Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 23 Mar 2001 12:38:32 -0800
From:      "Jonathan Graehl" <jonathan@graehl.org>
To:        "Alan L. Cox" <alc@imimic.com>
Cc:        "Freebsd-Net" <freebsd-net@freebsd.org>
Subject:   RE: Linux Vs. FreeBSD Networking Performance
Message-ID:  <NCBBLOALCKKINBNNEDDLGEAMDNAA.jonathan@graehl.org>
In-Reply-To: <3ABAEC0A.994C6D2C@imimic.com>

next in thread | previous in thread | raw e-mail | index | archive | help
> Yes, we do.  In fact, the difference between FreeBSD and Linux is
> greater
> than 2x.  On equivalent processors, we demonstrated 1900 polygraph
> req/sec
> on FreeBSD 4.2 and 720 polygraph req/sec on a 2.2.14 Linux kernel.  It's
> also worth mentioning that the response time for FreeBSD at 1900 req/sec
> was faster than Linux at 720 req/sec.
>
> There are other advantages to FreeBSD, but kqueue is definitely
> at the top of the list.
>
> Alan

What would it take to get Linus to give the nod to an implementation conforming
to the kqueue API?  I remember him saying that he only wanted it to work for
file descriptors, and to only allow one kqueue per process - neither of which I
agree with.  The abstraction penalty for the capability of multiple filter types
and kqueue-as-selectable-fd is as minimal as a table lookup and a pointer
indirection.  If the kqueue API is overengineered, well, then, so is the
Berkeley Sockets API.

--
Jonathan Graehl
  http://jonathan.graehl.org/


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-net" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?NCBBLOALCKKINBNNEDDLGEAMDNAA.jonathan>