[e2e] Mixed ECN and Non-ECN traffic flows.

[Mutlu Arpaci]

Hi,

in general, a better throughput performance for ECN-flows as opposed to
non-ECN flows should be expected, due to the fact that non-ECN packets are
simply dropped by the AQM mechanism causing retransmissions and RTOs. The
level of "unfairness" between ECN and non-ECN flows depends on the AQM
paramaters (max_p for RED) and the composition of the traffic. Quoting from
my dissertation:

" [...] for a given traffic scenario, the performance improvement of ECN
flows increases with increasing max_p.
In addition, for a given max_p, the performance improvement of ECN flows
increases as the number of NonECN flows increases. This second point implies
that as the traffic load increases, the NonECN connections are punished more
severely than ECN flows."

The question of "fairness" regarding ECN vs non-ECN flows is not
straightforward in my opinion, because one can argue that ECN-flows should
outperform non-ECN anyway (otherwise what's the point?). This may also
promote the usage of ECN in the network. Both ECN and non-ECN based AQM
implementations aim to avoid congestion in the network, but ECN achieves a
much higher resource efficiency, and provides better performance to the
end-user.

You can find my thesis at:

http://www.csc.gatech.edu/~mutlu/


thanks,
Mutlu Arpaci


> > > -----Original Message-----
> > > From: Chong Poh Kit [mailto:thesoothsayer at pd.jaring.my]
> > > Sent: Friday, October 11, 2002 5:53 AM
> > > To: end2end-interest at postel.org
> > > Subject: [e2e] Mixed ECN and Non-ECN traffic flows.
> > >
> > >
> > > Hi,
> > >
> > > I've done some simulations using a mixture of ECN and non-ECN
> > > capable TCP
> > > flows over NS-2 and I've discovered that at a bottleneck link
> > > with a single
> > > FIFO queue, ECN flows tend to grab more than their fair share of the
> > > bandwidth when an AQM (RED,BLUE) is used at the router.
> > >
> > > I believe this is due to the fact that ECN traffic needs more
> > > aggressive
> > > marking to control the rate of congestion while the non-ECN
> > > traffic suffers
> > > from the increased packet drops. Furthermore, the packet
> > > drops results in
> > > constant retransmissions and thus, a hit in the goodput of the non-ECN
> > > flows.
> > > This problem is exacerbated when the latency is low because a more
> > > aggressive packet dropping/marking probability is needed to
> > > control the
> > > throughput of TCP.
> > >
> > > I was wondering if this is a trivial problem as I've not seen
> > > any research
> > > done on it so far?
> > > My opinion is that as the percentage of ECN capable end hosts
> > > increases this
> > > would pose a problem of unfairness in the future.
> > >
> > > Also, could anyone direct me to any statistics that show the
> > > percentage of
> > > ECN capable end hosts currently deployed on the Internet?
> > >
> > > Thanks.
> > >
> > > Regards,
> > >
> > > Chong Poh Kit
> > > Post-graduate Student
> > > Multimedia University, Malaysia
> > > Tel : (6)016-3174831
> > >
>
>
>
>
> _________________________________________________________________
> Join the world's largest e-mail service with MSN Hotmail.
> http://www.hotmail.com
>