Start with recovery targets
Every workload needs two agreed numbers: the Recovery Time Objective (RTO), how long it may be unavailable, and the Recovery Point Objective (RPO), how much data may be lost. These come from business impact, not technology preference, and they decide the architecture and its cost.
Four disaster recovery patterns
| Pattern | Typical RTO / RPO | Relative cost | Good fit |
|---|---|---|---|
| Backup & restore | Hours / hours | Low | Internal tools, archives |
| Pilot light | Tens of minutes / minutes | Moderate | Core business apps |
| Warm standby | Minutes / seconds | Higher | Customer-facing services |
| Multi-site active-active | Near zero / near zero | Highest | Payments, critical platforms |
Backup practices that survive ransomware
- Follow 3-2-1: three copies, two media or services, one isolated copy.
- Use immutable storage such as S3 Object Lock, Azure immutable blobs or GCS bucket lock.
- Store backups in a separate account, subscription or project with distinct credentials.
- Encrypt backups with keys whose deletion requires multi-party approval.
- Back up configuration and infrastructure code, not only data.
- Alert on backup failures, deletion attempts and retention changes.
Native cloud tools
AWS
AWS Backup, Elastic Disaster Recovery, Aurora Global Database, Route 53 health-checked failover.
Microsoft Azure
Azure Backup, Azure Site Recovery, geo-redundant storage, Front Door and Traffic Manager.
Google Cloud
Backup and DR Service, regional persistent disks, Cloud SQL cross-region replicas, global load balancing.
Cross-platform
Terraform for rebuildable environments, Velero for Kubernetes, and runbook automation for repeatable failover.
Test recovery, not just backups
A backup is only proven when a restore succeeds within the RTO. Schedule quarterly restore tests, annual full failover exercises for critical services, and game days that simulate a lost region or compromised credentials. Record actual recovery times and fix the gaps found.
Know exactly how fast you can recover
We design, automate and test disaster recovery across AWS, Azure and Google Cloud.
Request a resilience review