[T2TRG] Review of draft-hong-t2trg-iot-edge-computing-03

[Marie-Jose Montpetit <[e-mail]>]

As promised I reviewed the draft-hong-t2trg-iot-edge-computing-03 and here
are my comments.

*General:*

This is a needed document as asummary of the current landscape in edge and
IoT is important as the field is rapidly expanding. As you will see in the
comments below, one of my issues is the fact that the document is big on
requirements and a bit less on implementation.  This is fine as long as it
would be advertised as such in the introduction. In a very dynamic
environment I agree that it will be hard to maintain a list of
implementations and I welcome the use of the appendix for that.

For me the weakness is in the use cases which I think should be expanded.

But in general it is a good paper.

*Some notes and typos:*

 Internet of Things was coined by Henry Holtzman of the MIT Media
Laboratory before the RFID people borrowed it :)

Repetition: Cloud computing has been a predominant technology which has
virtually *unlimited* capacity in terms of storage and processing power.
The availability of virtually *unlimited* - and actually it’s unlimited
only if we consider that energy is also unlimited and has zero cost… Which
is another reason to offload to the edge and maybe should be mentioned.

Today's cloud-based service models are not suitable for these applications.
- short list of reasons would be good here.

It is predicted that by 2019 - any update for 2021?

Edge computing, under certain aspects also referred to as fog computing -
fog computing is more associated with home or even personal computing
devices and edge is a bit wider in scope and worth mentioning even if you
make it clearer later in the document

Example of IoT Edge Computing Use Cases > Examples

The examples are very “traditional” would be worth adding a few more
(AR/VR, agriculture 4.0, intelligent vehicles…) if not it seems very
limited.

In the common aspects I wonder why there is no mention of IPFS.

In section 4 I think that the creation of more “intelligent” systems that
require IoT platforms to interconnect is another major challenge; for the
moment all IoT is developed in silos that connect the device(s) to the
cloud directly. We we want to add an “intelligence” or “data” layer for
responding to the service requirements then another challenge is to create
some open APIs or interface inside the IoT gateway(s) or computing nodes
that would allow local pipelining of the functionalities.

In OAM I think data discovery is also important for data driven systems to
“borrow” data to help in local decision making.

The Functional Components section is more a Functional Component
Requirements - unless you want to add specific examples.

In network computation in not just offloading as it also includes decision
making, local image or sensor processing and data reduction.

Data management is also a list of requirements and not focused on common
functionalities.


Happy Sunday and my best for a successful IETF week!

mjm
Marie-José Montpetit, Ph.D.
[e-mail]