nanog mailing list archives

Re: Observations of an Internet Middleman (Level3) (was: RIP


From: Matthew Petach <mpetach () netflight com>
Date: Fri, 16 May 2014 12:21:57 -0700

On Fri, May 16, 2014 at 12:14 PM, James R Cutler <
james.cutler () consultant com> wrote:


All this talk about symmetry and asymmetry is interesting.

Has anyone actually quantified how much congestion is due to buffer bloat
which is, in turn, exacerbated by asymmetric connections?


James R. Cutler
James.cutler () consultant com
PGP keys at http://pgp.mit.edu


I think you might have the cart before the horse.

If there's no congestion on a peering link,
buffering doesn't come into play, at least
not within the transport infrastructure.

We're not talking congestion on the last mile
side, we're looking at congestion on the
interconnect links between networks,
typically 10G or 100G ports.  Unless
you're running those links near or at
capacity, buffering should be a complete
non-issue.  And if you're running those
links at capacity, then the congestion
is due to too much traffic, period, not
to the size of buffers involved on either
side of the link.  ^_^;

Thanks!

Matt


Current thread: