Question

A financial services company is designing a new customer-facing web portal that will run as a containerized application. The application maintains in-memory user session state and must remain available even if an entire Availability Zone fails. Traffic is highly variable, spiking to 10x baseline during quarterly statement releases, and the company wants to minimize idle capacity costs while ensuring sub-second response times. The architecture team wants to avoid managing EC2 instances or patching operating systems. Which design BEST meets these requirements?

Accepted Answer

Run the containers on AWS Fargate behind an Application Load Balancer across multiple Availability Zones, store session state in an Amazon ElastiCache for Redis cluster with Multi-AZ enabled, and configure Application Auto Scaling on the ECS service based on request count.

Answer

Run the containers on an EC2-backed ECS cluster spread across Availability Zones, enable ALB sticky sessions to keep users pinned to a single task, and use an EC2 Auto Scaling group with scheduled scaling for the quarterly spikes.

Answer

Deploy the containers on AWS Fargate with an ALB, enable ALB session stickiness so in-memory session state stays on the same task, and scale the ECS service using a target-tracking policy on CPU utilization.

Answer

Run the containers on AWS App Runner with automatic scaling, and store session state in an Amazon DynamoDB table with on-demand capacity mode for durability across Availability Zones.