K8s Cost Optimizer / 容器成本优化专家v1.0.2

A platform team is asked by leadership to cut 30% off the monthly cloud bill before the next board meeting. They feed the agent 30 days of metrics-server data, the full node list, and AWS Cost Explorer exports to get a ranked set of dollar-quantified recommendations. The output becomes the agenda for the FinOps review with clear owners and ready-to-apply patches.

A company runs all workloads on expensive on-demand instances and wants a safe path to spot/preemptible capacity. The agent identifies which workloads are interruption-tolerant, models savings per instance family, and produces a phased four-week migration plan with node pool YAML. It also flags workloads that must stay on on-demand due to statefulness or SLAs.

Before signing a one- or three-year committed-use discount, a FinOps engineer wants to ensure they are not locking in waste. The agent analyzes p99 CPU and memory usage against requests, recommends right-sized requests, and recalculates the correct commitment level. This prevents over-committing to capacity that should be eliminated first.

An internal developer platform hosts 14 namespaces, several of which belong to deprecated projects. The agent detects idle namespaces by p95 CPU and memory thresholds, finds LoadBalancer services with no traffic for 30 days, and produces an archive plan. The team recovers capacity without a risky big-bang cleanup.

A retail platform sees Kubernetes costs spike before Black Friday due to overprovisioned requests and missing autoscaling. The agent identifies missing HPAs, suggests HPA target settings based on historical usage, and recommends right-sized requests so the cluster scales efficiently during peak. The result is predictable peak-season spend instead of emergency budget overruns.

A consulting or managed-service business that uses this skill to audit customer clusters and deliver ranked savings reports on a recurring basis. Pricing can be a percentage of realized savings or a fixed monthly retainer. It scales by serving many clusters with a repeatable, data-driven methodology.

A cloud cost management platform embeds this skill as a Kubernetes-specific optimization module inside its product. Customers connect their clusters and receive continuous recommendations with YAML patches and savings forecasts. It differentiates the platform from generic cloud cost tools by covering Kubernetes-native objects.

An enterprise platform team uses the skill internally to continuously reduce infrastructure spend and enforce cost-aware configuration standards. The savings directly improve gross margin and free budget for new initiatives. It also supports chargeback or showback models to engineering teams.