Skip to content

ajsalminen/xhprof-kit

 
 

Repository files navigation

About

Note: Using this on a production system is unsupported and potentially dangerous.

Setup

  1. Install XHProf PHP Extension
    • This guide assumes you have the PHP XHProf extension installed already.
  2. Install XHProf-Kit in the Drupal directory you want to benchmark
    • $ /where/is/xhprof-kit/setup.sh
    • This creates a xhprof-kit symlink in the directory.
    • This will also initialize and update the xhprof submodules.
    • Make sure symlinks are allowed in Apache configuration.
  3. Install drush registry rebuild project in drush folder. (optional, only needed for Drupal 7)
    • $ cd ~/.drush
    • $ drush dl registry_rebuild

Assumptions

Currently XHProf-Kit assumes that Drupal is installed at 127.0.0.1. If you use a virtual host, you will need to change the URL in find-min-web.sh.

Benchmarking

Create a scenario

  • Install Drupal and setup a page you want to benchmark.
  • Use devel and devel_generate to create a suitable page.
  • For example: 50 nodes on frontpage

Now lets assume the patch you want to benchmark is in:

  • core--issueno

and the vanilla core is in 8.x branch.

Find a base line

First you will need to find a base line, which is what you compare things to.

Use:

./xhprof-kit/benchmark-branch.sh 8.x

The output will be something like:

loop time: |0.345083s|5173f49dd982a|drupal-perf|8.x|<a href="http://127.0.0.1/xhprof-kit/xhprof/xhprof_html/index.php?run=5173f49dd982a&source=drupal-perf&extra=8.x&url=%2F" target="_blank">Profiler output</a>

The fields delimited by '|' are:

  1. "loop time" - Just a string
  2. loop time - The actual time Drupal needed to load the page.
  3. xhprof-identifier - An identifier for the run that can be used to see the run via xhprof_html/index.php utility.
  4. xhprof-source - A suffix used to distinguish runs - currently set to "drupal-perf".
  5. extra identifier - An identifier used internally by xhprof-kit; usually the branch name.
  6. Run URL - The url, where the XHProf output can be seen.

What you need as your baseline is the xhprof identifier: '5173f49dd982a'.

Note that as scenarios and core changes, so will your baseline, so be sure to always re-calculate it when starting a different benchmarking scenario.

Compare Branches

To now benchmark your baseline against 8.x again (to verify it is accurate) and against your new branch with the patch to test, use the following command:

./xhprof-kit/benchmark-branches.sh <your baseline identifier> baseline-8.x 8.x core--issueno

The output of this will be something like:

=== baseline-8.x..8.x compared (5173f49dd982a..5172adc3e447f):

ct  : 47,077|47,077|0|0.0%
wt  : 396,923|397,915|992|0.2%
cpu : 370,881|370,295|-586|-0.2%
mu  : 30,347,736|30,349,184|1,448|0.0%
pmu : 30,472,168|30,475,144|2,976|0.0%

=== baseline-8.x..core--issueno compared (5172ad6a9a6b6..5172ae167ba6d):

ct  : 47,077|47,207|130|0.3%
wt  : 396,923|399,899|2,976|0.7%
cpu : 370,881|373,186|2,305|0.6%
mu  : 30,347,736|30,440,776|93,040|0.3%
pmu : 30,472,168|30,564,656|92,488|0.3%

---
ct = function calls, wt = wall time, cpu = cpu time used, mu = memory usage, pmu = peak memory usage

Hints

  • Update 100 to 1000 in benchmark-branch.sh to have more accurate runs.
    • 1000 takes a while, but is very accurate. It depends on the speed of your machine and how dedicated it can be, if 100 runs are enough to find the minimum.
  • 'baseline-8.x' is just a name, its a good idea to point out what the benchmark is here, like 'core-stark-10-nodes'.

Benchmarks are saved in xhprof-kit/data/[your baseline identifier]-[baseline name].data files.

  • To later show a benchmark again use:
    • ./xhprof-kit/show-bench.sh [your baseline identifier] baseline-8.x

Uploading

As XHProf runs are really useful when published and others can take a look, the easiest is to upload them to a hosted installation of XHProf.

./xhprof-kit/upload-bench.sh <your API key> <your xhprof source ID> <your baseline identifier> baseline-8.x

You can get an API key and source ID by contacting me via my drupal.org contact form (http://drupal.org/user/693738) or Fabianx on FreeNode.

Due to security reasons I only give out API keys to people I know.

Hosting yourself

However you can also host runs yourself by changing just some little files:

  • Download xhprof to / of your server.
  • Install the xhprof submodule like usual.
  • Change all lionsad.de domains to your server
  • Add a hosted/api.php file like:
<?php

$api_keys = array(
  '8da2a3349ea4711eb59f57faff1eb05a' => 'drupal-perf-cottser',
);

The format is: API KEY => source identifier, which is used as xhprof source argument.

  • Apply the following diff to xhprof:
diff --git a/xhprof_lib/utils/xhprof_runs.php b/xhprof_lib/utils/xhprof_runs.php
index cde5ff5..561d410 100644
--- a/xhprof_lib/utils/xhprof_runs.php
+++ b/xhprof_lib/utils/xhprof_runs.php
@@ -90,6 +90,9 @@ class XHProfRuns_Default implements iXHProfRuns {
     // we use the xhprof.output_dir ini setting
     // if specified, else we default to the directory
     // in which the error_log file resides.
+    if (empty($dir)) {
+      $dir = '../../data/stored-runs/';
+    }
 
     if (empty($dir)) {
       $dir = ini_get("xhprof.output_dir");

About

Fabian's XHProf Benchmarking Kit

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Shell 50.5%
  • PHP 49.5%