Disaster Recovery
Disaster recovery (DR) is the set of policies, tools, and procedures organizations use to restore IT systems, data, and operations after a disruptive event such as a hardware failure, cyberattack, or natural disaster.
Definition
Disaster recovery (DR) is the set of policies, tools, and procedures organizations use to restore IT systems, data, and operations after a disruptive event such as a hardware failure, cyberattack, or natural disaster.
Overview
Disaster recovery is usually planned around two core metrics: Recovery Time Objective (RTO), how long systems can be down before the business is meaningfully harmed, and Recovery Point Objective (RPO), how much data loss (measured in time) is acceptable. Together they determine how aggressive — and expensive — a DR strategy needs to be, ranging from simple daily backups to fully mirrored standby environments that can take over in seconds. In cloud environments, DR strategies commonly rely on geographically separate regions, replicated Block Storage or object storage, and automated failover using a Load Balancer or DNS-based routing. Common patterns include backup-and-restore (cheapest, slowest), pilot light (a minimal always-on copy of critical systems), warm standby (a scaled-down but running duplicate), and multi-site active-active (near-instant failover at the highest cost). Regular testing of these plans — not just writing them — is what separates a real DR capability from a document that looks good in an audit. DR is closely related to but distinct from high availability: HA is about surviving individual component failures without downtime, while DR is about recovering after a broader outage that HA alone couldn't prevent, such as an entire data center or region going offline. Cloud architecture courses like AWS Solutions Architect and Cloud Security both treat DR planning as a core competency, since it directly affects an organization's resilience and compliance posture.
Key Concepts
- Defined by Recovery Time Objective (RTO) and Recovery Point Objective (RPO)
- Ranges from backup-and-restore to fully active-active multi-region setups
- Relies on data replication across separate physical or cloud regions
- Uses automated failover via load balancing or DNS routing
- Distinct from high availability, which addresses component-level failure
- Requires regular, realistic testing to be effective
Use Cases
Frequently Asked Questions
From the Blog
Agent Error Recovery: Retries, Fallbacks and Dead Ends
An agent recovers from a tool failure only if the failure reaches it as a readable observation rather than an exception. This covers turning errors into structured observations, deciding what the runtime retries versus what the model retries, and setting the give-up rules that stop a run looping forever.
Read More AI & TechnologyDesigning streaming UX for LLM responses: latency, cancellation and errors
Streaming changes the error model: the response fails after it starts. Learn cancellation, partial state, mid-stream failure recovery and honest progress feedback.
Read More