httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dean Gaudet <dgau...@arctic.org>
Subject Re: SAFE_UNSERIALIZED_ACCEPT
Date Mon, 18 Aug 1997 03:22:51 GMT
I considered partial serialization in a message a few months back ... 
saying that yeah it is probably worth looking into.  But when I found that
sysv semaphores were so extremely lame I was put off. 

Dean

On Sun, 17 Aug 1997, Ben Laurie wrote:

> Dean Gaudet wrote:
> > 
> > Ok maybe this is a good enough heuristic for deciding when to use
> > SAFE_UNSERIALIZED_ACCEPT.  Or maybe not.  I dunno.  Ok what I have here is
> > a well tuned server (this is the minimal #syscalls configuration).
> > Everything is identical between these two runs except that the first run
> > is with SAFE_UNSERIALIZED_ACCEPT, and the second without.  (Both results
> > are repeatable.)
> > 
> > So it looks like unserializing is beneficial on linux... then again, this
> > is a dual processor box and maybe it's flock lameness.  Bah.
> > 
> > Oh hrm.  Total throughput is higher with unserialized, but notice the
> > latency of the connection is worse.
> 
> Because the connection wakes up _all_ the children? Is it worth
> exploring partial serialization (i.e. use a semaphore to reduce the
> number of children accepting)?
> 
> Cheers,
> 
> Ben.
> 
> -- 
> Ben Laurie            |Phone: +44 (181) 994 6435|Apache Group member
> Freelance Consultant  |Fax:   +44 (181) 994 6472|http://www.apache.org
> and Technical Director|Email: ben@algroup.co.uk |Apache-SSL author
> A.L. Digital Ltd,     |http://www.algroup.co.uk/Apache-SSL
> London, England.      |"Apache: TDG" http://www.ora.com/catalog/apache
> 


Mime
View raw message