Fan-out

Problem

One event or write must be distributed to many consumers or write destinations.

Context

A new post → notify N followers; a config change → push to N services; a user action → update N cached views.

Solution

Publish the event once to a broker (pub/sub, event bus). The broker delivers to each subscriber. For 'push-based' fan-out (write to N destinations synchronously), use a fan-out worker with parallel writes and per-target retries.

Trade-offs

• Fan-out amplification: 1 event × 1M consumers = 1M messages
• Slow consumers can cause backup (with per-consumer queues, this is contained)
• Ordering across consumers is usually not guaranteed

Failure modes

• One target is slow and slows the whole fan-out (mitigate with async + timeouts)
• Fan-out storm on a viral post overwhelms the broker
• Duplicate deliveries because the broker retries on ACK loss

When to use

• Notification systems (post to 1M followers)
• Cache invalidation across regions
• Multi-consumer analytics pipelines

When NOT to use

• Single-consumer pipelines
• Ultra-low-latency real-time paths where fanning out adds too many hops