pgsql TIME: 1.2292530536652
adodb TIME: 3.7784049510956
pear TIME: 6.3551380634308
pdo TIME: 1.2426941394806

from selecting from 1521 rows over 250times from the DB in my laptop.

Surely that ODBC will be one of the slowest but I really don’t know for dbx !!!
I think that as it is based on vendors drivers it should be slower than using it directly, but will it be faster than ODBC, DB, & MDB2 ???

‘mysql’,
‘hostspec’ => ‘localhost’,
‘database’ => ‘mysql’,
‘username’ => ‘root’,
‘password’ => ‘pass’
);

if ($type == “DB”)
{
require_once(“DB.php”);

$db = DB::connect($dsn);
if (PEAR::isError($db)) { die($db->getMessage()); }
$db->setFetchMode(DB_FETCHMODE_ASSOC);

for ($i=1; $iquery( $sql);

if (PEAR::isError($res)) {die($res->getMessage());}

echo “”;
while( $res->fetchInto( $row ) )
{
process_row($row);
}
echo “”;
}
}
elseif($type == “MDB2″)
{
require_once(“MDB2.php”);

$mdb2 =& MDB2::connect($dsn);
if (PEAR::isError($mdb2)) { die($mdb2->getMessage()); }

for ($i=1; $iquery( $sql);

if (PEAR::isError($res)) {die($res->getMessage());}

echo “”;
while ($row = $res->fetchRow(MDB2_FETCHMODE_ASSOC))
{
process_row($row);
}
echo “”;
}
}
elseif ($type == “PDO”)
{
$pdo = new PDO(“mysql:dbname=”.$dsn['database'], $dsn['username'], $dsn['password']); //PDO::FETCH_ASSOC

for ($i=1; $i<=$num; $i++)
{
echo “”;
foreach ($pdo->query($sql) as $row)
{
process_row($row);
}
echo “”;
}
}
elseif ($type == “mysql”)
{
mysql_connect($dsn['hostspec'], $dsn['username'], $dsn['password']);
mysql_select_db($dsn['database']);

for ($i=1; $i<=$num; $i++)
{
echo “”;
if ($res = mysql_query($sql))
{
while($row = mysql_fetch_assoc($res))
{
process_row($row);
}
}
echo “”;
}
}
}

function process_row($row)
{
ob_start();
echo sizeof($row);
//print_r();
ob_end_clean();
}

$types = array(“PDO”, “DB”, “MDB2″, “mysql”);
$sql = ‘SELECT * FROM user’;

require($_SERVER['DOCUMENT_ROOT'].”/inc/class/class_timer.php”);
$timer = Timer::getInstance();
$timer->saveTime(’start’);
foreach ($types as $type)
{
dump_users($type, $sql, 500);
$timer->saveTime(‘dumped ‘.$type.’ driver’);
}

echo $timer->getReport(1);

?>

PDO, 2500 iterations: 2.94587302208
pg_query, 2500 iterations: 2.72805809975

So they’ve improved PDO in the last year or two.

It’s ~8% slower than direct pg_* functions, at least for relatively simple selects, as opposed to the ~16% when this post was originally written.

I left out ADOdb with the extension because, as I stated, my tests were to determine if there is it was an advantage or a disadvantage to add a large distribution such as the ADOdb library. Also, it is my feeling that PDO is widely adopted (and will become even more so) by web hosts than the ADOdb extension.

As to why I didn’t benchmark the mysql/mysqli functions – I was benchmarking database abstraction libraries. Of course they should perform faster than any abstraction library but that is beside the point if abstraction is required, right?

Are you suggesting that using any of these libraries are less suitable for use in a scalable application?

If you suspect that they would be nearly a tie, why leave it out? And if you are worried about performance, why even consider a host where you can’t use something like APC or the ADOdb extension? Usually this means a dedicated server, something where you have a lot more control.

You tests showed ADOdb with APC improving by more than 700%, on the low end. A good chunk of that is because of APC, since it doesn’t have to re-compile the PHP scripts every time. Which goes back to my point, using something APC offsets library size in a big way. Not entirely, but much better than without it.

It seems obvious on the surface that ADOdb is probably not going to be as fast as PDO, I don’t think anyone will find that surprising.

If you were looking to demonstrate performance issues, why didn’t you include benchmarks for the mysql/mysqli functions? Wouldn’t those be even faster than PDO?

Just a note on “performance” in general. It isn’t a substitute for scalability. Not that improved performance isn’t an important part of being scalable, but no amount of increased performance will make up for scalability.

I agree that the ADOdb extension would change things. I suspect that PDO and ADOdb performance would be nearly a tie.

I do have to disagree that the overhead to import code is small – as my tests showed, even with APC enabled, ADODdb performed at a bit less than half the speed of ADOdb. And since many hosting environments offer neither APC or the ADOdb extension, ADOdb is going to use a great deal more resources in most cases… so it should be a concern where performance is an issue.

Ultimately, my tests were to determine if there is it was an advantage or a disadvantage to add a large distribution such as the ADOdb library, and I understand that your tests were targeted differently.

True, it does not include the over head required to import the library code. For those worried about performance that over head is likely to be very, very, very small since they’ll be using some sort of cache like APC.

You technique also includes many other factors that I wasn’t interested in measuring, my only concern was the relative performance of talking do the database in different ways.

I read through the results of your tests, I didn’t see anything using the ADOdb extension, which speeds things up quite a bit.

Thanks for the benchmark. However, it does not take into consideration the real world aspect of library inclusion overhead.

See: http://www.tonylandis.com/articles/accurate-web-application-benchmarking-methodology.htm