
Guide
Backup and Disaster Recovery Guide: RPO, RTO, 3-2-1 and Restore Tests
A backup is a copy of data kept apart from the original; disaster recovery is the plan that puts systems back into service after a serious disruption. For each system, first agree on how much recent data the business can afford to lose (the recovery point objective, RPO) and how long it can stay down (the recovery time objective, RTO), terms defined in NIST SP 800-34 Rev. 1. Then keep three copies on two kinds of storage, one off-site and one offline or immutable; encrypt them, include cloud services, prove them with restore tests and keep a printed runbook.
Updated November 2026
RPO, RTO and maximum tolerable downtime
NIST SP 800-34 Rev. 1, the U.S. federal contingency planning guide, sets three values in its business impact analysis:
- Maximum tolerable downtime (MTD): the total time that the system owner or authorizing official is willing to accept for an outage of a business process.
- Recovery time objective (RTO): how long a system can stay unavailable before the impact on other systems and on the processes it supports becomes unacceptable; normally shorter than the MTD.
- Recovery point objective (RPO): the point before the outage to which data can be recovered from the most recent backup, a measure of how much data loss is tolerable.
The RPO sets how often copies are made: a nightly job cannot meet a four-hour RPO. The RTO sets how the restore is done: downloading a whole server from the cloud may exceed it, while a local copy or a standby machine may not. Agree on both values with the people who run each process, not only with IT.
The 3-2-1 rule and what ransomware changed
A 2012 paper produced for US-CERT, now part of CISA, states the 3-2-1 rule: three copies of any important file (the primary and two backups), on two different media types, one of them stored off-site. The #StopRansomware Guide (CISA, the FBI, the NSA and the MS-ISAC, October 2023) adds what the rule leaves out:
- Keep backups of critical data offline and encrypted, and regularly test their availability and integrity in a disaster recovery scenario.
- Ransomware actors look for accessible backups to delete or encrypt them, try credentials gathered inside the network against backup systems and exploit unpatched backup software.
- Immutable storage can protect stored data, but it should be used with caution: it does not meet the compliance criteria of certain regulations, and misconfiguration can impose significant costs.
NIST SP 800-209 defines immutability as locking data after it is created so that it cannot be altered or deleted. It also asks that the systems managing the copies kept for recovery from an attack be out of reach of regular production and backup credentials. In practice: backup administrator accounts outside the production domain, with multifactor authentication (MFA), and backup software kept patched.
What to back up, cloud services included
NIST SP 800-209 asks that a data protection plan cover every data asset, on premises or in the cloud, and that any deliberate omission be documented. A practical inventory:
- Servers and virtual machines: the whole system, so that it can be rebuilt, not only its files restored.
- Computers: work saved only on a laptop is lost with the laptop.
- Rebuild material: system software and drivers, which SP 800-34 Rev. 1 says to back up, and the “golden images” (preconfigured system templates), installers and license agreements that the #StopRansomware Guide asks organizations to keep.
- Configurations and keys: firewalls, switches, phone systems, the password vault and the recovery keys.
For email and other cloud services, Microsoft's shared responsibility page assigns customer data, configurations and identities to the customer in every model, software as a service included, and AWS states that customers are responsible for managing their data. The #StopRansomware Guide asks organizations to review each cloud's shared responsibility model and to back up data often, offline or cloud-to-cloud.
Backup frequency, types and encryption
SP 800-34 Rev. 1 asks backup policies to set a minimum frequency and scope (daily or weekly, incremental or full) according to how critical the data is and how often it changes. It describes three types:
- Full: everything selected; any file is easy to locate in it, but it can take a long time to run.
- Incremental: what changed since the last backup of any type; quick, but a restore needs the last full backup plus every incremental after it.
- Differential: what changed since the last full backup; a restore needs only that full backup and the latest differential.
SP 800-209 says the encryption protecting backup data should be at least as secure as that of the original data, and asks for encryption in transit as well. SP 800-34 Rev. 1 adds that keys should be stored apart from the encrypted backups yet stay available for recovery: a copy that nobody can decrypt is lost data.
Restore tests that prove a backup works
A job report that says “success” proves only that the job ran. SP 800-209 asks for periodic tests, at least monthly for critical data, to verify that backups are intact and can be restored; an end-to-end test restore into a sandbox for applications with strict restoration speed requirements; and a review of the backup plan at least once a year. NIST CSF 2.0 lists verifying the integrity of backups before restoring from them as recovery outcome RC.RP-03.
Time every full test and compare the result with the RTO. Restore into an isolated network, never over the live system, and record what was restored, how long it took and what failed. The managed IT services guide lists the restore-test question to ask a provider.
The disaster recovery runbook
The runbook is the written, step-by-step part of the plan. It should state:
- Who declares a disaster and who carries out each step, with a deputy for every role.
- The recovery order of the systems, with the RTO and RPO of each.
- Where every copy is kept, and how to reach it when the main site, the directory or email is down.
- Who holds the encryption keys and the backup administrator credentials.
- Restore steps for each system, with its golden image and the hardware to restore onto; the #StopRansomware Guide suggests retaining backup hardware.
- Contacts: suppliers, the internet provider, the insurer and whoever speaks to customers.
- After a cyberattack: rebuild on a clean, separate network from a copy verified as clean (the cybersecurity baseline guide covers the first hour of an incident).
The #StopRansomware Guide asks for a hard copy and an offline version of the incident response plan. Keep the runbook the same way, and update it after every test and every change of systems.
Field notes for storm- and outage-prone sites
SP 800-34 Rev. 1 counts UPS units, generators, water sensors and waterproof containers for backup media among common preventive controls. They matter most where storms and power cuts are routine:
- Distance. Among the criteria SP 800-34 Rev. 1 gives for off-site storage is the probability that the storage site is affected by the same disaster as the main site. A storm or a grid failure can cover a whole region: keep at least one copy in another region, not just in another building in the same city.
- Power. The same publication notes that a UPS usually provides 30 to 60 minutes for a graceful shutdown, and that a UPS and generator combination supplies power as long as the generator has fuel. Measure the UPS runtime at your real load: it must carry every device a backup job depends on, network included, and signal each one to shut down before the battery is empty.
- Water. A sensor in the ceiling or under the floor of the equipment room warns of a leak before it reaches the rack; removable backup media should leave the building in a waterproof container.
- Bandwidth. At 100 Mb/s, uploading or restoring 2 TB takes about 44 hours at full speed, so keep a local copy for fast restores.
Planning checklist
- List systems and data, cloud services included, and agree on the RPO and RTO of each with its owner.
- Set the frequency, backup type and retention of each tier of copies.
- Keep three copies on two kinds of storage, one off-site and one offline or immutable.
- Encrypt the copies, store the keys apart and give backup administration separate accounts with MFA.
- Test restores on a schedule and time them.
- Print the runbook and review it every year.
Frequently Asked Questions
Is RAID or a synced folder a backup?
No. SP 800-34 Rev. 1 presents redundancy that stores data on more than one drive as protection against the failure of a single drive. An accidental deletion, corruption or ransomware encryption reaches every drive of the array at once. Sync is no better: the #StopRansomware Guide warns that files encrypted on a computer can sync to the cloud and overwrite unaffected data. Neither replaces a separate copy with its own history of versions.
Do Microsoft 365 and other cloud services need their own backup?
Check rather than assume. The shared responsibility statements of Microsoft and AWS cited in this guide leave the data with the customer, so read the service's terms: how long deleted items can be recovered, whether that depends on settings you control, and whether someone holding your administrator account could erase them as well. If the answers leave a gap, add a cloud-to-cloud or offline copy.
What is the difference between business continuity and disaster recovery?
In NIST SP 800-34 Rev. 1, a business continuity plan (BCP) keeps processes such as payroll running during and after a disruption. A disaster recovery plan (DRP) restores information systems at an alternate site after a major, usually physical, disruption that denies access to the main facility. The plan for recovering a single system is called an information system contingency plan (ISCP). In everyday use, “disaster recovery” often covers all IT recovery, and backups serve every one of these plans.
How many restore points should be kept, and for how long?
NIST SP 800-209 sets no figures: it asks the data protection plan to state the frequency and retention of each tier of copies, giving “48 hourly snapshots, 30 daily backups” as an example, and to delete on purpose the copies that are no longer needed. To set the retention, estimate how long a problem such as silent data corruption or an intruder could go unnoticed, and keep at least one copy older than that. Record-keeping rules in your sector may impose minimums: check the rule that applies to you.
Sources
- NIST SP 800-34 Rev. 1, Contingency Planning Guide for Federal Information Systems (May 2010, updated November 2010)
- NIST SP 800-209, Security Guidelines for Storage Infrastructure (October 2020)
- NIST CSWP 29, The NIST Cybersecurity Framework (CSF) 2.0 (February 2024; recovery outcome RC.RP-03)
- CISA, #StopRansomware Guide (October 2023; written with the FBI, NSA and MS-ISAC through the Joint Ransomware Task Force)
- US-CERT (now CISA), Data Backup Options (2012; the 3-2-1 rule)
- Microsoft Learn, Shared responsibility in the cloud
- Amazon Web Services, Shared Responsibility Model
About this guide
This guide comes from GlobalTech Corp., an IT and cybersecurity company in Hormigueros, Puerto Rico, that has more than 25 years of experience. It provides secure data backup and disaster recovery with NAS and secure cloud backup, and works with brands such as Veeam and QNAP. It serves companies from one end of Puerto Rico to the other and, in Florida, supports organizations remotely and, when a project needs it, in person. Two pages describe that work: backup and disaster recovery services and backup and disaster recovery for Puerto Rico businesses. Sources reviewed in November 2026.
More guides: Technology Guides.