TIL: When Redis Streams beat Kafka

Reached for Kafka out of habit on a small pipeline and felt the weight of it. Redis Streams turned out to be the right call when:

• Throughput is moderate (thousands/sec, not millions)
• You already run Redis and want one fewer system to operate
• Consumer groups + acknowledgements are enough; you don’t need log compaction or infinite retention
• Latency matters more than durability guarantees

Kafka still wins for huge fan-out, replay-from-the-beginning semantics, and multi-day retention. But “we might need Kafka someday” is not “we need Kafka today.” Pick the boring tool you already run.