Reducing Datacenter Operational Carbon Emissions Effectively by Cooperating with the Grid

Abstract

Facing growing concerns about power consumption and carbon emissions, cloud providers are adapting datacenter loads to reduce carbon emissions. With datacenters exceeding 100MW, they can affect grid dynamics, so doing this without damaging grid performance is difficult.We study power adaptation algorithms that use grid metrics and seek to reduce datacenter (DC) operational carbon emissions. First, we consider local, online adaptation. Second, to reduce grid disruption that can arise from adaptation, we consider using an external coordinator that limits aggregate impact. Finally, we study a novel cooperative scheme, where datacenters first create a fullday adapted power plan, based on day-ahead information, and then share it with the power grid. This novel approach is a partnership that reflects the shared responsibility between datacenter and the grid for carbon minimization. For each, we report DC carbon emissions reduction and grid impacts.Results show that PlanShare, the novel cooperative scheme is superior to all online approaches considered, achieving a net benefit of 12.6% reduction in datacenter operational carbon emissions. PlanShare achieves 26%-132% better performance than the best online approach, and far larger for many. This benefit arises from allowing the grid optimization with full future DC load information. Further it enables efficient DC resource management with a 24-hour capacity schedule. It also decreases power cost for the entire grid (datacenters and other customers). CCS CONCEPTS• Applied computing → Data centers; • Hardware → Power and energy.