[clue-tech] TCP/IP and the Bufferbloat Problem

[Jed S. Baer]

On Wed, 08 Dec 2010 10:51:15 -0700
Jim Ockers wrote:

> Hi,
> 
> Collins Richey wrote:
> > On Tue, Dec 7, 2010 at 7:24 PM, Jed S. Baer
> > <cluemail at jbaer.cotse.net> wrote: 
> >> Fascinating read. Even if I didn't understand a lot of it.
> >>
> >> https://gettys.wordpress.com/2010/12/06/whose-house-is-of-glasse-must-not-throw-stones-at-another/
> >
> > Most of us have the general impression "lots of buffers - a good
> > thing", but obviously this is not always the case.
> >
> > A second vote for "I didn't understand a lot of it".

IP networking has always been something I wish I knew more about, but
never seem to get to studying it. But I do believe I grok the essence of
this, which is that the TCP congestion alogorithm is being fooled when a
buffer upstream isn't signaling (albeit indirectly) that packets are
coming in too quickly.

> I read this and it's curious that he never mentions anywhere what the 
> MTU size was and whether changing it had any effect on data transfer 
> rates or buffering efficiency.

All the posts so far are in the category 'bufferbloat':
https://gettys.wordpress.com/category/bufferbloat/

I don't recall seeing anything about MTU, but he did play with txqueuelen.

-- 
Ok, so we should be thinking of a lovable, cuddly, stuffed penguin
sitting down after having gorged itself on herring. Still with me? 
 -- Linus Torvalds