[Bloat] Packet drops, ECN and ECN+ [was: Thoughts on Stochastic Fair Blue]

[Juliusz Chroboczek <jch@pps.jussieu.fr>]

>>> [1] Aleksandar Kuzmanovic. The power of explicit congestion
>>>    notification. In Proceedings of the 2005 conference on Applications,
>>>    technologies, architectures, and protocols for computer
>>>    communications. 2005.

>> http://www.cs.northwestern.edu/~akuzma/doc/ecn.pdf
>> 
>> appears to be the paper Juliusz cited.

Indeed.  See Figure 3, where you clearly see TCP's admission issues in
the presence of ECN.  Compare to Figure 2, where dropping at high
congestion levels works around the issue (mostly).  The authors of the
paper suggest ECN marking SYN packets ("ECN+") to overcome this issue;
my reading of their data is that dropping at high congestion levels is
a good workaround.

(And in case you wonder why the Wikipedia page on ECN agrees with me --
I wrote it.)

>> I haven't had time to read the paper thoroughly, but I don't argue
>> with this - if the marking probability goes above 1/2 then you
>> probably have an unresponsive flow anyway.

Hmm... for TCP, a mark/drop rate of 1/2 means that the fair share is
slightly less than one packet per RTT, right?  While not very good,
that's quite likely to happen in The Real World, especially for low RTT
and browsers opening multiple connections per page.

(And yes, as far as I can see, transitioning to a low-latency Internet
will require increasing mark/drop rates dramatically.  I'll let you draw
your own conclusions about whether we can fight bufferbloat without ECN.)

> If nothing else, I take away from this paper that ECN should be applied
> (at least) on servers (and they advocate clients and routers) to TCP
> control packets (e.g. SYN and ACK packets) as well as data packets
> despite the potential (accepted admin legend???) that this might be a
> "bad thing" for reasons of aiding a potential SYN-flood attack vector. 

Unfortunately, other issues have been found with "ECN+" since the paper
was published, which is why it is no longer being proposed for deployment
on the Public Internet.  It's been a couple years since I've looked at
that stuff, though, so it might take me some work to find the reference.

--Juliusz