Super Mario errr Marco….

Der lustige kleine süsse Krakeeler am Ende des Comics ist Marco aka bate

Guckst du hier .

Ich bin begeistert - Marco ist richtig gut getroffen…

August 22, 2007 | Filed Under _trunk | 4 Comments 

Comparing communication speed between different machines in PHP

For a project we needed some sort of inter-server communication. Due to performance critical tasks like Click Tracking we decided that we need to investigate first the most reliable and fastest way to communicate.

1. Using Sockets

I created a Test Socket Server using more or less the demo script under http://de.php.net/sockets . This Server just takes what he receives and send it back. What is not implemented is multi threading and optimization - but it should be enough to get some reasonable results.

The Test Server looks like this…

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39

do {
 
 if (FALSE === ($buf = socket_read ($msgsock, 2048))) {
 
 	echo "socket_read() fehlgeschlagen: Grund: ";
 
 	echo socket_strerror ($ret) . "n";
 
 	break 2;
 
 }	if (!$buf = trim ($buf)) {
 
 	continue;
 
 }
 
if ($buf == 'quit') {
 
 	break;
 
 }
 
if ($buf == 'shutdown') {
 
 	socket_close ($msgsock);
 
 	break 2;
 
 }
 
$buf .= "n";
 
 socket_write ($msgsock, $buf, strlen ($buf));
 
} while (true);
 
 socket_close ($msgsock);
 
} while (true);

Now we need to create some client. In our case we keep the connection open so we could avoid opening and closing sockets. The client for testing looks like this:

1
2
3
4
5

$sh = socket_create(AF_INET, SOCK_STREAM, SOL_TCP);
 
if (! socket_connect( $sh, '192.168.20.33', 10000) )
 
die( 'Could not bind to Socket!'.PHP_EOL);
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

$dStart = microtime( true);
 
for( $iIdx = 0; $iIdx < $iMaxCyles; $iIdx++) {
 
 $aClient[$iIdx] = $iIdx;
 
 $aData = array( 'iIdx' => $iIdx );	// call
 
 $aRes = Comm_Socket( $sh, $aData);
 
 $sResult = $aRes['iIdx'];
 
$aServer[$iIdx] = $sResult;
 
}
 
$dDuration = microtime( true) - $dStart;
 
echo( $iMaxCyles." calls total: $dDuration sec. ".PHP_EOL);
 
echo( "Average:  ".$dDuration/$iMaxCyles." sec. per call ".PHP_EOL.PHP_EOL);
 
$aLost = array_diff( $aClient, $aServer);
 
if( count( $aLost)) {
 
 echo( "lost call/differences:".PHP_EOL);
 
 print_r( array_diff( $aClient, $aServer));
 
}

where as function Comm_Socket looks quite simple:

1
2
3
4
5
6
7
8
9
10
11

function Comm_Socket( $Socket, $mPostData)
 
{
 
 $sMessage = serialize( $mPostData);	socket_write( $Socket, $sMessage, strlen( $sMessage));
 
 $sRet = socket_read( $Socket,512,PHP_NORMAL_READ);
 
return unserialize( $sRet);
 
}

The results look like this:

1000 calls total: 0.473767995834 sec.
Average: 0.000473767995834 sec. per call

2. Using http via file_get_contents

Now let’s modify our testscript a bit and change our server so, that he takes any POST parameter and echo’s it.
This is easy to do - and our Client Function looks now like this:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37

function Comm_File_Get_Contents( $sUri, $mPostData)
 
{
 
 $fTimeout = 20;	$sHttpPostQuery =  http_build_query ( $mPostData );
 
$aStreamContextOpts = array(
 
 'http'=>array(
 
 'method'  => 'POST',
 
 'header'  => 'Content-type: application/x-www-form-urlencoded',
 
 'content' => $sHttpPostQuery
 
 )
 
);
 
$oContext = stream_context_create ( $aStreamContextOpts );
 
// inject a request read timeout
 
$aAddOpts = array(
 
'http' => array('timeout' => (float) $fTimeout )
 
);
 
stream_context_set_option( $oContext, $aAddOpts );
 
$sResponse = file_get_contents ( urldecode($sUri), false, $oContext );
 
return $sResponse;
 
}

Ok the results are quite interesting:

1000 calls total: 4.02230811119 sec.
Average: 0.00402230811119 sec. per call

Nearly 10 times slower compared to our Socket communication. For sure - by using lighttpd and tweaking the system we could improve performance - but still it’s a lot slower.

Also interesting are the results when we call the server not by IP but by URI - then the results look a lot worse:

1000 calls total: 156.423214912 sec.
Average: 0.156423214912 sec. per call

This is probably because our nslookup is quite slow ( we run Microsoft Small Business Server ) but still - the differences are dramatically.

Actually I could sum up now - but for the records I will try also the cURL way.

3. Using http via cURL

For using cURL we have to modify our client function a bit - the server can be left untouched.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

function Comm_cUrl( $sUri, $mPostData)
 
{
 
$mOptions = array( 'host' => $sUri,
 
'port' => 80,
 
'method' => 'POST',
 
'timeout' => 3,
 
'params' => $mPostData);$ch = curl_init();
 
curl_setopt( $ch, CURLOPT_URL, $sUri );
 
curl_setopt( $ch, CURLOPT_HEADER, 0 );
 
curl_setopt( $ch, CURLOPT_USERAGENT, @$_SERVER['HTTP_USER_AGENT'] );
 
curl_setopt( $ch, CURLOPT_FOLLOWLOCATION, 1 );
 
curl_setopt( $ch, CURLOPT_RETURNTRANSFER, 1 );
 
curl_setopt( $ch, CURLOPT_SSL_VERIFYPEER, 0 );
 
if (  'POST' == $mOptions['method'] ) {
 
curl_setopt( $ch, CURLOPT_POST, 1 );
 
$sPostString = "";
 
foreach( $mOptions['params'] as $key => $value ) {
 
$sPostString .= "$key=" . utf8_encode( $value ) . "&";
 
}
 
$sPostString = substr( $sPostString, 0, -1 );
 
curl_setopt( $ch, CURLOPT_POSTFIELDS, $sPostString );
 
}
 
$mData = curl_exec( $ch );
 
curl_close( $ch );
 
return $mData;
 
}

Using IP we get these results:

1000 calls total: 3.17922592163 sec.
Average: 0.00317922592163 sec. per call

Interesting - a bit faster then with file_get_contents.

Finally let’s also try an URI and we get:

1000 calls total: 4.69447493553 sec.
Average: 0.00469447493553 sec. per call

So - cURL obivously caches the nslookup what leads to much better results.

4. Conclusion

More or less we got what we expected. Taking into account that multi threaded socket communication is a real beast and can cause a lot of trouble my personal opinion is that http will work fine for us. There is a lot of potential of speeding things up - so the difference is tolerable for us. This might be different for you ;) 

Results:
Time    Avg      Typeof
0.47    0.00047	 Socket
4.02    0.00402	 file_get_contents
156.42  0.15642  file_get_contents by uri
3.17    0.00317  curl
4.69    0.00469  curl by uri
August 21, 2007 | Filed Under PHP related | 2 Comments 

Internet Hasen ?

Da können Internet und Telefon ja bald einpacken - ich kommuniziere ab sofort nur noch durch Ohren ziehen und drehen….

Nabaztag ist was, was wie ein Hase aussehen soll…

Und es kommt noch besser: Man kann auch von Hase zu Hase kommunizieren. Freunde können Nachrichten direkt an den eigenen Nabaztag schicken, die er einem dann vorliest. Andere, befreundete Hasen können direkt angegangen werden, indem man die Ohren des eigenen Nabaztags bewegt.

Vielleicht gibt es dann bald auch eine Adult-Version bei denen man auch an anderen Körperstellen ziehen und drehen kann…

via public-beta

August 20, 2007 | Filed Under _trunk | 1 Comment 

Neulich in Leipzig…

Real Life… eine Anleitung:

Du sollst/nicht:

Sollst ? Oder nicht ? :) Seltsam ? So stehts geschrieben - guckst du hier:

Leipzig….

August 17, 2007 | Filed Under _trunk | 2 Comments 

GEADN-Systematik zur Qualitätssicherung ?

Heute im IRC:

johann__: bei uns setzen wir uebrigens die GEADN-systematik zur qualitaetssicherung bei commits ein
johann__: dh. “Gepflegt eins auf die Nase”

johann__: ja, man darf da den motivationsschub durch gezielt eingesetzte koerperliche gewalt nicht unterschaetzen. Schon nach wenigen Rippenbruechen waechst das Bewusstsein fuer die notwendigkeit korrekt eingerueckter klammern erheblich.

johann__: auch muessige diskussionen bei architektonischen entscheidungen lassen sich mit gewalt auf eine sportliche zeit begrenzen.

johann__: ich mein, vor der deadline muss es ja nicht immer nur psychologischer druck sein, da ist gut dosierte pruegel eine viel ehrlichere und direktere art der teamkommunikation.

…

August 16, 2007 | Filed Under _trunk | 6 Comments 

Failed asserting that “double:0.999″ matches expected value “double:0.999″.

If you execute the following piece of code in a test using phpUnit….

$dRes = 400-399.001;
$this->assertEquals( 0.999, $dRes);

you will get…

“Failed asserting that <double:0.999> matches expected value <double:0.999>.”

That’s PHP…

The correct way would be “$this->assertEquals(0.999, round($dRes, 3));” as float handling in PHP is a bit strange. But I really love this Message - I start to think about what is reality and what not…

August 15, 2007 | Filed Under PHP related, _trunk | 5 Comments 

Das weckt Vertrauen….

bild-1.png

Ob ich da mal klicken soll ? Wusste gar nicht das mein Powerbook ein PC ist… der optimiert mir dann bestimmt die win.ini….

August 15, 2007 | Filed Under _trunk | Leave a Comment 

Facebook sourcecode leaked…

I know you might have read this somewhere else already but still - I like it - It’s sort of fun….

http://facebooksecrets.blogspot.com/

I like especially this part:

// Holy shit, is this the cleanest fucking frontend file you’ve ever seen?!
ubersearch($_GET, $embedded=false, $template=true);

August 14, 2007 | Filed Under _trunk | Leave a Comment 

linux developer…

Over the last years I could observe some strange attitudes in my various teams. Many developer love to work under linux - but in fact they often lack the skills to administrate the developer machine correctly.

Last week I had another incident: I could observe that the resolution of the laptop display did not fit. So I asked him why he doesn’t set the correct resolution - but after some tries in the Suse GUI I got the reply: “I can’t - but anyway I like it like it is otherwise everthing would be sooo small…”

I had a deeper look and realized that the external 21′ LCD which we provide was not used either - for similar reasons - well, what should I say ? Now I do have a really nice second display on my laptop…

August 12, 2007 | Filed Under _trunk | Leave a Comment 

Ruby is…

One of my Developers today about Ruby on Rails:

Ruby is somehow crazy, well if you use it a lot you might get used to it but still …. it’s a totally crazy language

No comments…

August 10, 2007 | Filed Under PHP related, _trunk | Leave a Comment 

Next Page →