Database Caching

Introduction

Caching avoids hitting the DB. Layers: query cache, ORM cache, Redis/Memcached in front, CDN for static. Done well, you cut DB load by 10×.

Understanding the topic

Core concepts to understand:

• Buffer cache — DB-internal hot pages in RAM.
• Materialized views — cached query results.
• Redis/Memcached — app-level cache.
• HTTP/CDN cache — for read-only public data.
• Cache invalidation is the hard part — design upfront.

Syntax reference

Visual workflow / architecture:

Client → CDN → API → Redis → DB
                ▲       ▲       ▲
                └ cold path falls through layers
                ▼ writes invalidate caches

Real-world use

GitHub, Twitter and Reddit all serve > 90% of reads from caches in front of their SQL DBs.

Best practices

• Cache hot reads, never writes.
• Pick a clear invalidation strategy (TTL, event, write-through).
• Monitor hit ratio.

Common mistakes

• Stale caches serving wrong data.
• Thundering herd on cache miss.
• Cache stampede — use stampede protection.

Hands-on exercise

Interview preparation — practice these questions:

• Q1. Cache invalidation strategies.
• Q2. Thundering herd — what is it?
• Q3. Where do you cache: CDN, app, DB?
• Q4. Materialized view as cache.
• Q5. Hit ratio target.