redis-deep

Redis Deep Dive

When to Use This Skill

• Implementing distributed locking (Redlock)
• Building rate limiters with sliding windows
• Designing pub/sub for real-time features
• Using Redis Streams for event log
• Choosing memory eviction policy

Core Principles

1. Redis is not a database for data you cannot afford to lose — AOF/RDB persistence is best-effort; design for Redis to be empty
2. One connection per operation is wasteful — use pipelining for batch operations (10× throughput improvement)
3. Key naming convention is critical — resource:id:field prevents collisions and enables pattern scanning
4. TTL everything — unbounded keys accumulate until OOM; if no TTL makes sense, use a large one (7 days)
5. Redlock requires majority quorum — N/2+1 nodes must confirm; never use single-node lock for distributed coordination

References available

• references/data-structures.md — 8 data structures with TypeScript examples, Redlock, sliding window rate limiter, pub/sub
• references/caching-patterns.md — eviction policies, pipeline batching, keyspace notifications, cache-aside pattern
• references/cluster-sentinel.md — Redis Cluster vs Sentinel, topology decisions, failover behavior