Target Tracking vs Step Scaling Policies

Target tracking scaling is the simpler and more commonly used dynamic scaling type: you declare a target value for a chosen metric (like 'keep average CPU at 50%'), and AWS's own scaling algorithm calculates and applies the needed capacity adjustments automatically, functioning similarly to a thermostat maintaining a set temperature. Step scaling gives more granular, explicit control: you define specific capacity change amounts tied to specific metric threshold ranges (for example, add 2 instances if CPU is between 70-90%, add 5 instances if CPU exceeds 90%), which is useful when different severities of load genuinely warrant different-sized responses rather than a smooth, algorithmically-calculated adjustment. Most workloads are well served by target tracking's simplicity; step scaling is chosen specifically when a team has enough operational history with a workload to know that certain load thresholds warrant disproportionately larger capacity jumps than a target-tracking algorithm would naturally apply.