ECS for Highly Available Container Workloads Across Three AZs?

Why the Answer Is Correct

Amazon ECS is a managed container orchestration service that runs tasks across the subnets and Availability Zones configured in a VPC. Setting the service desired count or minimum capacity to 3 with a spread task placement strategy using the Availability Zone attribute distributes tasks across three AZs, so the loss of one AZ still leaves the application running. Target tracking scaling adjusts capacity based on demand without manual tuning, which directly satisfies the least-operational-overhead requirement. Because the application is already containerized, moving it to ECS requires minimal application changes compared with re-platforming to Lambda or managing EC2 instances directly. This combination gives highly available container workloads with the least ongoing operational burden.

Why the Other Options Are Wrong

Option B uses Amazon EKS with self-managed nodes, which requires you to install and maintain Kubernetes node components, manage node groups, and handle more of the cluster lifecycle yourself. Option C uses EC2 Reserved Instances and a spread placement group, which spreads instances across AZs but leaves you managing the guest OS, container runtime, scheduling, patching, and scaling hooks for the application. Option D uses Lambda, which can run container images but is event-driven and not a drop-in replacement for a long-running containerized service; Lambda also does not use ECS-style task placement or a meaningful minimum task capacity of 3. In each wrong option, either the operational overhead is higher or the workload model does not match a continuously running containerized application.

Community Comment Notes

Community learners overwhelmingly picked option A, and the discussion centers on operational overhead. As osmk noted, Amazon EKS self-managed nodes require manually installing and configuring components such as "kubelet, kube-proxy, and Docker" and managing security groups, IAM roles, and subnets, whereas "Amazon ECS handles these tasks for you when you use the Amazon EC2 launch type". Another learner summarized the container-orchestration logic simply with "Containerized.. = ECS". A learner also asked why not Lambda, which highlights the common trap of treating serverless functions as equivalent to a managed container service; Lambda's event-driven model does not provide ECS task spreading across three AZs with a minimum capacity of 3.