AWS Elastic Disaster Recovery
Continuously replicates your servers into AWS so you can boot them as EC2 instances minutes after your primary site dies.
Verified against AWS documentation · 2026-09-09
Key facts
- 01An agent on each source server streams block-level changes into a low-cost staging area in your AWS account — EBS volumes and small replication servers, not an idle copy of production
- 02RPO in seconds, RTO in minutes; DRS converts the disks to boot natively on EC2 at launch time, so there's no replatforming
- 03Sources: on-prem physical servers, VMware/Hyper-V, another cloud, or EC2 instances in another Region or AZ
- 04Point-in-time recovery — launch from an earlier snapshot instead of the latest state (default 7 days of daily points, adjustable 1–365)
- 05Naming check: DRS is the successor to CloudEndure Disaster Recovery, retired for most customers in 2024 — older study material still says CloudEndure
Exam triggers
Phrases and scenarios that point here.
- "Minimize downtime and data loss", "RPO of seconds", "RTO of minutes", "disaster recovery to AWS"
- On-premises servers must come back up in AWS without rebuilding or re-architecting them → DRS
- DR standby required, but paying for a duplicate running environment is not → DRS staging area, not a hot standby
- Ransomware or data corruption, need to roll back to a known-good moment → DRS point-in-time recovery
- "Test the DR plan regularly without disrupting production" → DRS drills
Not this service
- Restorable copies of data, hours of RPO/RTO acceptable → AWS Backup
- One-way move to AWS with the source decommissioned afterwards → AWS Application Migration Service (MGN)
- Routing traffic and checking readiness for an app already running in multiple Regions → Route 53 Application Recovery Controller
Use cases
- 01A distributor running its warehouse system on two racks in the back office keeps a warm copy in a Region, so a burst pipe upstairs costs hours of downtime instead of weeks
- 02Quarterly DR drills for auditors — launch the whole recovery fleet into an isolated VPC, capture the proof, delete it, and production never notices
- 03Ransomware rollback — recover to a point in time from before the encryption started, rather than discovering last night's backup is already infected
- 04Cross-Region cover for EC2 workloads built single-Region that can't be re-architected for active-active before the compliance deadline
- 05Legacy Windows boxes nobody wants to touch — DRS copies the disks as-is, so nobody has to reinstall an app whose original engineer left in 2014
How it's charged
- 01Flat hourly fee per source server actively replicating — covers continuous replication, drills, recovery launches and point-in-time recovery, and doesn't change with disk count, storage size, or target Region
- 02That service fee is the small line item; staging EBS volumes and point-in-time snapshots usually cost more, and both scale with disk size and daily change rate
- 03Replication servers are ordinary EC2 instances in the staging subnet (small ones by default), billed as EC2 — one server handles the disks of several source servers
- 04Drill and recovery instances are billed as normal EC2 + EBS for as long as they run — leaving a drill fleet up over a weekend is the classic surprise bill
- 05Longer point-in-time retention means more snapshots and more storage cost; stop replicating a server and its DRS charge stops with it
