[master] 15ec79c Concentrate the acceptor-setup-session code.

Federico Schwindt fgsch at lodoss.net
Wed Mar 25 16:35:50 CET 2015


Attached is a patch in that effect.

I have no empirical evidence this will improve things but I believe Fastly
and others are doing it with good results.



On Sat, Mar 21, 2015 at 12:57 PM, Federico Schwindt <fgsch at lodoss.net>
wrote:

> Yeah, me not reading the code properly.
>
> So this could potentially help.
>
> On Wed, Mar 18, 2015 at 5:40 PM, Poul-Henning Kamp <phk at phk.freebsd.dk>
> wrote:
>
>> --------
>> In message <
>> CAJV_h0bubvWpPGBogGQRK7+rGTjHJ6rTXbf8kWQJU3L5JVMz4Q at mail.gmail.com>
>> , Federico Schwindt writes:
>>
>> >My understanding and what I've read is that if you have multiple threads
>> >accepting connections it'll will behave much better when running in
>> >multiple cores which is pretty much everyone these days.
>> >
>> >Now looking at master, it looks we only have one acceptor per address and
>> >not per pools.
>>
>> We have one thread per socket per pool.
>>
>> --
>> Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
>> phk at FreeBSD.ORG         | TCP/IP since RFC 956
>> FreeBSD committer       | BSD since 4.3-tahoe
>> Never attribute to malice what can adequately be explained by
>> incompetence.
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://www.varnish-cache.org/lists/pipermail/varnish-dev/attachments/20150325/bab1d648/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-Set-SO_REUSEPORT-in-the-socket-if-available.patch
Type: text/x-patch
Size: 1500 bytes
Desc: not available
URL: <https://www.varnish-cache.org/lists/pipermail/varnish-dev/attachments/20150325/bab1d648/attachment.bin>


More information about the varnish-dev mailing list