httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
Subject Re: cvs commit: httpd-2.0/modules/http http_core.c
Date Wed, 24 Jan 2001 22:14:36 GMT

> > > Do we have any performance testing to back this up? Why would sendfile of
> > > less than 8k be slower than read-file/send-socket?
> > > 
> > > I'd much rather see it remain simple and always use sendfile when a file is
> > > supplied.
> > 
> > No, I was seeing us use sendfile on 10 bytes of information from a file,
> > which is insane, because that is all we send, even if there is another 9K
> > file two buckets later.  I needed a number, and I planned to do testing to
> > ensure it is the current number.  I chose 8K, because it is logical and it
> > matches the point at which we do not convert a file to an MMAP.
> Cool... I look forward to the numbers. I can't say that I know them either
> :-), but am a little worried about making "optimizations" like these without
> knowing.
> btw, constructing an MMAP is reasonably expensive, so the conversion point
> is handy. sendfile() is just a function call, so I don't think we'll see the
> same overhead as MMAP-creation.

The overhead isn't the limiting factor here.  Take a look at the example I
sent earlier.  If we send all files using sendfile, we are likely to see
some VERY small packets with byterange requests.


Ryan Bloom               
406 29th St.
San Francisco, CA 94131

View raw message