← Back to blog

Cloud Disaster Recovery: Plan, Tiers and Tests

November 13, 2026
Cloud Disaster Recovery: Plan, Tiers and Tests

Cloud disaster recovery (cloud DR) means keeping a recoverable copy of your systems, data and configuration in a public cloud, so you can bring the business back after an outage, ransomware attack or site loss. It works when you decide, system by system, how much downtime and data loss you can tolerate, and then pay only for the recovery pattern that meets that target. This guide covers what that means, how it differs from backup, what AWS and Azure give you, and how to plan, test and maintain it. We run backup, disaster recovery and business continuity planning for clients as part of one managed IT and cybersecurity contract, so that is the angle we take.

AccountNext-Nexus
Plan for a More Resilient Recovery
Nexus brings disaster recovery, backup and business continuity planning together with managed IT and cybersecurity services.
Visit

TL;DR

  • Cloud DR copies applications, data and infrastructure to a cloud region so you can restart them there. Backup alone only keeps data.
  • Two numbers drive every design choice: RTO (how long you can be down) and RPO (how much data you can lose).
  • AWS describes four patterns, from backup and restore to multi-site active/active. Azure covers the same ground with continuous replication and failover. Cost rises as recovery time falls.
  • A plan nobody has tested is a guess. Tier your systems, then test each tier on a schedule.
  • Regulated data (HIPAA, PCI-DSS, SOC 2, ISO 27001) adds encryption, access and evidence requirements to the DR design.

Table of Contents

What cloud disaster recovery is

Disaster recovery is the process of restoring critical systems and data after an unexpected disruption, and it is a core part of business continuity planning, according to Microsoft Azure's cloud computing dictionary. Cloud DR applies that idea to a public cloud. TechTarget describes it as a combination of strategies and services that back up data, applications and other resources to a public cloud or a dedicated provider, so that after a disaster the affected resources can be restored to your own data center or to a cloud provider to resume normal operation.

A "disaster" is wider than a flood or a fire. In practice, the events that trigger a recovery are:

  • Ransomware or another cyberattack that encrypts or deletes production systems.
  • A failed change, a bad patch or an operator mistake that corrupts a database.
  • Loss of a server room, a data center or a cloud region.
  • A prolonged power or network failure at your main site.
  • A failed cloud migration or a misconfigured storage account.

The cloud changes the economics. A traditional DR site means buying and housing duplicate hardware that sits idle. In the cloud you can keep a small footprint running (or nothing at all) and scale up only when you need it. That is why cloud DR now suits small and mid-sized businesses as well as large enterprises.

Why organizations need it

Azure's dictionary lists the practical gains of a well-structured plan: minimal downtime, data protection through regular backups and redundant storage across regions, cost control by avoiding the expense of prolonged outages, compliance with industry regulations, and confidence among customers and stakeholders. Each of those is a business argument, not a technical one.

Three reasons come up most often when we talk to owners and IT leaders:

  1. Downtime has a daily cost. Every hour a billing system, patient portal or order platform is offline, revenue and trust leak out. A DR plan turns an open-ended outage into a bounded, rehearsed one.
  2. Attackers target backups. Ransomware crews look for backup repositories first. Recovery copies that sit in a separate account or region, with separate credentials, are far harder to destroy than copies on the same network.
  3. Auditors ask for it. SOC 2, HIPAA, PCI-DSS and ISO 27001 all expect documented continuity and recovery arrangements, and evidence that you exercise them.

Cloud backup vs disaster recovery

This is the most common point of confusion, and it decides whether you are protected or only think you are.

Cloud backupCloud disaster recovery
PurposeKeep a copy of data so you can restore files or databasesRestart whole systems (servers, networks, applications) so the business can operate
What is protectedData, sometimes system imagesData, configuration, infrastructure definitions, network paths, access
Typical question it answers"Can we get yesterday's file back?""Can we be trading again by 2 p.m.?"
Recovery timeHours to days, depending on size and where it is restoredSet by a target (RTO) and built into the design
Needs a runbookRarelyAlways: order of restart, owners, contacts, failback

Backup is the foundation of DR, not a substitute for it. A set of backups tells you nothing about where the servers will run, how users will reach them, who restarts what, or in which order. The widely used 3-2-1 approach (three copies, two media types, one off-site) is a good baseline for the backup layer, and the 3-2-1 backup rule explains it with examples. Our own guide on the 3-2-1 backup rule for SMBs adds an offline copy, which matters when the threat is ransomware.

How cloud disaster recovery works

Whatever the provider, the mechanics have the same five parts:

  1. Protect. Data and, where needed, whole servers are copied to a recovery location, either as periodic backups or as continuous replication.
  2. Detect. Monitoring spots the failure or attack and someone with authority declares a disaster. Round-the-clock monitoring matters here, because a recovery that starts six hours late misses its target before it begins.
  3. Fail over. Systems are started in the recovery environment in a defined order: identity and network first, then databases, then applications.
  4. Operate. Users are redirected, and the business runs from the recovery environment while the primary is repaired.
  5. Fail back. Once the primary site or region is healthy, data created during the outage is synchronized back and traffic returns.

The difference between approaches is how much of step 1 is kept warm and how much is built on demand at step 3.

RPO and RTO: the two numbers that set the design

Every DR conversation reduces to two targets. As TechTarget explains, RTO is the maximum acceptable downtime for a system or business process, and RPO is the maximum amount of data that can be lost during an outage.

A simple way to remember it: RPO looks backward from the failure (how old can the last good copy be?) and RTO looks forward (how long until we are running again?).

Examples of how different systems end up with different targets:

SystemBusiness impact of an outageSensible direction
Online payments or patient schedulingImmediate revenue or care disruptionMinutes of RTO and RPO, continuous replication
Email and collaborationHigh, but work can pause brieflyHours of RTO, frequent backups
Internal reportingLow in the short termA day or more is acceptable
Archived recordsMinimalBackup and restore is enough

The figures must come from a business impact analysis (BIA), not from what the technology can do. IT teams tend to set targets from what is easy to build, and the business then discovers the gap during an outage. Our guides on RTO vs RPO with BIA-backed targets and on starting disaster recovery planning with a BIA walk through how to derive them.

Disaster recovery on AWS

AWS documents four disaster recovery options in the cloud: backup and restore, pilot light, warm standby and multi-site active/active. Its Elastic Disaster Recovery service also lets you use an AWS Region as the recovery target for a workload hosted on-premises or on another cloud.

The four patterns trade cost against recovery speed:

PatternWhat runs in the recovery RegionRecovery speedRelative cost
Backup and restoreNothing until needed; data copies are storedSlowest (hours)Lowest
Pilot lightCore data layer live; compute switched offFasterLow to moderate
Warm standbyA scaled-down but complete copy of the workloadMinutesModerate to high
Multi-site active/activeFull production in two or more placesNear zeroHighest

A practical way to choose: put each workload in the cheapest pattern that still meets its RTO and RPO. A customer-facing platform may justify warm standby, while internal tools sit comfortably in backup and restore. If you are moving to AWS anyway, build the recovery design into the migration; our eight-step cloud migration checklist shows where DR sign-offs belong.

Disaster recovery on Azure

On Azure, the core service is Azure Site Recovery. Microsoft's documentation says Site Recovery provides infrastructure resiliency for business-critical workloads through continuous replication, failover and recovery, helping keep apps and workloads running during outages. It can replicate from on-premises environments into Azure, or between Azure regions.

What to weigh when using Azure for DR:

  • Region pairing and distance. Choose a recovery region far enough from production that one event cannot hit both, and check data residency rules for the data you hold.
  • Identity. Recovery is useless if staff cannot sign in. Plan how your directory and multi-factor authentication work in the recovery region.
  • Network design. Address ranges, DNS, firewalls and VPN or private links must exist in the recovery region before the failover, not be improvised during it.
  • Recovery order. Define which virtual machines start first and which depend on others.
  • Cost visibility. Replication, storage and any standby compute are ongoing charges, so tag and review them.

The same logic applies to Google Cloud. Many organizations run workloads across more than one provider, and a DR design can use that to its advantage. Our multi-cloud strategy guide covers the risk and cost side of that decision.

How to build a cloud DR plan, step by step

The sequence below is what a good plan looks like in any cloud. At each step we note where it usually goes wrong, and what we take on for clients who contract us for it.

1. Run a business impact analysis

List the business processes, the systems behind them, the cost of an hour of downtime, and the data that cannot be recreated. Interview process owners rather than guessing from the IT side. The common failure is a BIA done once and never updated after a new system goes live. We run business continuity planning and carry the BIA into the recovery design, so targets and architecture stay linked. See also our guide to business continuity planning you can test this week.

Process owner explaining business-critical work

2. Set tiers with RTO and RPO

Group systems into three or four tiers, each with its own targets. Fewer tiers are easier to test and to price. The failure here is putting everything in the top tier, which makes the plan unaffordable and so never fully funded.

3. Choose a recovery pattern per tier

Map each tier to backup and restore, pilot light, warm standby or active/active on AWS, or the equivalent on Azure or Google Cloud. We design and manage cloud environments across AWS, Azure and Google Cloud, so one team can compare the options against your targets rather than recommending whichever platform it happens to know.

4. Secure the recovery copy

Encrypt data in transit and at rest, keep recovery credentials separate from production ones, and make at least one copy immutable or offline so that an attacker who controls production cannot erase it. Data encryption and backup and disaster recovery sit in the same service line at Nexus, and 24/7 threat detection and monitoring covers the environment around them.

5. Write the runbook

A runbook names who declares a disaster, the order of restart, how to reach staff and vendors if email is down, and how to fail back. The failure is a runbook that lives only on the server that just failed. Store copies where they can be reached during an outage.

6. Test on a schedule

A restore that has never been run is an assumption. Test the top tier more often than the bottom one, and record the actual recovery time against the target. Our guide to disaster recovery testing frequency by system tier sets out a cadence you can adapt.

7. Maintain it

Re-run the BIA after major changes, review costs, rotate credentials, and update the runbook after every test. A DR plan decays as systems change, and the usual cause of failure is drift between the plan and reality.

Security and compliance in cloud DR

Your recovery environment holds the same data as production, so it carries the same obligations. In practice:

  • Healthcare organizations that handle protected health information need backups and recovery copies that meet HIPAA requirements, and evidence that restores work. Our guides on HIPAA compliant backups for US health IT and the HIPAA compliant cloud playbook go through the detail.
  • Payment data brings PCI-DSS scope into the recovery region as well as production.
  • SOC 2 and ISO 27001 audits look for documented continuity plans, tested recovery and access control over recovery systems.

We carry out compliance assessments for SOC 2, HIPAA, PCI-DSS and ISO 27001, so the DR design and the audit evidence come from the same engagement instead of two vendors who do not talk to each other.

Common mistakes

  • Treating backup as DR. Data without a place to run it is not recovery.
  • Single-account recovery. If the same administrator account controls production and recovery, one stolen credential ends both.
  • Ignoring dependencies. Applications depend on DNS, identity, licensing, certificates and third-party APIs. Map them.
  • Never testing failback. Teams rehearse the move to the recovery site, then discover that returning is harder.
  • Fragmented ownership. When backup, cloud, security and compliance are bought from four vendors, nobody is accountable for the whole recovery.

Promo

If you want backup, disaster recovery, cloud management, security monitoring and compliance handled together, that is how we structure our work: Nexus offers cybersecurity, cloud solutions, managed IT services, data protection and compliance and risk services under one unified SLA, with a single point of accountability and fixed-fee monthly pricing.

AccountNext-Nexus

Small teams with limited budgets and regulated organizations such as healthcare providers both use that model to avoid managing several vendors during an incident. Start from the Nexus home page or go straight to the services overview to scope a recovery design for your tiers.

Our perspective

Most outages that turn into disasters do so because responsibility is split. The backup vendor says the copies are fine, the cloud team says the servers are fine, the security team is looking at an alert, and nobody owns the clock that is running against the business.

That is why we consolidate. When monitoring, incident response, cloud management, backup and compliance sit under one agreement, one team is accountable for getting you back, and the recovery targets you set in planning are the same ones we are measured against during an incident.

Our advice is to start with the business impact analysis, tier your systems honestly, buy only the recovery speed each tier needs, and test every tier on a schedule. Flat monthly pricing helps here: you can budget the DR program without waiting for a surprise invoice after a failover.

Nick - Sr. Executive

FAQ

What is cloud disaster recovery in simple terms?

It is a way of keeping copies of your data, applications and system settings in a public cloud so you can restart them there if your main site fails. After the incident you can run from the cloud until the primary environment is repaired, then switch back.

What is the difference between cloud backup and disaster recovery?

Backup stores copies of data so you can restore files or databases. Disaster recovery restores whole working systems, including servers, networking, access and the order in which they start, against a defined recovery time. Backup is one ingredient of a DR plan.

What is the difference between RPO and RTO?

RPO is the maximum amount of data you can afford to lose, which sets how often you copy data. RTO is the maximum downtime you can accept, which sets how fast you must restore. Different systems can and should have different targets.

Does AWS provide disaster recovery?

Yes. AWS documents four options: backup and restore, pilot light, warm standby and multi-site active/active. Its Elastic Disaster Recovery service can use an AWS Region as the recovery target for workloads running on-premises or on another cloud.

How often should we test our cloud DR plan?

Test according to tier: your most critical systems most often, lower tiers less often, and always after a major change. Record real recovery times against targets and update the runbook each time.

Sources