* Zach Brown <[EMAIL PROTECTED]> [010421 06:47] wrote:
> [apologies for missing the original post and replying to a reply..]
>
> > > - A round-robin token-passing scheme to determine which process gets
> > > to do the accept(). Turns out it's very bad to just have all the
> > > processes do an accept(), since every time there's a new connection
> > > *all* the processes wake up. The context switches totally kill
> > > performance. But a properly tuned round-robin scheme works great.
>
> In my apache tuning adventures, including the insanity at the zd mindcraft
> tests, I've never seen accept() hurding be a real measurable problem for
> the simple reason that when you're under load your waiters aren't waiting
> in accept(), they're off doing work. The only time this actually occurs
> is when you're entirely idle and get a new connection.
>
> or so the numbers have lead me to beleive. Its still an annoying
> design, but has someone come up with real numbers to show that accept()
> hurding is a problem for waiters that do real work after accept() ?
Accept herding isn't a problem under FreeBSD because the kernel doesn't
allow it to happen.
--
-Alfred Perlstein - [[EMAIL PROTECTED]]
Daemon News Magazine in your snail-mail! http://magazine.daemonnews.org/
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
