Re: Patch: accept filters for NetBSD

[Thor Lancelot Simon <tls%rek.tjls.com@localhost>]

On Tue, Jan 29, 2008 at 08:19:29PM +0200, Elad Efrat wrote:
> Thor Lancelot Simon wrote:
> 
> >>What's the motivation of adding the accept filters? I understand one may
> >>be performance -- are there any relevant benchmarks? Is it possible to
> >>hear more about why this is necessary, and what are planned future
> >>extensions, if any?
> >
> >It is not necessary.  It provides an opportunity to optimize the processing
> >of certain application-layer protocols (the ones that like to make a lot
> >of connections at once)
> 
> I'm afraid I don't agree benchmarks are unnecessary.

You asked "Is it possible to hear more about why this is necessary".  I
responded that "it is not necessary".  Given that I think you probably
misunderstood my response.

I'm glad to get you some benchmark numbers.  It's not practical for me
to generate them using my actual application for accept filters, mostly
because of the amount of explanation I'd have to provide about how it
works to make its performance metrics interpretable by others.  But I
can try to throw something together for you with Apache.

>From my point of view, Apache is so much slower than the applications I
deal with that the performance characteristics may be dramatically
different (large Apache configurations seem to handle perhaps 10,000
connections/second; my proprietary application is about an order of
magnitude better, but is not a pure-userspace application).  But I will
get you what I can!

I think the effect can be observed using inetd, and that it's a useful
illustration, so I will still polish that code up as, at least, an
example.

Thor