[Ndn-interest] Application of ICN in the IoT
Aisling O' Driscoll
a.odriscoll at cs.ucc.ie
Fri Sep 15 00:37:05 PDT 2017
Thank you so much for you detailed and thoughtful responses. It’s great to hear your insight on the topic as well as a longitudinal perspective.
I certainly look forward to talking to you and those mentioned below at ACM ICN.
From: Lixia Zhang [mailto:lixia at cs.ucla.edu]
Sent: Thursday 14 September 2017 04:09
To: Aisling O' Driscoll
Cc: ndn-interest at lists.cs.ucla.edu
Subject: Re: [Ndn-interest] Application of ICN in the IoT
On Sep 13, 2017, at 5:19 AM, Aisling O' Driscoll <a.odriscoll at cs.ucc.ie> wrote:
I’m new to the mailing list so greetings! I recently posted on the ICNRG mailing list and Lixia Zhang kindly pointed me in this direction as a mailing list of interest.
I’m attending the ACM ICN conference in 2 weeks so looking forward to meeting many of you there.
as I know, several people from the NDN team will be attending ICN,
Prof. Alex Afanasyev from FIU (like you, he's also new there),
Prof. Burke Jeff from UCLA, and
Prof. Christos Papadopoulos from Colorado State.
hope you'll meet them all there; there may be others that I missed.
I recently joined the academic staff in the Department to Computer Science in University College Cork (Ireland) and am interested in investigating ICN, particularly its application within the IoT and edge networks. My background is in network communication protocols, specifically geo-routing and location service protocol design for ad-hoc networks.
many people here share that interest as well, as you know NDN is really a great fit for ad hoc networking.
This is an area that is new to me but one that I’m interested in exploring. If I anyone has any suggestions/opportunities for the best way to connect in with existing work in this space or is searching for an able and interested person to further explore some of the existing research challenges please let me know.
People here would be really interested in collaborating on IoT.
Alex and some other NDN team students are part of the full day tutorial team for ICN IoT on Tuesday Sept 26, hope that could offer some useful info to you and serve as a starting point to explore collaboration.
Meanwhile if you are interested in reading some papers: browse NDN publication page at https://named-data.net/publications/
(we plan to sort papers into different categories but just have not got there yet)
A high level question for you that I would really be interested in hearing your views on:
There seems to be a school of thought that adoption of ICN (whatever the architecture) may never come about because of the amount of effort and money already invested in TCP/IP.
during my graduate school days, I heard exactly the same reasoning about telephone network infrastructures.
A more optimistic view is that the deployment will most likely being a hybrid ICN over IP with a slow phase out over 20-30 years.
couple of thoughts here to share:
1/ Depending on how one defines "slow phase", one could say that TCP/IP took a very "slow phase" from the inception to the eventual roll out.
e.g. people credit Cerf/Kahn's seminal 1974 paper as the first blueprint of TCP/IP, yet it took about 20 years before Web showed up, and some more years before the exponential growth of the Internet
(I sort of recall the famos Bill Gates "Internet Tidal Wave" memo was around mid 90's, urging MicroSoft to catch Internet revolution, i.e. MS was yet to move to TCP/IP by then)
2/ also depending on how one defines "a hybrid ICN over IP": upon birth a new architecture necessarily runs on top of the existing infrastructure -- if this is called hybrid, then IP was hybrid over telephone networks: people with gray hair can all remember old days of dialing up to Internet, even ARPAnet was built over leased telephone lines. But personally I would not call this as hybrid, but over : IP can run over anything, both Ethernet and dailup carry IP datagrams, and IP does not care.
early last year KC and I did an interview with ;login to answer some FAQs, the ending remark there touched on this question:
Extending this to the IoT, I would foresee that a hybrid stack isn’t feasible as the goal is to lighten the stack not increase its complexity which may not be feasible for resource constrained devices.
I completely agree with you here.
So does sufficient quantitative evidence exist to provide a very strong and compelling motivation for why IoT should transition from the CoAP/UDP/RPL/IPv6/6LoWPAN/802.15.4 stack to an ICN approach in terms of network performance benefits OR is the argument that IoT is still relatively in its infancy and thus a clean slate approach is still feasible given that ICN is naturally a better fit without providing a strong migration argument?
all great questions.
personal opinion: in addition to IoT deployment being still in its infancy, there is another important factor that enables IoT to take on a different design: being at the edge, there is no dependency on the deployed infrastructure to make changes, before you or I decide to try some entirely new IoT products, so that my home wont freeze when some code 500 miles away fails, like the story told here http://www.computerworlduk.com/galleries/infrastructure/top-10-software-failures-3599618/
(just type in "cloud failures" to google, there is never a shortage of new stories)
I do note that some quantitative evaluation was conducted in a sub-section in a paper “ICN in the IoT: Experiments with NDN in the Wild” (really interesting paper).
It seems like there is significant further work to be done to provide quantitative motivation. Is this a fair observation? I’d be interested to hear your opinions.
I'll leave the discussion on the paper to a later time. But I think there may be a qualitative comparison, even before one goes quantitative.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Ndn-interest