Mailing List Archive

>
> Something's missing....
>
> Brian

Someone's brain is missing....

Something's missing....

Perhaps something like, "Please note that the test conditions
correspond to a *very* busy Web site. Unless your installation has
similarly heavy traffic, on the order of several hundred thousand
hits a day, the overhead doesn't matter much, and recent versions of
any software available will yield decent performance on a properly
configured machine. Just be sure to keep the files on a disk local to
the server, and stay the hell away from AFS and the goddam
automounter", or something like that? 30 CGI hits a *second* ---
geesh!

[. As a caveat --- the above statement is only actually true if the
only difference between the server API and the CGI implementations
is the CGI startup overhead. There are other things you can do in
some circumstances which might make a difference in performance;
for instance, a search engine function could preload parts of the
database at config time, saving the expense of doing so once per
hit. In those cases, use of an API might indeed be worth it for
performance alone.

In that connection, one of the things that makes the study's
results harder to interpret than they might be is that it says
nothing about what the CGI and NSAPI transactions called upon
the server to do...

Still, giving CGI scripts a performance boost has always struck
me as the least interesting thing about server APIs... ]

MHO, of course, but it's borne out by at least a little experience
running a busy server on a fairly slow machine...

rst

These reports are just bullshit.

Oh, come on, Andrew. While I do think performance concerns are somewhat
overemphasized in the community in general (after all, NCSA 1.3 serves
most of its users just fine, and we all know how slow that was), the
results of the study (that using a server API instead of CGI is worth
nearly a factor of two in performance at saturation loads) are completely
believable.

Of course, they're not the only server out there which *has* an API,
but marketing our server is not their problem. The study was done to
test the performance impact of *one* API, in *one* server, and the
results of that test look perfectly sensible to me. Let's not toss
around invective without due cause.

rst

> Gee, who would have thought a server object-code API would have beat
> CGI? Not me.
>
> http://home.netscape.com/comprod/server_central/performance_whitepaper.html
>
> Something's missing....

You forgot to cross yourself. ;)

These reports are just bullshit.

> Brian

Ay.

On Tue, 14 Nov 1995, Andrew Wilson wrote:
> Hey I'm British, smileys are implicit. ;)

This is the most important thing to remember about this list - maybe it
should go in the FAQ.

Brian

--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
brian@organic.com brian@hyperreal.com http://www.[hyperreal,organic].com/

> most of its users just fine, and we all know how slow that was), the
> results of the study (that using a server API instead of CGI is worth
> nearly a factor of two in performance at saturation loads) are completely
> believable.

Which leads me to believe that perhaps the study would be
helpful in showing that the apache server lends good performance.

>
> Of course, they're not the only server out there which *has* an API,
> but marketing our server is not their problem. The study was done to
> test the performance impact of *one* API, in *one* server, and the
> results of that test look perfectly sensible to me. Let's not toss
> around invective without due cause.
>
> rst
>


--
sameer Voice: 510-601-9777
Community ConneXion FAX: 510-601-9734
The Internet Privacy Provider Dialin: 510-658-6376
http://www.c2.org (or login as "guest") sameer@c2.org

rst:
> These reports are just bullshit.
>
> Oh, come on, Andrew. While I do think performance concerns are somewhat
> overemphasized in the community in general (after all, NCSA 1.3 serves
> most of its users just fine, and we all know how slow that was), the
> results of the study (that using a server API instead of CGI is worth
> nearly a factor of two in performance at saturation loads) are completely
> believable.

I agree.

> Of course, they're not the only server out there which *has* an API,
> but marketing our server is not their problem. The study was done to
> test the performance impact of *one* API, in *one* server

I can read.

> , and the
> results of that test look perfectly sensible to me. Let's not toss
> around invective without due cause.

Let's not. This list should not be used for offering uninformed
critique of the marketing and management methods of other
organisations, which was admittedly my own agenda.

> rst

Hey I'm British, smileys are implicit. ;)

Ay.