Question

A retail company runs a customer-facing web application on an Amazon EC2 Auto Scaling group behind an Application Load Balancer. Traffic follows a highly predictable daily pattern: it ramps up sharply every morning at 8:00 AM as customers log in. The team currently uses target tracking scaling based on average CPU utilization, but they observe elevated latency and HTTP 503 errors for the first 10-15 minutes each morning because instances cannot launch and warm up fast enough to meet the sudden demand. Application startup (bootstrapping and cache warming) takes about 6 minutes per instance. The team wants to improve reliability without permanently over-provisioning. What is the MOST effective way to address this?

Accepted Answer

Enable predictive scaling on the Auto Scaling group in addition to the existing target tracking policy, and configure an instance warm-up period that accounts for the 6-minute startup time.

Answer

Lower the target tracking CPU threshold to 30% so that new instances launch earlier when demand begins to rise each morning.

Answer

Replace the Auto Scaling group with a fixed desired capacity sized for peak morning load to eliminate scaling delays entirely.

Answer

Add a step scaling policy triggered by ALB request count so that instances scale based on incoming requests instead of CPU.