCH? _______________________________________________ 6tisch mailing list [email protected]<mailto:[email protected]> https://www.ietf.org/mailman/listinfo/6tisch _______________________________________________ 6tisch mailing list [email protected]<mailto:[email protected]> https://www.ietf.org/mailman/listinfo/6tisch _______________________________________________ 6tisch mailing list [email protected]<mailto:[email protected]> https://www.ietf.org/mailman/listinfo/6tisch _______________________________________________ 6tisch mailing list [email protected]<mailto:[email protected]> https://www.ietf.org/mailman/listinfo/6tisch _______________________________________________ 6tisch mailing list [email protected]<mailto:[email protected]> https://www.ietf.org/mailman/listinfo/6tisch -- Jonathan Simon Linear Technology, Dust Networks product group 32990 Alvarado-Niles Road, Suite 910 Union City, CA 94587 (510) 400-2936<tel:%28510%29%20400-2936> (510) 489-3799<tel:%28510%29%20489-3799> FAX [email protected]<mailto:[email protected]> ******************LINEAR TECHNOLOGY CORPORATION CONFIDENTIAL****************** This e-mail transmission, and any documents, files or previous e-mail messages attached to it may contain confidential information that is legally privileged. If you are not the intended recipient, or person responsible for delivering it to the intended recipient, you are hereby notified that any disclosure, copying, distribution or use of any of the information contained in or attached to this transmission is STRICTLY PROHIBITED. If you have received this transmission in error, please immediately notify me by reply email or by telephone at 510-400-2936<tel:510-400-2936> and delete the original transmission and its attachments without reading or saving in any manner. Thank you. _______________________________________________ 6tisch mailing list [email protected]<mailto:[email protected]> https://www.ietf.org/mailman/listinfo/6tisch
_______________________________________________ 6tisch mailing list [email protected] https://www.ietf.org/mailman/listinfo/6tisch
