This is a graphical front end designed to store and present the profiling information provided by the Facebook created XHProf profiling tool.
Related Tools:
- XHGui - Uses MongoDB as a backend (rather than MySQL)
- XHProf.io - Uses MySQL backend, more normalized storage schema
This branch/clone/whatever git calls it of the official Facebook GUI does a few things:
- It includes a header.php document you can use with PHP's auto_prepend_file directive. It sets up profiling by initilizing a few variables, and settting register_shutdown_function with the footer. Once started profiles are done when requested (?_profile=1), or randomly. Profiled pages display a link to their profile results at the bottom of the page (this can be disabled on a blacklist based for specific documents. e.g. pages generating XML, images, etc.).
- For tips on including header.php on an nginx + php-fpm install take a look at: http://www.justincarmony.com/blog/2012/04/23/php-fpm-nginx-php_value-and-multiple-values/
- The GUI is a bit prettier (Thanks to Graham Slater)
- It uses a MySQL backend, the database schema is stored in xhprof_runs.php
- There's a frontend to view different runs, compare runs to the same url, etc.
Key features include:
- Listing 25, 50 most recent runs
- Display most expensive (cpu), longest running, or highest memory usage runs for the day
- It introduces the concept of "Similar" URLs. Consider:
- http://news.example.com/?story=23
- http://news.example.com/?story=25 While the URLs are different, the PHP code execution path is likely identical, by tweaking the method in xhprof_runs.php you can help the frontend be aware that these urls are identical.
- Highcharts is used to graph stats over requests for an easy heads up display.
Requirements:
- Zlib library in PHP: http://php.net/manual/en/zlib.installation.php (Used to compress serialized data)
- Database backend
- MySQL, or MySQLi for data storage fully supported
- Support for SQL Server now in beta stages #jumpInCamp
Work that we're still doing:
- The aggregation functionality is ignored completely
- The code is... a mess. Deadlines do that to you, we're working on it
- The default table schema isn't indexed all the places it needs to be
- Easier ways to diff URLs