Disaster recovery

Disaster recovery is the part of resilience engineering that deals with events large enough that the running system cannot absorb them — a Region-wide impairment, a destructive deployment, a data corruption that replicated everywhere before anyone noticed.

Two numbers drive every decision here. The Recovery Point Objective is how much data the business will accept losing, measured in time. The Recovery Time Objective is how long it will accept being unavailable. Both are business commitments, not technical preferences, and both should be agreed per workload before any architecture is drawn.

AWS disaster recovery strategies sets out the four options AWS defines and what each costs in money and in recovery time. The definitions the numbers rest on — RPO and RTO, business continuity versus disaster recovery, high availability versus fault tolerance — are in high availability and disaster recovery concepts, which is worth reading first if any of those terms are being used loosely in the room.