tinc devel - Dec 2015 - [PATCH] Receive multiple packets at a time

Dave Taht, on Wed 02 Dec 2015 14:13:27 +0100, wrote:
> More recently Tom Herbert was working on udp encapsulation methods in
> the kernel "foo over udp"
> 
> https://www.netdev01.org/docs/herbert-UDP-Encapsulation-Linux.pdf
> 
> https://lwn.net/Articles/614348/
> 
> which preserve things important at high rates like GRO/GSO.
Yes, FOU will probably get the highest rates, but it doesn't support
even basic encryption, and is harder to set up for normal people :)
> > At least for now we could commit the recvmmsg part?
> 
> Not my call. It is a linux only thing, so far as I know.
ATM yes.  I wouldn't be surprised that other OSes adopt it, though.

Samuel

Dave T?ht
Let's go make home routers and wifi faster! With better software!
https://www.gofundme.com/savewifi


On Wed, Dec 2, 2015 at 2:26 PM, Samuel Thibault
<samuel.thibault at ens-lyon.org> wrote:
> Dave Taht, on Wed 02 Dec 2015 14:13:27 +0100, wrote:
>> More recently Tom Herbert was working on udp encapsulation methods in
>> the kernel "foo over udp"
>>
>> https://www.netdev01.org/docs/herbert-UDP-Encapsulation-Linux.pdf
>>
>> https://lwn.net/Articles/614348/
>>
>> which preserve things important at high rates like GRO/GSO.
>
> Yes, FOU will probably get the highest rates, but it doesn't support
> even basic encryption, and is harder to set up for normal people :)
I guess my meta point is driven by my headaches. Getting per packet
processing to scale up past 100Mbit is hard without offloads even on
embedded hardware considered "high end". (I was recently testing the
peeling[1] feature of cake[2] only to see it cut performance by over
half). Getting ANY modern platform past 10gbit is *really* hard. [3]

So while recvmmsg and optimizations like that are needed, some harder
thinking about the other brick walls coming up would be helpful..
(which is why I stopped)
>> > At least for now we could commit the recvmmsg part?
>>
>> Not my call. It is a linux only thing, so far as I know.
>
> ATM yes.  I wouldn't be surprised that other OSes adopt it, though.
Given how slow other OSes are evolving, I would not hold my breath,
and retain code paths that worked with them.

[1] https://lists.bufferbloat.net/pipermail/cake/2015-November/000846.html
[2] http://www.bufferbloat.net/projects/codel/wiki/CakeTechnical
[3]
http://netoptimizer.blogspot.se/2014/10/unlocked-10gbps-tx-wirespeed-smallest.html

Dave Taht, on Wed 02 Dec 2015 14:41:35 +0100, wrote:
> I guess my meta point is driven by my headaches. Getting per packet
> processing to scale up past 100Mbit is hard without offloads even on
> embedded hardware considered "high end".
In my tests I was getting 800Mbps with common laptop Gb ethernet
devices.
> >> > At least for now we could commit the recvmmsg part?
> >>
> >> Not my call. It is a linux only thing, so far as I know.
> >
> > ATM yes.  I wouldn't be surprised that other OSes adopt it,
though.
> 
> Given how slow other OSes are evolving, I would not hold my breath,
> and retain code paths that worked with them.
Sure!

Samuel