Topic: Load times and crontabs (php from console)

I'm running some speed tests on the current alpha of 1.3, just for fun.

And I was wondering if running the php pages from console over and over again would skew the results? (is there any internal cashing for php etc)

Current script (.sh):
php -f index.php
php -f viewtopic.php
php -f viewforum.php

this is repeated 100 times.

Post hoc ergo propter hoc

Re: Load times and crontabs (php from console)

Yes, PHP caches page executions.

3 (edited by Tjalve 2007-10-30 19:16)

Re: Load times and crontabs (php from console)

ok thanks, and I forgot about the query_cache_size in mysql....

So I guess there is no way too get a real load time if you just run the same 3 files over and over ?

Post hoc ergo propter hoc

Re: Load times and crontabs (php from console)

Is this gonna work better?
x= over 10k atleast
Fill _topic with same topic with 1/40 of _posts
Fill _posts with x posts the same 40 and change topic_id every 40 post.
lower query_cache_size a bit (under 8mb)
Run wget and crawl through the site.


Would this work better,  just what too know before I use 2 hr of my time.

Post hoc ergo propter hoc

Re: Load times and crontabs (php from console)

Well, you're pretty much getting real load time running those files, because that caching will always kick in wink

But filling the database would be a good idea, yes.

6 (edited by Tjalve 2007-10-30 21:05)

Re: Load times and crontabs (php from console)

This is the current setup...
Index.php, viewtopic.php (with same $pid) and viewforum.php (with same $id) 100 times (takes about 20sec) (running wget -r -1 -0 url now)
Show_avatar and show_sig is set to 0, so it don't load any off-site images.

Forum Stats
Users: 888
Topics: 4,753
Posts: 36,173

Index lists:
4 cats
22 forums
extra nav links
announcement

Viewforum lists:
One forum with 47pages (1385 topics)

Viewtopic lists:
One post with 40 post (2 pages, some bbcodes and nice blocks of text)

The load time is then inserted into a table via fotter.php.
then a simple "sum() / count(*)" and rounded up too 3 des.

Post hoc ergo propter hoc