Re: [aqm] Immediate ECN: Autotuning AQM for RTT

[John Leslie <[e-mail]>]

Bob Briscoe <[e-mail]> wrote:
> 
> "Immediate ECN" slides:
> 
> <http://bobbriscoe.net/presents/1311ietf/1311tsvarea-iecn.pdf>
> 
> PS. This talk fell off the end of the TSVAREA agenda. It's mostly 
> relevant to AQM, but I didn't originally bring it to AQM, because it 
> affects 3 wgs: tsvwg, aqm & tcpm.

   (responding only on AQM...)

> In the AQM wg, there was dismay about CableLabs not including 
> anything about ECN in DOCSIS3.1. This talk is about AQM dynamics; and 
> how ECN can take out the 100ms of delay that CoDel and PIE introduce 
> - it's essentially about auto-tuning for RTT.

   I don't entirely follow what Bob is proposing; but in essence he
proposes that packets which would _not_ otherwise be dropped get an
ECN mark, and let the transport apply "smoothing" to correct for the
RTT of buffering necessary to fill the available bandwidth.

   There are various ways to accomplish this, and I won't consider
them here: the essential question is whether we consider latency
important enough to diddle with ECN signaling to allow short-RTT flows
to react to congestion sooner than long-RTT flows (rather than have
the forwarding node guess what RTT is proper).

   I strongly endorse work in that direction.

> It gives an interim recommendation for hardware designers that there 
> should be a second instance of the AQM algo for ECN packets so that 
> it can be configured with different parameters (think of WRED instead of 
> RED).
> 
> Specifically, for ECN packets:
> interval = 0 (for CoDel)
> max_burst = 0 (for PIE)

   As I understand it, Bob wishes to apply different parameters for
ECN-capable packets in order to remove the allowance for buffering
sufficient to cover the RTT. This is definitely not the only way to
think about the issue; and Bob could do a very good job of explaining
other ways (which I likely would prefer).

   But at this point, Bob is simply asking whether the whole idea
is worth talking about in the IETF...

====

   The AQM Friday agenda looks tight; so I'm not expecting Bob's
slides to be covered then. To be blunt, I will attend CLUE instead
unless ECN issues will be covered Friday...

   Regardless, I intend to follow the AQM list carefully.

--
John Leslie <[e-mail]>