Profiling and Caching

Measure before optimizing: use Xdebug's profiler, SPX, Blackfire or Tideways to see where time goes, and check database query counts. Common wins: fix N+1 queries, add indexes, cache expensive computations in APCu (per-server, in-memory) or Redis (shared), cache HTTP responses (Cache-Control, ETags, reverse proxies), and move slow work to queues.

<?php
// Cache-aside with PSR-16 (simple cache) backed by Redis
function popularProducts(Psr\SimpleCache\CacheInterface $cache, ProductRepository $repo): array
{
    $key = 'products:popular:v1';
    $cached = $cache->get($key);
    if ($cached !== null) {
        return $cached;
    }
    $products = $repo->popular(limit: 20);
    $cache->set($key, $products, ttl: 300 + random_int(0, 60));   // jitter
    return $products;
}

// APCu for per-server config/derived data
$routes = apcu_entry('routes:v42', fn() => buildRouteTable());

Common pitfalls

• Leaving Xdebug enabled in production (large slowdown).
• Caching per-user data under shared keys.
• Forgetting to version or invalidate cache keys after deploys.

Key points

• Profile first
• APCu per server, Redis shared
• Version keys and add TTL jitter

Further reading: PHP manual: APCu