Multi-Region active-active with Route 53 and DynamoDB global tables
Two Regions, both live, no standby to promote: the most expensive DR strategy, and the one with the lowest RTO and RPO.
Both Regions serve live traffic at the same time
Active-active (also called multi-site active/active) means two or more Regions each run a full, independent copy of the stack (a load balancer, a stateless app tier, and a local replica of the data) and both take real user traffic at the same time. There is no idle standby waiting to be promoted.
Amazon Route 53 sits in front and answers each user's DNS query with the endpoint that is best for them, usually via latency-based routing, so a user in Europe resolves to the European Region and a user in North America resolves to the North American Region. Because both Regions are already warm and scaled, a Regional failure is handled by simply steering all traffic to the survivor: the recovery time objective (RTO) is near zero because nothing has to be built or promoted first. The cost is that you pay for full capacity in every Region all the time, which is why this is the most expensive of the four disaster recovery strategies (backup and restore, pilot light, warm standby, and multi-site active/active).
Active-active steady state: both Regions warm and taking live traffic
Trace
In steady state, two Regions each run a full, independent, warm copy of the stack (public ALB, an Auto Scaling app fleet, a local cache, and a writable DynamoDB replica). Route 53 latency-based routing splits users so North American users resolve to us-east-1 and European users resolve to eu-west-1. Nothing is idle: a Regional failure is absorbed by steering all traffic to the survivor, which is why RTO is near zero.
The data layer is what makes it active-active
The stateless tiers are easy to duplicate. The hard part is the database, and this is where Amazon DynamoDB global tables do the work. A global table is a single table with replicas in multiple Regions, and every replica is writable: this is multi-active replication, not a primary with read replicas. An application in either Region reads and writes its local replica, and DynamoDB asynchronously propagates each change to the other Regions, typically within a second.
Because writes can happen in two Regions at once, DynamoDB resolves conflicting writes to the same item with a last-writer-wins rule: the most recent write, by timestamp, is the one that survives, and the other is discarded. Cross-Region reads are eventually consistent. This gives a recovery point objective (RPO) measured in seconds and an RTO near zero for the data tier, because there is no primary to promote when a Region fails. The trade-off is that the pattern is a poor fit for workloads that need a single global source of truth for every write (for example strict inventory decrements or a financial ledger) where last-writer-wins could silently drop an update.
DynamoDB global tables: multi-active writes, async replication, last-writer-wins
Trace
The data layer is what makes the pattern active-active. A DynamoDB global table keeps a writable replica in every Region, so each Regional app reads and writes its local replica with low latency. Changes fan out asynchronously to the other replicas, typically within a second. When the same item is written in two Regions at once, DynamoDB converges with last-writer-wins by timestamp: the newest write survives and the other is discarded, which is why the pattern is unsafe for a strict ledger or inventory decrement.
What this teaches for the exam
Match the routing policy to the goal. Latency-based routing sends each user to the lowest-latency healthy Region; geolocation or geoproximity routing pins users to a Region for data-residency or business reasons. In every case, attach Route 53 health checks so that when a Region's endpoint fails, Route 53 stops returning it and traffic shifts to the healthy Region automatically.
For controlled, tested failover rather than relying only on health checks, Route 53 Application Recovery Controller (ARC) adds readiness checks and routing controls that you can flip like a manual switch. The decision rule the exam wants: choose active-active when you need the lowest possible RTO and RPO and can accept the cost and the last-writer-wins conflict model; choose warm standby or pilot light when cost matters more than shaving the last seconds off recovery.
Route 53 health checks and ARC: automatic and controlled failover
Trace
Failover is a Route 53 job. Latency records send each user to the lowest-latency Region, and a health check probes each Regional endpoint; when a Region's endpoint fails, Route 53 stops returning that record and traffic shifts to the survivor automatically, with no replica to promote. For controlled, tested failover, Route 53 Application Recovery Controller (ARC) adds readiness checks and routing controls you can flip like a manual switch.
