Re: [Coin] Challenges for routing systems

[Adrian Farrel <[e-mail]>]

Thanks Dave, you ask some important questions.

> Hi - thanks for the comprehensive list to get a parsing/decomposition
> of the possible research areas to include in a work program. It would
> also be valuable in general for IRTF folks to get a sense for which
> research groups around the world are working in which of these areas
> and what some of their recent publications are.
>
> I don’t follow routing research much beyond issues related to the
> narrow case of ICN, so I haven’t seen much in most of these areas in
> the conferences and journals I do follow. 

Yes. We started work on a "literature survey" of sorts, and you can find it at https://datatracker.ietf.org/doc/draft-king-irtf-semantic-routing-survey/ and then we got swamped! As Jon C pointed out, the concept of semantic routing is not new: we found a lot of schemes (some weirder than others) going back a lot of years, and some that have just been introduced. Since the last revision of the survey we have been bombarded with references and papers - we should (like good soldiers) update the survey, but time and resources are limited and we are behind with this work (anyone got a spare researcher?).

What we learned was that these projects (some research, some engineering) typically limited themselves to specific environments or use cases. They did not consider their wider applicability, and they did not consider many features considered essential in networking. That's what led us to produce the "challenges" document - it seemed to us that whatever research or engineering was going on, it was not thinking about how routing systems work and what the schemes under investigation needed to consider in order to assess their value.

I am aware of a research project being put together at the moment that includes an element of "semantic routing" with an attempt to make this more general (based on a "unifying abstraction" as Jon puts it), but as this is at a pre-fund-application status, I don't think I should talk about it like it is really happening.

> Of the things on this list, which do you think are good topics
> for discussion and research agendas for COINRG? My reading
> of is that COINRG would be a candidate venue for part of #4,
> but not for any of the other items. What do others think?

Your opinion here is really helpful. I am relatively new to COIN, but I am reliably assured that substantial parts of our work should be able to live comfortably within COIN.

In as much as COIN concerns itself with forwarding packets at the IP-layer (i.e., hop-by-hop, not in an overlay or higher layer, and not selecting the next remote transit processing node) I should think that all of the points must be in scope.

Cheers,
Adrian

> On Feb 3, 2022, at 2:53 AM, Adrian Farrel <[e-mail]> wrote:
> 
> Hi again,
> 
> This email introduces some of the challenges that exist for
> routing/forwarding systems.
> 
> Our motivation in putting this document together has been our experience
> with many pieces of research that are based either on simple networks or on
> very stable networks. Such approaches can show that a new routing system can
> work, but they don't evaluate the utility of the system in real-world
> networks. Thus, we are attempting to collect guidance for the research
> community to help make routing research more grounded and so more valuable
> as input to the engineering community.
> 
> While our personal drive behind this work comes from Semantic Routing (and
> particularly observing which Semantic Routing research work is in need of
> additional effort), we hope that our ideas will be valuable across the
> research community where-ever new approaches to packet routing/forwarding
> are being considered. As with the core Semantic Routing work, our focus here
> is at layer 3 (IP layer), packet-level routing and forwarding. We are not
> spending much time considering the application of routing at higher layers
> or through overlays: while many similar considerations may exist, the nature
> of things is that as you abstract upwards, the number of network nodes and
> links becomes fewer, the network becomes more stable, and the problem space
> becomes simpler.
> 
> I won't repeat the contents of
> http://datatracker.ietf.org/doc/draft-king-irtf-challenges-in-routing in
> this email since the purpose of the draft is to provide the more detailed
> basis for discussion and record, but it may be helpful to give context by
> listing out the top-level considerations that we have so far:
> 1.  What is the scope of the routing proposal?  
>      Global, backbone, overlay, gateway-based, limited domain?
> 2.  What will be the impact on existing routing systems?
> 3.  What path characteristics are needed to describe the desired
>     IP-layer paths and as input to route computation?
> 4.  Can we solve these routing challenges with existing routing tools
>     and methods? Is new hardware needed? Do we need new routing
>     protocols?
> 5.  Do we need new management tools and techniques?
> 6.  What is the impact on the security of the routing system?
>      What are the implications to privacy?
> 7.  What is the scalability impact on routing systems?
> 8.  To what extent can the proposed routing scheme be applied
>     to multicast transmission schemes?
> 9.  What aspects need to be standardized?
> 
> We have been discussing this work within the IETF (specifically the Routing
> Area discussion list and the RTGWG) because the engineers there will have a
> perspective on what gets forgotten and what problems they have seen with
> deploying new routing techniques, but the purpose of the work is to give
> input to researchers. And, of course, we have started to take this material
> to workshops and conferences to spread the word.
> 
> In the context of COIN, we are interested to know how we could make this
> more relevant to what COIN researchers are doing, and we would love to have
> feedback on whether this is useful, scary, or irrelevant.
> 
> Best,
> Adrian
> 
> -- 
> Coin mailing list
> [e-mail]
> https://www.irtf.org/mailman/listinfo/coin