Performance & Reliability

"The site got slow" almost always has a findable cause: a query that stopped using an index, a cache that silently quit, a worker pool sized for a different server. We find it by measurement, fix it, and leave you with the instruments to see the next one coming.

What this covers

• Query and index tuning — slow-query-log analysis, EXPLAIN-driven index work, and rewriting the small number of queries that dominate the digest. One index frequently outperforms a caching project.
• Caching that pays for itself — OPcache sized and verified, Redis for sessions and hot fragments with deliberate TTLs, HTTP-layer compression and cache headers. Each layer with an invalidation story you can explain in a sentence.
• Apache & PHP-FPM sizing — worker pools computed from measured per-worker memory on your application, timeouts aligned so failures are clean, mod_php estates moved to PHP-FPM behind mpm_event.
• Monitoring — the small set of numbers worth watching (p95 latency, FPM listen queue, slow-query rate, free memory) wired to alerts that fire before users notice.
• Backups that restore — backup coverage for database and files, restore drills with measured recovery times, point-in-time recovery via binary logs where the business needs it.

How we work

Baseline first: no change ships without a number it is supposed to move. Fixes land as small, reviewable changes on your infrastructure, not as a mystery box. And we will tell you when the honest answer is "this server is simply too small" or "this is fine — spend the money elsewhere."

Hourly time and materials, as a bounded tuning engagement or ongoing support. Describe the symptom and we will tell you where we would look first.