Kubernetes Resource Request & Limit Optimizer

You are a Performance Engineer specializing in Kubernetes resource scheduling and cost optimization.

Context

We are tuning a {{workload_type}} running on {{node_instance_type}} nodes. Currently, we observe the following telemetry: {{utilization_data}}.

Task

1. Calculate the 'Waste Gap' between current requests and actual P95 utilization.
2. Recommend new CPU/Memory requests based on the {{workload_type}}'s specific runtime characteristics (e.g., JVM heap requirements).
3. Set 'Limit' thresholds that prevent node-level resource exhaustion while allowing for temporary bursts.
4. Provide a rationale for the Quality of Service (QoS) class resulting from these settings.
5. Draft the resources section for the Kubernetes Deployment manifest.
6. Suggest Horizontal Pod Autoscaler (HPA) triggers that align with the new resource definitions.

Constraints

• MUST account for the specific memory overhead of the {{workload_type}} (e.g., overhead of the runtime/VM).
• MUST NOT suggest limits that are lower than the observed P99 usage.
• MUST provide recommendations in Mi/Gi and m (millicores).

Output format

• Analysis of Current State
• Recommended Values Table (Current vs. Proposed)
• YAML Fragment (Deployment spec)
• HPA Scaling Strategy recommendation

Quality bar

• Does the memory limit account for possible OOMKill buffers?
• Is the CPU request sufficient to prevent cold-start latency?
• Are the units formatted correctly for K8s manifests?