Benchmark: Reading Directory Size

[apinto]

Hello,
I was having trouble with the calculation of disk space taking to long (around 5min), so I decided to check what was going on.


First I checked the Daemon script and found the class it uses to get the disk space, it is located here:

Code:

/etc/sentora/panel/dryden/fs/filehandler.class.php

Readin the file you go to the GetDirectorySize class and you can read:

PHP Code:

   /**
     * Returns the full size of a directory and all child objects.
     * @author Bobby Allen (ballen@bobbyallen.me)
     * @param string $directory The filesystem path to the directory
     * @return int The directory size in bytes.
     */
    static function GetDirectorySize($directory) {

        error_reporting(0);

        $size = 0;
        if (substr($directory, -1) == '/') {
            $directory = substr($directory, 0, -1);
        }
        if (!file_exists($directory) || !is_dir($directory) || !is_readable($directory)) {
            return -1;
        }
        $handle = opendir($directory);
        if ($handle) {
            while (($file = readdir($handle)) !== false) {
                $path = $directory . '/' . $file;
                if ($file != '.' && $file != '..') {
                    if (is_file($path)) {
                        $size += filesize($path);
                    } elseif (is_dir($path)) {
                        $handlesize = self::GetDirectorySize($path);
                        if ($handlesize >= 0) {
                            $size += $handlesize;
                        } else {
                            return -1;
                        }
                    }
                }
            }
            closedir($handle);
            return $size;
        } else {
            return false;
        }
    } 


Its everything ok with this, it works perfectly, however I decided to benchmark ( using microtime() ) and got the following results (I do have HUGE directories...)

Code:

START Calculating disk Usage for all client accounts..
Disk usage for user "zadmin" is: 177763599 (173.6 MB) in: 54.743 seconds
Disk usage for user "user2" is: 135473500 (132.3 MB) in: 72.311 seconds
Disk usage for user "user3" is: 393436099 (384.2 MB) in: 90.218 seconds
END Calculating disk usage in: 217.274 seconds

I was amazed, 217.274 seconds   that would explain the 5 min wait time for each daemon to run.

I decided to make a quick change, as Sentora is supporting Linux only (officially), why not use the du command?
So I created a new class:

PHP Code:

static function GetDirectorySizeDU($directory) {

        error_reporting(0);

        $size = 0;

        if (substr($directory, -1) == '/') {
            $directory = substr($directory, 0, -1);
        }
        if (!file_exists($directory) || !is_dir($directory) || !is_readable($directory)) {
            return -1;
        }
        $handle = opendir($directory);
        if ($handle) {

        @exec("du -c -b ".$directory."/".$k, $output);
            $use = explode("\t", array_pop($output));
            @$size += $use[0];

            return $size;
        } else {
            return false;
        }
    } 


Probably can be improved, however I did not spent too much time looking into it, but, lets see the results:

Code:

START Calculating disk Usage for all client accounts..
Disk usage for user "zadmin" is: 185160975 (180.8 MB) in:12.438 seconds
Disk usage for user "user2" is: 142506332 (139.2 MB) in:11.265 seconds
Disk usage for user "user3" is: 404528067 (395 MB) in:15.667 seconds
END Calculating disk usage in:39.372 seconds

WOW!! 
From 217 seconds to 39 seconds!!
This is a 456% better perfomance!!
Any thoughts one why this might not work?
I mean, this is just great