Backend & Database Performance — Singleflight, Redis & High-Concurrency APIs

Measure before optimizing. Performance work without empirical profiling is guessing — and guessing leads to premature optimization that adds architectural complexity without moving latency metrics.

This skill governs 100% backend, database, server runtime, and API performance. It equips AI agents and backend engineers with rigorous patterns to minimize Time to First Byte (TTFB), eliminate database bottlenecks, protect Node.js/Bun event loops, and design resilient caching architectures.

• API response time (TTFB / p95 / p99) violates SLA budgets (>200ms).
• PostgreSQL queries suffer from sequential scans, buffer misses, or lock contention.
• N+1 query cascades detected in ORM data access (Prisma, TypeORM, Drizzle, Kysely).
• High throughput causes cache stampedes / thundering herds against backing stores.
• Heavy CPU operations or synchronous I/O freeze the Node.js / Bun single-threaded event loop.
• Database connection pool exhaustion or transaction timeouts occur under load.

When NOT to use:

• Frontend DOM, Core Web Vitals, CSS, or browser asset bundling — use frontend-performance-lighthouse.