[Ndn-interest] nfd and iot?

Mahmoudi, Charif (IntlAssoc) charif.mahmoudi at nist.gov
Mon May 23 07:59:53 PDT 2016


Hi Marc,

Our main support to address that issue is a name delimiter based on naming conventions. for example. If the domain is /dom1. A composition of can be managed in an intuitive way.
/dom1/actuator1/dom1/agregator1/dom1/sensor1/dom1/sensor2
|--------1------------|---------2-------------|-------3-----------|---------4---------|
We may consider a cross domain composition. In that case, the aggregator should know about all the composition domains.
We are experimenting the use of NLSR information to know about the reachable domains in order to consider them in an automatic way.
For example, if the NLSR database contains routes for /dom1, /dom2, and /dom2/sub1. We are investigating the behavior when using these entrees as delimiter for our composition :
/dom1/actuator1/dom2/agregator1/dom2/sub1/sensor1/dom1/sensor2
|--------1------------|---------2-------------|-------3-----------------|---------4---------|

That been said, I agree with you,  without naming conventions, it will be too many incertitude on the naming limits.

Thanks for your feedback, I hope that we will get more feedback from this mailing list and on the nest NDN workshop.

Charif

From: Marc.Mosko at parc.com [mailto:Marc.Mosko at parc.com]
Sent: Monday, May 23, 2016 10:26 AM
To: Mahmoudi, Charif (IntlAssoc) <charif.mahmoudi at nist.gov>; gabesilva2004 at yahoo.com; ndn-interest at lists.cs.ucla.edu
Subject: Re: [Ndn-interest] nfd and iot?

I provided an incomplete description of the 2nd case where “left most child” could fail even without packet loss.

A – B – C – D - E
    |-----------|

In this example, node A and E are the consumers from publisher C.  Node E can send its interests via B and D.  So, it could send the interest for /sensor/11 via D and /sensor/12 via B.  Node A, when sending an interest for “/sensor, exclude up to and including 10, left most child,” would then get /sensor/12 because E had already pulled that via B.  It would not get /sensor/11 because that data was never pulled down via B.  Node A would need some other mechanism to detect that it missed a message (again, I used simple monotonic sequence numbers so it would be easy, but if you use something else like timestamps its not so easy).

Marc

From: Ndn-interest <ndn-interest-bounces at lists.cs.ucla.edu<mailto:ndn-interest-bounces at lists.cs.ucla.edu>> on behalf of Marc <Marc.Mosko at parc.com<mailto:Marc.Mosko at parc.com>>
Date: Monday, May 23, 2016 at 10:10 PM
To: "charif.mahmoudi at nist.gov<mailto:charif.mahmoudi at nist.gov>" <charif.mahmoudi at nist.gov<mailto:charif.mahmoudi at nist.gov>>, "gabesilva2004 at yahoo.com<mailto:gabesilva2004 at yahoo.com>" <gabesilva2004 at yahoo.com<mailto:gabesilva2004 at yahoo.com>>, "ndn-interest at lists.cs.ucla.edu<mailto:ndn-interest at lists.cs.ucla.edu>" <ndn-interest at lists.cs.ucla.edu<mailto:ndn-interest at lists.cs.ucla.edu>>
Subject: Re: [Ndn-interest] nfd and iot?

While I agree that ICN can offer a lot of advantages over IP based IoT solutions, I do not think it’s quite so simple to use opportunistic caching as a replacement for something like an MQTT broker.  The broker offers several processing features, like the QoS attribute and wildcarding subscriptions.  Using opportunistic caching, consumers can miss publications if the cache gets evicted or maybe just due to adversarial timing of messages.

I looked at the poster, but was not able to determine if there was some mechanism you used beyond opportunistic caching and normal NDN data matching.  I think there are some shortcomings to using only those mechanisms in a service that is supposed to deliver (perhaps reliably) all messages in a publication stream to all currently subscribed clients.

Depending on the timing of publications and Interests, a client could miss messages.  For example, a sensor is publishing /sensor/1, /sensor/2, … /sensor/N.  If a consumer is asking for these with exclusions and “left most child” to try to walk them in order, it could still miss one due to packet loss.  Lets say it gets 1 … 10, then 11 is lost in network, so 12 is the “left most child” at the closest router.  It would then skip to 12, and unless you do some recovery mechanism at the consumer, 11 is lost without even know it.  If one is using timestamps or some other non-monotonic sequence number, then it would be very hard to know about the omission.  Even without packet loss, one could still miss messages.  If two consumers are pulling data but at slightly different rates, one could be on “10” and the other pulls “11” and “12” before the next interest from the first arrives.  “12” is now the “left most child” so it would miss 11.

Marc

From: Ndn-interest <ndn-interest-bounces at lists.cs.ucla.edu<mailto:ndn-interest-bounces at lists.cs.ucla.edu>> on behalf of "Mahmoudi, Charif (IntlAssoc)" <charif.mahmoudi at nist.gov<mailto:charif.mahmoudi at nist.gov>>
Date: Monday, May 23, 2016 at 9:06 PM
To: "gabesilva2004 at yahoo.com<mailto:gabesilva2004 at yahoo.com>" <gabesilva2004 at yahoo.com<mailto:gabesilva2004 at yahoo.com>>, "ndn-interest at lists.cs.ucla.edu<mailto:ndn-interest at lists.cs.ucla.edu>" <ndn-interest at lists.cs.ucla.edu<mailto:ndn-interest at lists.cs.ucla.edu>>
Subject: Re: [Ndn-interest] nfd and iot?

https://github.com/usnistgov/NamedIoT/files/272882<https://github.com/usnistgov/NamedIoT/files/272882/ndn-poster.pptx>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/ndn-interest/attachments/20160523/9507701c/attachment.html>


More information about the Ndn-interest mailing list