Why Go Multi Region
Users on different continents suffer high latency talking to a single region, and one region can fail entirely. A multi region database places data in several geographic regions to lower latency for local users and survive a regional outage.
The Latency Wall
Physics sets a floor: a round trip between distant regions takes tens to over a hundred milliseconds. A write that must be synchronously confirmed by another region inherits that delay. So the central tension is between consistency and latency.
Common Patterns
- Single write region One region owns all writes; others hold read replicas. Local reads are fast everywhere, but writes always cross to the home region, so distant writers pay the latency.
- Read local write home A refinement where reads serve from the nearest replica while writes route home, accepting replication lag for reads.
- Multi write Several regions accept writes, which removes the write latency penalty but introduces conflicts when two regions edit the same data at once.
Failure and Data Residency
Multi region also brings disaster recovery: lose a region and another takes over. It can satisfy data residency laws by pinning certain records to a required region. The harder the consistency guarantee across regions, the more each write pays in cross region round trips.
Key idea
Multi region databases cut read latency and survive regional outages, but synchronous cross region consistency collides with the latency wall set by physics.