The pattern
Many systems read far more than they write. Read replica fan out keeps one primary that takes writes and copies those changes to several replicas that only serve reads. Read traffic spreads across the replicas.
How it scales
- The primary handles all writes in one place, keeping correctness simple.
- Each new replica adds more read capacity.
- A read router or proxy spreads queries across the pool.
The catch is lag
Replicas apply changes a moment after the primary, so they can be stale. This is replication lag.
- A user who just wrote data may read an old value from a replica.
- For freshness sensitive reads, route to the primary or use read your writes routing.
The pattern is cheap and powerful for read dominated systems, as long as the application tolerates small staleness on most reads.
Key idea
Send writes to one primary and fan reads across many replicas to scale read heavy traffic.