Question

Difficulty: HardMulti-AZ, Multi-Region Architectures and Disaster Recovery (DR)

A company hosts a critical global multiplayer gaming backend in us-east-1. The backend utilizes Amazon Aurora MySQL for player profiles, session states, and game configuration data, and an Application Load Balancer (ALB) with Amazon ECS Fargate for the web API layer. The company needs to establish a Disaster Recovery (DR) strategy in us-west-2 with a Recovery Time Objective (RTO) of less than 1515 minutes and a Recovery Point Objective (RPO) of less than 11 minute. They also want to minimize monthly standby running costs. Which of the following architectures meets these requirements?

  1. Configure Amazon Aurora Global Database with the primary cluster in us-east-1 and a secondary cluster in us-west-2. In us-west-2, deploy the ECS Fargate service with the desired task count set to 00. Configure Amazon Route 53 failover routing with health checks. During a failover event, promote the secondary Aurora cluster to primary, scale the ECS Fargate task count to the required capacity, and update the DNS records.Answer
  2. B
    Deploy an Amazon Aurora MySQL database with a cross-region Read Replica in us-west-2. Configure the ECS Fargate service in us-west-2 to run at full capacity. Use Amazon Route 53 latency-based routing to automatically failover application write traffic to the replica in us-west-2 if the primary region becomes unhealthy.
  3. C
    Implement a Backup and Restore strategy. Set up AWS Backup to take hourly snapshots of the Aurora database and copy them to Amazon S3 Glacier Flexible Retrieval in us-west-2. In us-west-2, keep the ECS Fargate task definition registered. Upon disaster, retrieve the backup from S3 Glacier using standard retrieval, restore the database, and launch the ECS Fargate tasks.
  4. D
    Configure a Pilot Light strategy. Scale the ECS Fargate service in us-west-2 to 00 tasks. To minimize database standby costs, stop the secondary Aurora DB instance in the secondary region. Configure an AWS Lambda function to start the instance during failover, expecting the database to start instantly and serve traffic with a recovery time of under 22 minutes.

Answer

Configure Amazon Aurora Global Database with the primary cluster in us-east-1 and a secondary cluster in us-west-2. In us-west-2, deploy the ECS Fargate service with the desired task count set to 00. Configure Amazon Route 53 failover routing with health checks. During a failover event, promote the secondary Aurora cluster to primary, scale the ECS Fargate task count to the required capacity, and update the DNS records.
The configuration using Amazon Aurora Global Database and scaling ECS Fargate tasks to 00 in the secondary region represents a Pilot Light strategy. Aurora Global Database keeps the primary and secondary databases synchronized via storage-level replication with sub-second replication latency, meeting the 11-minute RPO. Since no Fargate tasks are running in the secondary region during normal operations, standby compute costs are minimized. During a disaster, promoting the Aurora secondary cluster takes less than a minute, and scaling up the ECS Fargate tasks to full capacity takes a few minutes, which easily meets the 1515-minute RTO requirement.

Step-by-Step Solution

1
Analyze RPO requirements.
To meet an RPO of less than 11 minute, database changes must be replicated to the secondary region continuously. Amazon Aurora Global Database uses storage-based replication with a typical latency of less than 11 second, easily meeting this constraint.
Choosing the replication technology first ensures the data loss window (RPO) is minimized.
2
Analyze RTO and standby cost requirements.
To minimize standby running costs, compute resources (ECS Fargate tasks) in the disaster recovery region should be scaled to 00 when idle. Promoting an Aurora Global Database secondary cluster takes under 11 minute, and provisioning ECS Fargate containers takes 22 to 33 minutes. This easily fits within the 1515-minute RTO.
This establishes a Pilot Light strategy, which has minimal running standby costs while still satisfying the RTO.
3
Evaluate failover and DNS routing.
Using Route 53 active-passive failover routing policy allows traffic redirection once health checks fail. During failover, the secondary Aurora cluster is promoted to primary (becoming read-write), and the ECS Fargate tasks are scaled up to match the production workload.
A clean failover orchestration guarantees consistent application state and traffic redirection.

Key Concept

Disaster recovery (DR) strategies on AWS involve trading off cost (standby resources) against Recovery Time Objective (RTO) and Recovery Point Objective (RPO). For RPO < 11 minute and RTO < 1515 minutes with minimized costs, a Pilot Light strategy using Aurora Global Database (with secondary cluster running a minimal reader) and scaled-down compute (ECS Fargate task count set to 00) is the most optimal architecture.
Rate this question