Choosing a Disaster Recovery Pattern for the Data Tier

Choosing a disaster recovery pattern for a data tier only means something once it’s tied to specific numbers, not general intentions like “highly available” or “resilient.” Backup and restore, warm standby, and active-active all solve the same underlying problem: staying usable after a serious failure. However, they trade off cost, complexity, and recovery speed in different ways. The right choice depends entirely on which numbers are actually in play.

Here’s a concrete version of that trade-off. A production data tier needs a disaster recovery (DR) choice that matches numbers, not hopes. Assume an availability target of 99.9%, peak traffic of around 200 requests per second to the primary read path, and a data residency constraint that keeps all customer data in one region. Monthly DR spend is capped at a small multiple of steady state, and a team of two people handles both deploys and on-call incident response, so anything chosen has to be operable by that same team under pressure.

Now apply pressure. The business sets a strict RPO of five minutes, a moderate RTO of 60 minutes, and compliance requires encryption and an audit trail for every recovery action. With that set of constraints, backup and restore, warm standby, and active-active are all technically possible, but each pushes cost and operational burden in a different direction.

What a DR pattern means in practice

A DR pattern is a runbook promise. It ties ...