Cloud & Infrastructure
High Availability Is Not Disaster Recovery: Explain RTO and RPO
Explain RTO and RPO with a worked incident timeline, then distinguish recovery targets from actual recovery performance.
Accumentum® ·
The recovery time objective (RTO) is the target maximum time to restore a service. The recovery point objective (RPO) is the target maximum tolerable data-loss window, expressed in time. In the drill below, recovery takes 42 minutes and the restored data is 12 minutes behind. A 45-minute RTO passes while a 10-minute RPO is missed.
AWS Well-Architected Recovery Objectives
Teams sometimes describe a multi-region deployment as "disaster recovery" before they have defined what the business needs to recover. High availability reduces interruptions in expected failure modes. Disaster recovery addresses how a workload returns after disruptive events. The distinction becomes operational when you define recovery time objective (RTO), recovery point objective (RPO), dependencies, authority, and test evidence.
RTO is the target time to restore a workload after an interruption. RPO is the maximum acceptable interval of potential data loss expressed as time. Both are business requirements that drive technical design, cost, and testing. Do not treat them as guaranteed outcomes. A service can have replicated servers yet rely on a single administrator, key store, identity provider, or data pipeline that compromises recovery.
How to practice
- Define service impact: Ask which transactions or users are critical and what downtime or data loss is tolerable.
- Map dependencies: Include identity, DNS, networking, secrets, data, integrations, and operating personnel.
- Choose an approach: Compare backups, pilot light, warm standby, and multi-site designs against cost and recovery targets.
- Prove recovery: Rehearse failover or restoration, reconcile data, measure elapsed time, and document unresolved risk.
The workplace scenario
An order-processing service uses servers in two regions but writes to one database region. The architecture may improve front-end availability yet not meet the order recovery objective. Before recommending a new design, clarify accepted data loss, financial reconciliation, cross-region dependencies, and the owner's ability to authorize failover. The answer should explain why the selected approach meets the stated requirement rather than praising a particular cloud feature.
Calculate the gap your test actually measured
In a fictional restore drill, a service becomes unavailable at 09:00. The team brings the application up at 09:31, and the owner verifies that staff can complete a new order at 09:42. The latest consistent recovered order is from 08:48. If the approved targets are 45 minutes to resume the order service and 10 minutes of tolerable data loss, the measured recovery time is 42 minutes and the recovery point is 12 minutes before the outage. One target passes and the other misses by two minutes. Both results belong in the report.
Record what the service owner means by “resume.” A green process indicator at 09:31 would imply a different conclusion if the team had never agreed that completing an order was the acceptance test. Also record how you established the last consistent order. A backup job’s successful completion time alone may not tell you which transactions the recovered system contains.
Use the miss to choose the next investment
If restoring the application consumed little time but verifying the business workflow took most of the drill, buying faster compute may have limited effect. If the data gap exceeded tolerance, investigate the protection and recovery process with its owner. Run the same business transaction on the next approved drill and compare the two measurements independently. A single “recovered successfully” label would hide the reason to invest.
A common mistake
A successful backup restore proves that an image is readable. It does not prove the entire application can reopen. Include data consistency, identity access, integrations, and operational acceptance. Be precise about what the test did and did not demonstrate.
Try it yourself
Design two recovery options for a fictional payments service with different RTO and RPO targets. Write assumptions, likely operating cost differences, dependencies, test schedule, and the business owner who accepts residual risk.
Does a second cloud region automatically achieve zero downtime?
No. Replication, consistency, routing, operational procedures, and recovery testing determine outcomes. Zero downtime should not be assumed.
Put it into practice
Practice architecture tradeoffs in PathHire's Cloud Solutions Architect or Cloud Engineer pathways and use Accumentum cloud training to reinforce design principles.