Question

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

A logistics telemetry company hosts its fleet tracking application in the us-east-1 Region. The architecture consists of Amazon EC2 instances in an Auto Scaling Group behind an Application Load Balancer, and a primary Amazon RDS for PostgreSQL database instance. The company needs to design a disaster recovery (DR) plan in the us-west-2 Region. The DR plan must support a Recovery Time Objective (RTO) of 1515 minutes, a Recovery Point Objective (RPO) of 55 minutes, and minimize ongoing infrastructure costs. Which DR strategy meets these requirements?

  1. A
    Establish a Pilot Light recovery environment in us-west-2. Configure the RDS PostgreSQL database in us-east-1 as a Multi-AZ deployment with the standby replica located in us-west-2 to ensure automatic, zero-downtime cross-region database failover. Deploy the application tier with stopped EC2 instances in us-west-2.
  2. B
    Establish a Warm Standby recovery environment in us-west-2. Configure an RDS cross-region read replica in us-west-2. Run the EC2 Auto Scaling Group at minimum capacity in us-west-2. Configure Amazon Route 53 with a latency routing policy to automatically redirect all user traffic to us-west-2 if the us-east-1 region becomes unavailable.
  3. Establish a Pilot Light recovery environment in us-west-2. Configure an RDS cross-region read replica in us-west-2. Deploy an Elastic Load Balancer and an Auto Scaling Group with a desired capacity of 00. During a disaster, promote the read replica to a standalone database and scale the Auto Scaling Group to the required production capacity.Answer
  4. D
    Establish a Backup and Restore recovery environment in us-west-2. Schedule hourly RDS snapshots and copy them to an Amazon S3 bucket in us-west-2, using S3 Glacier Flexible Retrieval for cost optimization. In a disaster, restore the database from Glacier and launch EC2 instances from AMIs.

Answer

Establish a Pilot Light recovery environment in us-west-2 by configuring an RDS cross-region read replica and an Auto Scaling Group with a desired capacity of 00, then promoting the replica and scaling the Auto Scaling Group during a disaster.
The correct strategy uses Pilot Light. An Amazon RDS cross-region read replica provides asynchronous replication with very low lag (usually seconds), easily meeting the 55-minute RPO. In the event of a disaster, the replica is promoted to a primary database, and the Auto Scaling Group is scaled up from 00 to the required capacity. This sequence completes within the 1515-minute RTO. Because no EC2 instances run in the recovery region during normal operation, this approach minimizes ongoing costs.

Step-by-Step Solution

1
Evaluate the RTO and RPO requirements against the DR strategies.
An RPO of 55 minutes requires active database replication. S3 backups copied hourly (with Glacier retrieval times) fail the RPO. An RTO of 1515 minutes allows for either Warm Standby or Pilot Light, as both can scale up or promote databases within this window.
Eliminates Backup and Restore options which cannot meet the RPO of 55 minutes or the RTO of 1515 minutes.
2
Evaluate the database replication capability across AWS regions.
RDS Multi-AZ is a single-region high-availability feature and cannot place a standby replica in a different AWS region.
Eliminates the option proposing cross-region Multi-AZ standby replication.
3
Compare Pilot Light and Warm Standby based on the cost minimization constraint.
Pilot Light keeps the application tier shutdown (desired capacity of 00 for ASG), which incurs zero EC2 compute costs during idle periods. Warm Standby keeps a minimum number of EC2 instances running, incurring continuous compute costs. Route 53 latency routing also fails to provide failover without health checks.
Identifies the Pilot Light option as the most cost-effective solution that meets all constraints.

Key Concept

Disaster Recovery (DR) strategies using Pilot Light with cross-region read replicas to meet tight RTO/RPO targets while minimizing idle costs.
Estimated Time:2m 0s
Rate this question