CarbonScaler: Leveraging Cloud Workload Elasticity for Optimizing Carbon-Efficiency

Due to inherent variations in energy's carbon intensity, temporal shifting has become a key method in reducing the carbon footprint of batch workloads. However, temporally shifting workloads involves searching for periods with lower carbon intensity, which increases the workload's completion time. In this paper, we present CarbonScaler, a new approach that reduces carbon emissions of batch workloads without extending their completion time. Our approach relies on applications' ability to change their compute demand by scaling the workload based on fluctuations in energy's carbon intensity. We present a carbon-aware scheduling algorithm, a Kubernetes-based prototype, and an analytic tool to guide the carbon-efficient deployment of batch applications in the cloud. We evaluate CarbonScaler using real-world applications and show that it can yield 33% carbon savings without extending the completion time and up to 32% extra carbon savings over state-of-the-art suspend-resume policies when completion time is flexible.