perf number stuff (was: Re: [PATCH] ap_r* performance patch)

[Greg Stein <gs...@lyra.org>]

On Thu, Jan 18, 2001 at 07:47:12PM -0800, rbb@covalent.net wrote:
> 
> > > > [ I presume 2.0 has higher latency in handling the request, and then we slam
> > > >   it all onto the network at once ]
> > > 
> > > I've got a feeling it's the 9K buffer we use as opposed to the 4K 1.3
> > > uses.  It would be interesting to tweak that.
> > 
> > Yup. I'll give it a quick run and see what happens....
> > 
> > *blink* ... the numbers are the same. The strace shows that the writev() has
> > an extra item (since one is limited to the 4k length), so the change
> > definitely got in there. It is simply that there is no discernible speed
> > difference (in my testing, at least).
> 
> might be interesting to see a strace from 2.0 and 1.3 side by side.  There
> are a couple of read calls that we don't need, but I haven't had a chance
> to track them down yet.  hmmmmm.....

The strace came from a "telnet" session, so there will definitely be some
extra read() calls as Apache looks for the rest of the request, fails, then
moves into a select() to wait for it.

But for the rest: yes, a side-by-side will be great. Didn't Dean do one of
those already, but then we got sidetracked by the autoindex ugliness?

Cheers,
-g

-- 
Greg Stein, http://www.lyra.org/

Re: perf number stuff (was: Re: [PATCH] ap_r* performance patch)

[rb...@covalent.net]

> The strace came from a "telnet" session, so there will definitely be some
> extra read() calls as Apache looks for the rest of the request, fails, then
> moves into a select() to wait for it.

But we are seeing extra read calls even with HTTP/1.0 requests.

> But for the rest: yes, a side-by-side will be great. Didn't Dean do one of
> those already, but then we got sidetracked by the autoindex ugliness?

He ran a couple 2.0 strace's, but he doesn't still have the 1.3 strace's
as far as I know.

Ryan

_______________________________________________________________________________
Ryan Bloom                        	rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------