Bug 1026968

Description of problem:

When making calls to a mysql gear (using php, as an example) the calls or connections to provide adequate performance when threaded. 

How reproducible:
PHP SCRIPT: 

<?php

// Create connection
$dns = '127.10.1.2';
$port = '3306';
$user = 'admin8gNAWKu';
$pass = 'FqWI5TxtkYw8';
$db = 'mysql';

echo "DNS='$dns', PORT='$port', USER='$user', PASS='$pass', DB='$db' \n";
echo "Testing connection...";

$start = microtime(true);

$mysqli = new mysqli($dns, $user, $pass, $db, $port);

// $con=mysqli_connect($dns,$user,$pass,$db,$port);
// Check connection
if ($mysqli->connect_errno)
  {
  echo "Failed to connect to MySQL: " . mysqli_connect_error();
  } else {
  echo "Connected!";
  }

echo $mysqli->host_info . "\n";
$result = $mysqli->query("SELECT COUNT(*) FROM user");

while ($row = $result->fetch_assoc()) {
    echo " id = " . $row["COUNT(*)"] . "\n";
}

$time_taken = microtime(true) - $start;
echo $time_taken;

?>

Steps to Reproduce:
1. rhc app create bench php-5.3 mysql-5.1
2. Copy code (script above) into gear 
   - I ssh in, create a file in app-root/data/ and copy its contents into that file. 
3. Average the time against sequential calls vs threaded calls.
   - Sequential:
for x in $(seq 1 2); do php -f test.php; done | grep TIME | awk '{ total += $2; count++ } END { print total/count }'
   - Threaded:
for x in $(seq 1 2); do php -f test.php & done | grep TIME | awk '{ total += $2; count++ } END { print total/count }'

You will see that by treading the calls (or backgrounding each test) they queue up and cause performance issues to be noticed. This is likely caused by queuing of the database connections. Are there any best practices for avoiding this, or things that can be done to improve the databases performance in threaded scenarios.