Databases — Modeled and Tuned

Most performance problems I meet are database problems wearing a frontend costume. I design schemas deliberately, measure before optimizing, and pick engines by workload instead of fashion.

Engine choices I actually make

Engine	When I choose it
PostgreSQL	Default OLTP: panels, APIs, anything with relations and constraints.
MySQL / MariaDB	Shared-hosting realities and legacy stacks.
MongoDB	Document-shaped, rapidly evolving data — session vaults, logs.
Redis	Caching, queues, rate limiting, session state.
Firebase / Supabase	Zero-ops products and prototypes that still need auth + data.

• Schema design — normalization where integrity matters, deliberate denormalization where reads dominate.
• Query tuning — EXPLAIN plans, index strategy, N+1 hunting.
• Migrations — versioned, reversible, rehearsed on staging.
• Backups — automated, tested, and actually restorable (the only kind that count).