Kubernetes
Map pods, jobs, namespaces, nodes, and scheduling context to GPU energy.
Overview
Kubernetes gives Matcha the cluster context needed to connect GPU telemetry with actual workloads. By reading pod, namespace, job, and node metadata, Matcha can estimate which workloads were active during specific GPU energy windows.
Configuration Steps
Install the Matcha agent as a Kubernetes DaemonSet.
Grant read-only access to pod, node, and namespace metadata.
Map GPU devices to Kubernetes workloads.
Enable workload labels for tenants, models, or teams.
Confirm attribution views in Matcha.
