Re: [6tisch] New Version Notification for draft-karaagac-6tisch-int-00.txt

[Abdulkadir Karaağaç (UGent-imec) <[e-mail]>]

Dear Pascal, all,
Indeed, we have already implemented and validated the described mechanism in a real network. Then we thought that this work might be also interesting for the 6TiSCH community and it might be really beneficial to receive feedback from you all. Therefore, we decided to turn this work to an Internet Draft.

Thank you very much for your nice comments and very useful feedback. A new version of the draft will shortly follow which will be adapted based on your comments. 

In addition, I tried to respond your comments and questions below:

* Since you are using IEs Fig 1 could represent the insertion before the IPv6 payload.
-> I agree. We will update the figure in a way that represents the real position of the IEs. Maybe [Header+IEs+Payload]

* There could be a Fig describing the typical format of the inserted IE at the end of 2.2
-> We will try to add a descriptive figure for a typical INT IE format.

* A discussion on the chances of loss that are augmented as the frame are enlarged could be useful in the intro
-> Agreed!

* A discussion of the INT header used as a hidden channel could be useful in the security section.
-> Agreed!

* It seems that the data can only be reported on packets that are going outwards. You may want to indicate that in a classical non-storing network with the root acting as backbone router, all packets fly through the 6BBR and can be instrumented. 
-> Agreed!

* Because of the Q bit it seems that the INT IE is also used in packets going down: the source routing header may make it complicated to instrument such packets. More description on the operation and size of inserted header in downwards packets could be useful. Maybe we could make the query a different IE altogether and save the Q bit for the future. 
-> Indeed, although that part is not that mature, we also aimed a support for the downwards packets. I guess further consideration is needed in this regard.

* If the telemetry relates to a transmission by a child, then the node ID is not enough to identify the link since the child has multiple parents.
-> If the parent node is also able to add INT data (including node ID) to the same packet, then it will be possible to identify the link from these node IDs. If not, then as you said, the node id will not be sufficient for determining the links. Then, maybe the nodes can leverage on the previous or following telemetry information and try to infer the link?

* Apparently in HbH mode 1 only the first nodes may insert data if the network is deep and the room limited. 
-> It depends on the INT insertion strategy. Actually, a probabilistic approach (as described in 2.3.2) can enable middle nodes to insert data to some extend.

* In HbH mode 2, how do we balance the use of the space so that all hops have an equal chance to report their information? Is that what 2.3.1 is all about? 
-> Actually, we tried to aim a balance by using a probabilistic INT strategy in 2.3.2. 

* In section 2.3
** I believe that if nodes use different strategies some may starve. Like for the objective function, there should be a single strategy, that may have to be tuned for the shape of the network.  
-> I agree that more flexible and tunable strategies are required for larger networks. But I am not sure about the use of network-wide single strategy.

** in 2.3.1 if all nodes source traffic and not only the deepest ones, actually the nodes near the root may have more opportunities, it really depends on the shape of the network.
** in 2.3.2 we seem to assume that only nodes deep in the network source packets. In reality, say that there can be 3 values stored, then the number of chances is proportional to the number of children and grandchildren, and nodes deep in the network may have less of these.
-> If you consider the whole network and all available traffic, these comments correct. But we are trying to balance/compare the INT insertion opportunities of different nodes for a single flow.

** it could be good that a node that refrains to insert redundant information may still indicate that already sent information (sequence) is unchanged. Along that line of thought how do we handle loss? A same sequence sent several times?
-> Actually, we think that each INT entry carries telemetry information for that particular packet. If that packet is lost, so the INT entries are too. 

* Security section should indicate that the payload IE is modified at each hop and there is no guarantee that the original information is preserved.
-> Agreed!

Kind regards,
Abdulkadir

PS: It is possible to reach further details about our work in our paper named "In-Band Network Telemetry in Industrial Wireless Sensor Networks".


-----Original Message-----
From: Pascal Thubert (pthubert) <[e-mail]> 
Sent: Tuesday, January 14, 2020 14:11
To: Abdulkadir Karaağaç (UGent-imec) <[e-mail]>; [e-mail]
Subject: RE: New Version Notification for draft-karaagac-6tisch-int-00.txt

Dear authors

This draft strikes me as very mature for a 00, and a very useful piece of work as well.
It looks like you have implemented and tried it out already, correct?

A few comments and questions:

* Since you are using IEs Fig 1 could represent the insertion before the IPv6 payload.
* There could be a Fig describing the typical format of the inserted IE at the end of 2.2
* A discussion on the chances of loss that are augmented as the frame are enlarged could be useful in the intro
* A discussion of the INT header used as a hidden channel could be useful in the security section
* It seems that the data can only be reported on packets that are going outwards. You may want to indicate that in a classical non-storing network with the root acting as backbone router, all packets fly through the 6BBR and can be instrumented. 
* Because of the Q bit it seems that the INT IE is also used in packets going down: the source routing header may make it complicated to instrument such packets. More description on the operation and size of inserted header in downwards packets could be useful. Maybe we could make the query a different IE altogether and save the Q bit for the future. 
* If the telemetry relates to a transmission by a child, then the node ID is not enough to identify the link since the child has multiple parents.
* Apparently in HbH mode 1 only the first nodes may insert data if the network is deep and the room limited. 
* In HbH mode 2, how do we balance the use of the space so that all hops have an equal chance to report their information? Is that what 2.3.1 is all about? 
* In section 2.3
** I believe that if nodes use different strategies some may starve. Like for the objective function, there should be a single strategy, that may have to be tuned for the shape of the network. 
** in 2.3.1 if all nodes source traffic and not only the deepest ones, actually the nodes near the root may have more opportunities, it really depends on the shape of the network.
** in 2.3.2 we seem to assume that only nodes deep in the network source packets. In reality, say that there can be 3 values stored, then the number of chances is proportional to the number of children and grandchildren, and nodes deep in the network may have less of these.
** it could be good that a node that refrains to insert redundant information may still indicate that already sent information (sequence) is unchanged. Along that line of thought how do we handle loss? A same sequence sent several times?
* Security section should indicate that the payload IE is modified at each hop and there is no guarantee that the original information is preserved

Great work!

Pascal





> -----Original Message-----
> From: 6tisch <[e-mail]> On Behalf Of Abdulkadir 
> Karaagaç
> (UGent-imec)
> Sent: mardi 14 janvier 2020 10:43
> To: [e-mail]
> Subject: [6tisch] FW: New Version Notification for 
> draft-karaagac-6tisch-int- 00.txt
> 
> Dear all,
> We have submitted a new draft describing an efficient and timely 
> monitoring solution for 6TiSCH Networks. We believe that this 
> mechanism can become a very powerful tool for 6TiSCH Networks with the 
> contribution of the 6TiSCH community.
> 
> All kinds of feedback and comments are very welcome. Thanks a lot!
> 
> Kind regards,
> Abdulkadir
> 
> 
> -----Original Message-----
> From: [e-mail] <[e-mail]>
> Sent: Saturday, January 11, 2020 23:21
> To: Jeroen Hoebeke (UGent-imec) <[e-mail]>; Abdulkadir 
> Karaağaç (UGent-imec) <[e-mail]>
> Subject: New Version Notification for draft-karaagac-6tisch-int-00.txt
> 
> 
> A new version of I-D, draft-karaagac-6tisch-int-00.txt has been 
> successfully submitted by Abdulkadir Karaagac and posted to the IETF repository.
> 
> Name:		draft-karaagac-6tisch-int
> Revision:	00
> Title:		In-band Network Telemetry for 6TiSCH Networks
> Document date:	2020-01-11
> Group:		Individual Submission
> Pages:		14
> URL:            https://www.ietf.org/internet-drafts/draft-karaagac-6tisch-int-
> 00.txt
> Status:         https://datatracker.ietf.org/doc/draft-karaagac-6tisch-int/
> Htmlized:       https://tools.ietf.org/html/draft-karaagac-6tisch-int-00
> Htmlized:       https://datatracker.ietf.org/doc/html/draft-karaagac-6tisch-int
> 
> 
> Abstract:
>    This document describes In-band Network Telemetry for 6TiSCH
>    Networks, offering a flexible monitoring solution with minimal
>    resource consumption and communication overhead while supporting a
>    wide range of monitoring operations and strategies for dealing with
>    various network scenarios and use cases.  It enables 6TiSCH networks
>    to collect per-packet and per-hop monitoring information by
>    piggybacking telemetry information onto the data packets by
>    exploiting the remaining space in the IEEE 802.15.4e frames, thus not
>    impacting network behavior and performance.  This document also
>    discusses the data fields and associated data types for 6TiSCH INT
>    mechanism.
> 
> 
> 
> 
> Please note that it may take a couple of minutes from the time of 
> submission until the htmlized version and diff are available at tools.ietf.org.
> 
> The IETF Secretariat
> 
> _______________________________________________
> 6tisch mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/6tisch