Spatiotemporal Carbon-aware Scheduling in the Cloud: Limits and Benefits

Abstract

As the demand for computing continues to grow exponentially and datacenters are already highly optimized, many have suggested leveraging computing workload's spatiotemporal flexibility. However, different workloads may have different degrees of flexibility, including execution deadlines, data protection laws, or latency requirements. These constraints, along with many others, limit the potential benefits of carbon-aware spatiotemporal workload shifting; the achievable benefits of these approaches are unclear-an aspect not addressed by prior research. Accurately quantifying the achievable benefits of carbon-aware spatiotemporal workload scheduling is critically important, as many in research and industry are already devoting significant time and resources to realize these benefits. To address the problem, we conduct a large-scale longitudinal analysis of carbon-aware spatiotemporal workload shifting to answer the following research question: What are the maximum carbon emission reductions that can be achieved due to temporal and spatial workload shifting for different types of cloud workloads and in different parts of the world?