> No, it's telling _from_ kernel to userspace that request arrived, or
> that it's finally sent. And context switch between multiple processes
who needs multiple processes? My sily siginfo server has done 3500
requests a second with a single thread over localhost in 2.3.x.
yes, it cheats ;)
> What applications improve their performance after "thundering herd" on
> accept/select problem's solution?
siginfo events on the listening socket, pointable at individual threads
with F_SETOWN.
> of certain kind of applications (HTTP and possibly FTP servers -- even
> for SMTP and NNTP the possible improvements aren't noticeable because of
> heavy processing involved).
no, this model works well for any system that needs to manage lots of
concurrent nonblocking I/O. serving NNTP clients with this model would
make a ton of sense.
anyway, the interfaces already mostly exist to do a really well done
userspace i/o engine that can serve static http very quickly. Its simply
a matter of writing the code. I'm joining Dean in his nice hacking.
I'm not meaning to deter people from messing around with a kernel httpd.
I just don't think its the right approach to the problem. What happens
with smb isn't fast enough? ldap? nntp?
also notice that the newer apache using this stuff will also be usable
under solaris/*bsd, etc, etc. does khttpd even work under 2.3 yet? 2.0?
god forbid, a non linux unix?
-- zach
- - - - - -
007 373 5963
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/