On Mon, Jul 18, 2016 at 5:05 PM, Neil Shepperd <[email protected]> wrote:
> October 2015:
> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=eb9fae328faff9807a4ab5c1834b19f34dd155d4
>
> Sounds like the speed up is pretty subtle, so I'm not surprised I didn't
> notice...

It is not so much the speed ups in loss recovery in rack that are
concerning me, it is the impact on congestion avoidance (which is sort
of undefined in the ietf draft). I have been observing a dramatic
decline in observed packet loss and a sluggishness in grabbing newly
available bandwidth - as well as oddities when congestion avoidance is
hit after exiting slow start. I was looking at traces and not seeing
the window open up at all...  even with reno... as well as one where I
was seeing queues drain completely and a loss of utilization at 10mbit
in cake.

Regrettably I have so many different pieces in flight (cake, wifi
device drivers, now this) that taking a few steps back to look at the
issues seems to be needed. Are we really tracking dropped packets
correctly?

So I can add a new one - What happens if I disable rack and/or test
with an older kernel?

> On Mon, 18 Jul 2016 at 05:05 Dave Taht <[email protected]> wrote:
>>
>> https://www.ietf.org/proceedings/96/slides/slides-96-tcpm-3.pdf
>>
>> --
>> Dave Täht
>> Let's go make home routers and wifi faster! With better software!
>> http://blog.cerowrt.org
>> _______________________________________________
>> Cake mailing list
>> [email protected]
>> https://lists.bufferbloat.net/listinfo/cake



-- 
Dave Täht
Let's go make home routers and wifi faster! With better software!
http://blog.cerowrt.org
_______________________________________________
Cake mailing list
[email protected]
https://lists.bufferbloat.net/listinfo/cake

Reply via email to