Definition of the RTO RPO pair
The RTO RPO pair brings together the two objectives an organisation sets itself for recovering an activity after an IT disaster. The RTO (recovery time objective) is the maximum tolerable period of interruption: the time by which the activity must have resumed. The RPO (recovery point objective) is the maximum tolerable data loss: the age, counted as a period of time, of the data one accepts not recovering.
These definitions reflect common usage. Business continuity is the subject of a standard, NF EN ISO 22301 (November 2019), titled “Sécurité et résilience - Systèmes de management de la continuité d’activité - Exigences” (Security and resilience, business continuity management systems, requirements) according to its AFNOR record.
EU law uses a different vocabulary. Article 12(6) of the DORA Regulation, Regulation (EU) 2022/2554, speaks of “recovery time and recovery point objectives”. Financial entities determine them for each function, taking into account in particular whether it is critical or important.
Two questions put to the business
The RTO answers one question: how long can this activity remain stopped? The RPO answers another: how much work is one prepared to lose or re-enter?
Both questions are addressed to the business. They are asked activity by activity, during the business impact analysis, and the answer depends on what an hour of downtime or an hour of lost data costs. Shortening an objective requires additional resources, for example a standby environment or more frequent backups. The choice is therefore a trade-off between the cost of the disaster and the cost of the protection.
How each objective is read
| RTO | RPO | |
|---|---|---|
| Counted | Forwards, from the interruption | Backwards, from the incident |
| Caps | The downtime | The age of the data recovered |
| Depends on | The recovery procedure and architecture | The frequency of backups, logging, replication |
| Proven by | The duration observed during a recovery test | The age of the data recovered during a restore test |
Two conventions are written down with the objective, because they vary from one organisation to another. The first is the starting point of the RTO: the start of the incident, its detection, or the decision to trigger the plan. The second is the meaning of the word “recovery”: service delivered in degraded mode, or a return to normal operation.
RTO, RPO, MTTR, availability: not to be confused
The RTO and RPO are ceilings set in advance. The values observed during a test are measurements, to be compared with those ceilings.
The MTTR (mean time to restore, or to repair depending on the author) is an average calculated over past incidents. The RTO applies to a disaster scenario defined in advance.
An availability rate adds up the interruptions over a period. It does not cap the duration of any single interruption. Service level objectives are presented in our entry on SRE and on our SRE and observability page.
Finally, backup frequency is a means. A backup every four hours only meets an RPO of four hours if the latest backup can be restored.
Worked example: an order-taking application
Take a distributor whose order-taking application is backed up every night. The sales director sets an RTO of four hours and an RPO of one hour.
Two gaps emerge. With a nightly backup, an incident at the end of the day would lose up to a full day of orders: meeting one hour requires logging transactions continuously or replicating them. The first restore test, for its part, takes nine hours: meeting four hours requires a standby environment prepared in advance.
The cost of these two arrangements goes back to the sales director, who confirms the objectives or relaxes them. Setting the objectives with the business, the plans and the tests are described on our cyber resilience page.