What is cloud disaster recovery?
Cloud disaster recovery (cloud DR) is a plan for restoring applications, data, and infrastructure after an outage, cyberattack, hardware failure, or other disruptive event. It uses cloud resources to store recovery copies, replicate workloads, or run standby systems, instead of relying on a second physical data center. The right approach depends on how quickly systems must return, how much data the business can afford to lose, and the budget.
RTO and RPO
Two targets shape every disaster recovery plan:
- Recovery time objective (RTO) is the maximum acceptable downtime for a system. An RTO of two hours means the system should be back within two hours.
- Recovery point objective (RPO) is the maximum acceptable data loss, measured as time. An RPO of 15 minutes means the business can lose at most about the last 15 minutes of changes.
These are targets, not guarantees. Test them to see whether your setup meets them.
The four main DR strategies
AWS describes four common strategies, ordered from the cheapest and slowest to the most costly and fastest. Other providers describe similar approaches.
| Strategy | How it works | Typical recovery time | Cost |
|---|---|---|---|
| Backup and restore | Backups are stored in the cloud. Infrastructure is built and data restored after a disaster. | Hours | Lowest ongoing cost |
| Pilot light | Core data is replicated and essential services are kept ready, but other components stay off until needed. | Tens of minutes | Moderate |
| Warm standby | A scaled-down but working copy of the environment runs continuously and scales up during recovery. | Minutes | Higher |
| Multi-site active/active | Workloads run in two or more locations and serve traffic at the same time. | Near zero | Highest |
Match the strategy to each system’s RTO and RPO, rather than choosing one strategy for everything. Critical services may justify warm standby or active/active. Internal tools may be well served by backup and restore.
Examples
These examples are illustrative, not drawn from a specific customer:
- An online store takes orders all day. Its checkout service might use warm standby, with an RPO of minutes, because lost orders cost money. Its archive of old product images can use backup and restore, with a much longer RTO.
- An internal reporting tool that staff use on weekdays might accept an RTO of one business day. Backup and restore, with nightly backups and periodic restore tests, could be enough.
Backups, high availability, and disaster recovery are different
- Backups keep copies of data so you can restore it, including after deletion or corruption.
- High availability keeps a service running through component failures, often by spreading it across availability zones.
- Disaster recovery restores service after a larger failure, such as the loss of a whole region or site.
Each addresses a different failure. Replication alone is not a backup, because it copies deletions and corruption to the other site as well. Keep point-in-time backups too.
Multi-AZ and cross-region protection
Spreading resources across availability zones protects against the failure of one data center. Protecting against the failure of a whole region requires copies in a second region. Some providers also offer geo-redundant services that replicate data between paired regions.
Protect backups from ransomware
Ransomware often targets backups as well as production data. Immutable storage keeps backups from being changed or deleted for a set period. Examples include S3 Object Lock, AWS Backup Vault Lock, and Azure immutable blob storage.
The 3-2-1 rule is a simple baseline: keep three copies of your data, on two different types of media, with one copy offsite. A common extension, 3-2-1-1-0, adds one immutable or offline copy and zero errors after backup verification.
Test your plan
An untested DR plan is hard to trust. Testing can take several forms:
- A tabletop exercise, in which the team walks through the plan together.
- A restore test, in which backups are restored and checked for integrity.
- A partial failover, in which one system moves to its recovery environment.
- A full drill, which exercises the whole plan, including communication.
Many organizations test at least once a year, and the right frequency depends on policy and risk. Record how long each step takes, and compare the results with your RTO and RPO.
Shared responsibility
The provider is responsible for the resilience of the cloud itself, such as its data centers and regions. The customer is responsible for resilience in the cloud. That covers the backups, replication, configuration, and failover logic for its own workloads. Moving to the cloud does not mean the provider handles disaster recovery for you.
Tools
Common options include AWS Elastic Disaster Recovery and AWS Backup, Azure Site Recovery and Azure Backup, Google Cloud Backup and DR, and third-party products such as Veeam and Zerto.
Common misconceptions
- Backups alone are not a DR plan. You also need restore procedures, recovery targets, and tests.
- High availability is not disaster recovery. A highly available system can still fail with the region it runs in.
- Replication also copies corruption and deletions. Point-in-time backups are still needed.
- Moving to the cloud does not mean the provider handles your DR. Responsibility is shared.
- An untested plan should not be trusted.
An SLA’s availability figure does not replace a recovery plan. The life cycle of a cloud SLA explains what those figures do and do not promise.