Re: [Codel] [PATCH net-next] codel: use Newton method instead of sqrt() and divides

[dave taht <dave.taht@gmail.com>]

On 05/13/2012 11:33 PM, dave taht wrote:
> On 05/13/2012 11:17 PM, Eric Dumazet wrote:
>> On Sun, 2012-05-13 at 23:00 -0700, dave taht wrote:
>>> I am sitting here with the latest fq_codel implementation on 
>>> cerowrt, and
>>> this new 16 bit change.
>>>
>>> With either interface, running netserver on the router, I get
>>> ~260Mbit/sec out of it,
>>> with fq_codel on. It's quite marvelous, as ping times stay in the 2-3ms
>>> range...
>>>
>> If you disable TSO and GSO, do you still get good numbers ?

This is the first time I've had these puppies not crash in a month under 
workloads like this.

*Everything* I have has TSO and GSO off by default. That stuff has cause 
me no end of grief.

But for you...

I just fired up a long duration ipv6 test with TSO/GSO on the driving 
host, at 1Gbit, with fq_codel on all sides... with  10 TCP_MAERTS, 10 
TCP_STREAMS, over pure IPv6

and am running test for 6000 seconds.  and I'm going to dinner. :)

At the moment I see ping times 1.553 avg, 5.999 max, I can do a CDF plot 
of the next run.

Actually I think ipv4 fragmentation is still busted in cero, but I'm 
getting ther....