Can you recommend a cache store for rate limiting in a high-throughput API?

Yes — for high-throughput API rate limiting, the best cache store is usually Redis.

Why Redis is a strong choice

• Very fast: in-memory with low latency
• Atomic operations: crucial for correct rate limiting under concurrency
• Supports counters and TTLs: ideal for fixed windows, sliding windows, token buckets, etc.
• Widely supported: easy to integrate with most stacks
• Scales well: Redis Cluster can handle larger workloads

Good alternatives

• Memcached: fast, but weaker for rate limiting because it lacks richer atomic patterns and persistence features
• DynamoDB / Cassandra: useful if you need durability and massive horizontal scale, but typically slower and more complex for per-request rate limiting
• Local in-process cache: only good for single-instance or approximate limiting; not suitable for distributed APIs

Recommendation by use case

• Most APIs: Redis
• Very large distributed systems: Redis Cluster or a managed Redis service
• If you need durable counters across restarts: Redis with persistence enabled
• If you need ultra-low latency with approximate limits: combine local token bucket + Redis for coordination

Practical tip

Use Redis operations like:

• INCR + EXPIRE for fixed-window limiting
• Lua scripts for atomic token bucket or sliding-window logic

If you want, I can also suggest:

1. a specific rate-limiting algorithm, or
2. a Redis schema / Lua script for implementation.