Carbon Aware KEDA Operator

Overview

The Carbon Aware KEDA Operator is a Kubernetes operator that aims to reduce carbon emissions by helping KEDA scale Kubernetes workloads based on carbon intensity. It can use carbon intensity data from third-party sources such as WattTime, Electricity Map or any other provider, needs no changes to application or workload code, and works with any KEDA scaler. An administrator creates a CarbonAwareKedaScaler custom resource for a scaledObject or scaledJob, and the operator updates the maxReplicaCount field according to current carbon intensity, defining a ceiling for allowed replicas rather than computing a desired replicaCount, which stays with KEDA and the Horizontal Pod Autoscaler. Typical uses are low-priority, time-flexible workloads such as non-critical backups, batch processing, analytics and ML training.

Editorial verdict

Fits Kubernetes platform teams already running KEDA that want a low-risk way to make batch or ML workloads carbon-aware without touching application code: a custom resource caps replica counts by grid intensity and leaves desired replicas to KEDA and the autoscaler. The trade-off is that it sets a ceiling rather than scheduling to the cleanest hour, the effect depends on the configured intensity data source, and it only helps workloads that can flex. No measurement is included, so savings need separate monitoring. Teams without KEDA, or with latency-sensitive workloads, should not start here.

Written by the GreenAIX editors from the recorded fields above. Vendors do not review or approve verdicts.

Use cases

• Throttle batch and ML training jobs when grid carbon intensity is high
• Cap KEDA replica counts based on carbon intensity thresholds
• Add carbon awareness to existing KEDA workloads without code changes
• Schedule flexible dev and test workloads around cleaner grid periods

Sources

• Carbon Aware KEDA Operator GitHub repository
• Carbon Aware KEDA Operator README
• Carbon Aware KEDA Operator MIT licence