[Coin] RG formation - IAB Feedback

["Schooler, Eve M" <[e-mail]>]

Hi All,
We presented a COIN PRG overview to the IAB Monday morning 11/18 at Colin's request as a necessary step en route to possibly graduate from PRG to RG. Below are the comments we received. We plan to discuss briefly at the meeting on Friday.
Eve
(on behalf of J/E/M)

Feedback from IAB

-        Christian Huitema: There should be greater consideration given to privacy and security issues and solutions - early on

-        Ted Hardie:

*        One thought: look at Commercial 5G systems and their definition and architecture of the Edge, because the 5G community likely can provide guidance on an approach to Privacy/Security. For example around their Slices solution. They offer a relationship between the customer and the slice provider.

*        Can the work in DCs shed light on the APIs for COIN - will cloud-based computations have the same API as Edge-only computations?

-        Stephen Farrell:

*        Without considering security/privacy, the architecture is misplaced, and relying on homomorphic ciphers for security will be difficult. [Elsewhere, in hallway conversations, growing consensus that data will remain encrypted throughout its life cycle - what's the implication of that?]

*        What's been the most interesting finding so far?

*        Marie-Jose:

?  Noa Zilberman's paper - that it came out of the Data policy community - how to sell services to customers

?  Programmable networking devices

*        Jeffrey:

?  How significantly in-network compute can improve the distributed systems inside the DCN, probably also applicable to edge computing

?  Co-Flow notion: using in-network compute to process the packets from correlated flows

*        Eve:

?  Not only a convergence of compute and network -  BUT ALSO storage

?  The need to ponder WHAT can the IETF bring to the discussion/solution space, in the way of protocols and architecture?

?  For example what will we leverage, extend, throw-away and offer as new in the way of existing standards? For example, for Edge Data Discovery, should we use, extend, propose an alternate solution to DNS.

-        Brian Trammell:

o   Yes! Stress the Internet focus, as compared with other communities looking at Edge computing and COIN!

o   Noted that there's a fundamental tension: nodes that do the computation need access to the data; other nodes need to be prevented from this; e.g., been thinking about opportunistically storing data in other peoples' nodes but at the same time be able to ensure they cannot actually see it

o   Security model is critical if the work is to be relevant.

o   Goal should be to bring some of the work into the world.

-        Wes Hardaker:

o   Threat models vary for different applications. Different model for security when there is a single owner of the resources and data vs the security solution when there are different owners, as with COIN - e.g., threat models for IoT

o   Would be just as useful to Identify the things that are not suited - dream big and fail

o   Suggested some use case focus - seems like we haven't identified a primary or main use case

-        Re security/privacy know-how - continue to bring some/more of our questions to the DIN RG, and leverage solutions under scrutiny there

-        Recommendation: find people in the research community who are approaching the COIN problem from the computing standpoint vs network perspective

*        The way things are scheduled from DC standpoint vs once resources are liberated from the DC and are more distributed

*        What is sent in the clear on the lines or not

-        Have we considered CDNs in COIN? Yes, not only CDNs but reverse CDNs (rCDNs)