Stabilizing and Extending a Misaligned Compute Environment

Table of Contents

1. Challenge

   Diagnose sources of operational instability and inefficiency

   1. Review workload, capacity, scaling, and alarm metrics
   2. Analyze Auto Scaling activity and alarm history
   3. Identify which control surfaces contribute to instability, unnecessary scaling activity, or operational noise
2. Challenge

   Apply targeted corrections to scaling and monitoring controls

   1. Modify the Auto Scaling configuration so capacity follows workload demand on the Application Load Balancer. Scaling must be driven by the RequestCountPerTarget metric in the AWS/ApplicationELB namespace.
   2. Create CloudWatch alarms on RequestCountPerTarget (namespace AWS/ApplicationELB) with two scaling actions:
      • A scale-out action whose name contains Request Scale Out, triggered by high request load per target.
      • A scale-in action whose name contains Request Scale In, triggered by low request load per target.
   3. Improve cooldown and threshold behavior so scaling reacts to sustained demand rather than short spikes, avoiding rapid scale-out and scale-in cycles.
   4. Simplify monitoring by removing or consolidating low-value alarms that duplicate or conflict with the request-based scaling alarms.
3. Challenge

   Validate improved system behavior

   1. Generate controlled workload against the environment
   2. Observe resulting scaling activity and monitoring behavior
   3. Confirm that the environment responds predictably with reduced instability and improved signal quality