A stateless service has variable demand and should scale around average CPU utilization of 55%. The team does not want to maintain several manual threshold steps. Which policy is appropriate?

1. A target-tracking policy on a suitable average CPU utilization metric with a target of 55%.
2. A dashboard with no scaling action or capacity controller.
3. A fixed capacity change once each year without observing demand.
4. A policy using cumulative lifetime CPU time as if it were average utilization.