From: Paul <pa...@qu...> - 2005-10-10 22:36:14
|
As a follow up to my previous email, I do wonder how sane the attached patch might be to apply. In terms of php performance, xdebug output goes down by about 4-5 times. What I've not tested is whether the attached breaks anything. I think it's reasonably sane but the functions that I've hooked are used in multiple places throughout the code. Schnitzel: are you able to see if the attached makes any difference to you, or are you running 1.0rc2 in production? (I certainly wouldn't run the patch in production as it's had ~0 testing) Paul -----Original Message----- From: man...@li... [mailto:man...@li...] On Behalf Of Paul Sent: 10 October 2005 23:09 To: man...@li... Subject: RE: [Mantisbt-dev] Fwd: Performance problem with mantis 1.0.0rc2 >>>> From looking at top, the cpu hog is Apache, not mysql. Out of interest, and the amount of testing/coding this has got is about 30 seconds in total - i.e. I wouldn't be 100% sure whether this does actually work but... If you make the following change to authentication api, does it make any difference to performance? authentication_api: 54a55,57 > global $g_cookie_test; > if($g_cookie_test) > return true; 417c420 < global $g_cache_current_user_id; --- > global $g_cache_current_user_id, $g_cookie_test; 445c448,452 < return ( 1 == db_num_rows( $result ) ); --- > $g_cookie_test = false; > if( 1 == db_num_rows( $result ) ) { > $g_cookie_test = true; > return ( true ); > } >>> For a query that takes me 0.01 seconds using the mysql console, it >>> takes apache/mantis 0.36 seconds: How are you calculating this? What adodb database type are you using? From some php debugging tools, for 91 queries for me to run on the view all bugs page I'd be inclined to say it takes a few hundred milli-seconds. If it's apache being slow, I'd be somewhat included to think it's due to some of the generic functionality in mantis e.g. the above code change for me almost halved the number of internal php function calls when generating a page. For the most part, the biggest performance gains will come from any database performance optimisation, but as the code base has become more generic, it might be worth adding some hooks to cut down on the amount of processing work the webserver has to do. Paul ------------------------------------------------------- This SF.Net email is sponsored by: Power Architecture Resource Center: Free content, downloads, discussions, and more. http://solutions.newsforge.com/ibmarch.tmpl _______________________________________________ Mantisbt-dev mailing list Man...@li... https://lists.sourceforge.net/lists/listinfo/mantisbt-dev |