Mailing List Archive

Alternative PHP Cache and benchmarking
Has anyone used Alternative PHP Cache:

http://pecl.php.net/package/APC

with mediawiki?

I just installed it on a wiki and noticed about a 30% improvement
on serving the Main page. Also, what utilites are you using for
benchmarking speed optimizations? I just wrote a simple script
to parse the returned Main page for the "Served by" line and
used the time reported there. Is this legit?
Re: Alternative PHP Cache and benchmarking [ In reply to ]
Jeremy Milum wrote:
> Has anyone used Alternative PHP Cache:
>
> http://pecl.php.net/package/APC
>
> with mediawiki?

We used it for a while on Wikipedia, but gave it up as we had a lot of
problems with it not updating its cache, or generally getting stuffed
up. That was quite a while ago so maybe it's improved since.

Then for a while we used PHP-Accelerator, which is free-as-in-beer but
not open source. It did a pretty good job and was more consistent, but
every once in a while it didn't update quite right. Adding methods to
classes for instance sometimes mysteriously failed to work.
http://www.php-accelerator.co.uk/

A few months ago we switched to Turck MMCache. I haven't noticed any
problems with it, and it's open source:
http://turck-mmcache.sourceforge.net/

I don't have any hard numbers handy, sorry, just offhand anecdotes.

If you want to do whole-page-view benchmarking, you might consider using
the 'ab' benchmarking tool that comes with Apache. You can set it to run
multiple or concurrent runs, and can set custom headers (for instance to
enable gzip compression).

-- brion vibber (brion @ pobox.com)
Re: Alternative PHP Cache and benchmarking [ In reply to ]
On Mon, 28 Jun 2004 22:56:01 -0700, Brion Vibber <brion@pobox.com> wrote:
> If you want to do whole-page-view benchmarking, you might consider using
> the 'ab' benchmarking tool that comes with Apache. You can set it to run
> multiple or concurrent runs, and can set custom headers (for instance to
> enable gzip compression).

Thanks for the info, I will try those other cache programs and see
what works the best. I used the ab benchmark and my results were not
exactly what I expected (fancy that..) The time reported for the
script generation does not always correlate with the data from the ab
program. I guess I have some more testing to do.