[Iccrg] Explicit feedback

[John Leslie]

Michael Welzl <michael.welzl at uibk.ac.at> wrote:
> 
> There is no doubt that explicit feedback from routers is useful
> for a congestion control mechanism - this has been illustrated
> by mechanisms like XCP and Quickstart (and several others).

   It's obvious that sending packets we know will be discarded is
sub-optimal. The question is, how can we know they'll be discarded?
Clearly, this is an area deserving research.

> Yet, such mechanisms are critical because they require additional
> work in routers. So what is realistic? How much extra effort
> can one really assume from routers, when the scalability of
> a mechanism must not be constricted?

   At the research level, I'm not sure this question arises.

   Clearly, in the "real-world" Internet, there are routers which
cannot spare any per-packet processing time. But these "should" not
be the ones experiencing the congestion. (They get maximal advantage
of aggregation.)

   In principle, it should be possible to convey congestion status
in routing updates, and avoid introducing packets to congested routes
at packet rates which cannot be sustained.

   (It's not clear how we might retrofit that into current Internet
routing; that's another research area...)

> And how should the provision of such feedback be specified - e.g.
> with pseudocode, as in the appendix of the XCP paper, or something
> else...?

   Not an issue. Presumably we're smart enough to code from any
sufficiently detailed explanation.

> This kind of knowledge is missing in the congestion control
> landscape. Perhaps we can we make a change here.

   I doubt we can make much of a change at the backbone in the short
term. But research into what sorts of congestion information might be
carried in routing updates seems quite useful.

--
John Leslie <john at jlc.net>