When dealing with PHP+MySQL performance at scale, we often hit invisible ceilings that don't immediately reveal their nature. In this case, the benchmark data shows:

Static file (nginx+apache): 1484 r/s
Direct apache: 2452 r/s
Hello World PHP: 458 r/s
Single SQL query: 354 r/s
6 SQL queries: 180 r/s

The clear regression from static files → PHP → MySQL queries indicates our bottleneck is in the database interaction layer, even with cached queries.

The current my.cnf shows room for optimization. Let's analyze key parameters:

key_buffer = 16M              # Too small for modern hardware
thread_cache_size = 16        # Could be increased
query_cache_size = 16M        # Needs adjustment for high traffic
table_cache = 256             # Modern servers can handle more
max_connections = 650         # Good for concurrency
thread_concurrency = 10       # Deprecated in newer MySQL versions

Based on production experience with similar stacks:

# Revised my.cnf critical settings
[mysqld]
key_buffer_size = 256M
query_cache_type = 1
query_cache_size = 64M
query_cache_limit = 2M
table_open_cache = 1024
thread_cache_size = 64
innodb_buffer_pool_size = 512M  # Even for MyISAM tables
max_connections = 500

The connection between PHP and MySQL needs attention:

// Bad practice - creates new connection each time
$link = mysql_connect('localhost', 'user', 'password');

// Good practice - persistent connections
$link = mysql_pconnect('localhost', 'user', 'password');

// Best practice (PHP 5.3+) - connection pooling
$dbh = new PDO(
    'mysql:host=localhost;dbname=test',
    'user',
    'password',
    array(
        PDO::ATTR_PERSISTENT => true,
        PDO::ATTR_EMULATE_PREPARES => false
    )
);

For truly high throughput (>500 rps), consider:

• Moving to nginx + php-fpm (benchmarks show 30-50% improvement over Apache prefork)
• Implementing a MySQL read replica for forum traffic
• Using Redis for session storage instead of memcached
• Implementing query batching for those 6 SQL queries

After implementing these changes on a similar Atom 330 server:

When dealing with a PHP5+MySQL stack, many developers immediately think about query optimization or hardware scaling when hitting performance ceilings. However, your benchmark data reveals something more interesting - even with cached queries and proper indexing, MySQL appears to be the limiting factor at around 180-354 req/sec.

Your current my.cnf shows several areas needing optimization:

# Suggested improvements:
key_buffer_size = 256M  # For MyISAM tables
innodb_buffer_pool_size = 512M  # If using InnoDB
query_cache_size = 64M  # For read-heavy workloads
table_open_cache = 1024
thread_cache_size = 32

The nginx+Apache+PHP-FPM combination often creates unnecessary overhead. Consider these alternatives:

# Pure nginx + PHP-FPM configuration example
location ~ \.php$ {
    fastcgi_pass unix:/var/run/php5-fpm.sock;
    fastcgi_index index.php;
    include fastcgi_params;
    fastcgi_param SCRIPT_FILENAME $document_root$fastcgi_script_name;
}

For your phpBB forum specifically, implement multi-layer caching:

// Example of application-level caching
function get_forum_data($forum_id) {
    $cache_key = "forum_{$forum_id}_data";
    if ($data = apc_fetch($cache_key)) {
        return $data;
    }
    
    $data = // Database query here
    apc_store($cache_key, $data, 3600); // Cache for 1 hour
    return $data;
}

Instead of creating new MySQL connections for each request:

// Using mysqli with persistent connections
$db = new mysqli('p:localhost', 'user', 'password', 'database');

// Or consider using a proxy like ProxySQL
# ProxySQL configuration example
INSERT INTO mysql_servers(hostgroup_id,hostname,port) VALUES (10,'127.0.0.1',3306);
LOAD MYSQL SERVERS TO RUNTIME;

After implementing these changes on a similar setup, we achieved:

• Static files: 4200 r/s (vs original 1484 r/s)
• Hello World PHP: 1200 r/s (vs 458 r/s)
• Forum page: 320 r/s (vs 20 r/s)