Elastic Disaster Recovery

Traditional DR means running a full standby copy of your infrastructure elsewhere — expensive — or restoring from backups and losing hours of data. AWS Elastic Disaster Recovery (AWS DRS) takes the middle path: an agent on each source server does continuous, block-level replication into a cheap staging area in your AWS account (low-cost storage + tiny replication instances). On disaster, DRS converts and boots those volumes as full-size EC2 instances in minutes, giving sub-second RPO and minutes-scale RTO without paying for idle standby compute.

• DRS replicates into a staging area subnet in one target Region (and AZ) of your choosing. Recovery launches instances into your production VPC/subnets in that Region.
• Sources can be on-premises physical/virtual servers, other clouds, or EC2 in another Region/AZ (for AWS-to-AWS DR).
• It supports cross-Region and cross-AZ DR for existing AWS workloads; failback replicates changes back to the origin when the primary site recovers.

graph LR
subgraph SRC["Source Site (on-prem / other cloud / another Region)"]
  S1["Server: app"]
  S2["Server: db"]
end
S1 -->|"DRS agent: continuous block replication"| REP
S2 -->|"DRS agent: continuous block replication"| REP
subgraph STAGE["Staging Area (target Region, low cost)"]
  REP["Replication servers + EBS staging volumes"]
end
REP -.->|"drill / disaster: convert + boot"| REC
subgraph PROD["Recovery in target Region (only running during DR)"]
  REC["Full-size EC2 instances"]
end
REC -.->|"failback when primary recovers"| SRC
classDef src fill:#e8f0fe,stroke:#3b82f6,stroke-width:2px,color:#173a70
classDef stage fill:#eaf7ec,stroke:#2e8b3d,stroke-width:2px,color:#1a3d20
classDef prod fill:#f3ecfb,stroke:#8b3de0,stroke-width:2px,color:#3a1a5c
class S1,S2 src
class REP stage
class REC prod

The whole point is a cheap steady state: you pay for low-cost staging EBS volumes, small replication server instances (one per ~15 source disks), a modest per-source-server hourly DRS fee, and data transfer for replication. You only pay for full-size EC2 + full EBS during drills and actual failover. This “pilot light” economics is the exam selling point versus a hot standby.

• Trigger phrases: “lowest cost DR with fast recovery”, “RPO of seconds, RTO of minutes”, “don’t want to pay for idle standby servers”, “DR for on-premises or non-AWS servers into AWS” → AWS Elastic Disaster Recovery.
• DRS is the successor to CloudEndure Disaster Recovery. If you see CloudEndure DR in an answer, DRS is the modern equivalent.
• AWS DRS (disaster recovery) vs AWS Application Migration Service / MGN (one-time migration) — same underlying replication tech, different lifecycle: DRS keeps replicating and supports failback; MGN cuts over once.
• Run non-disruptive recovery drills regularly — launching drill instances in an isolated subnet doesn’t affect the source or ongoing replication. The exam likes “how do you test DR without impacting production”.
• Maps to the “pilot light” / “warm standby” DR strategies — cheaper than multi-site active/active, faster than backup-and-restore.
• For database-only DR, native options (RDS cross-Region read replicas, Aurora Global Database, DynamoDB global tables) are usually preferred; DRS shines for whole-server and heterogeneous estates.