The right distance comes from a trade-off between replication latency and risk separation.
- Synchronous replication (every write waits for the backup site’s acknowledgment — storage arrays, critical databases): latency caps the distance. Count ~1 ms RTT per 100 km of fiber: beyond 100 km (~2 ms), write performance degrades. The sweet spot: 15–60 km — far enough for distinct buildings and power feeds, close enough for synchronous.
- Asynchronous replication / DRP: distance is no longer a technical constraint — it becomes a requirement: 200 km or more, to leave the shared risk zone (flood, earthquake, regional power outage).
Also verify the actual interconnection between the two sites (common carriers, declared circuits): distance is useless if the two data centers cannot be properly linked.