Carbon-Aware Scheduling
A technical approach that dynamically schedules computing workloads according to grid carbon intensity signals. Deferrable tasks are placed in time windows with surplus clean electricity and in low-carbon regions, thereby reducing the carbon emissions of computing.
Timeline (last 90 days)
Carbon-aware scheduling is especially feasible for batch tasks (training, offline inference, data preprocessing), which can be delayed by several hours to wait for peaks in wind and solar output.
Latency-sensitive AI services are difficult to subject to carbon-aware scheduling, because users will not accept extra latency in exchange for waiting for clean power.
Carbon-aware scheduling requires hourly or even five-minute-level grid signals, not annual averages.
All Facts (3)
Carbon-aware scheduling is especially feasible for batch tasks (training, offline inference, data preprocessing), which can be delayed by several hours to wait for peaks in wind and solar output.
50%UnverifiedLatency-sensitive AI services are difficult to subject to carbon-aware scheduling, because users will not accept extra latency in exchange for waiting for clean power.
50%UnverifiedCarbon-aware scheduling requires hourly or even five-minute-level grid signals, not annual averages.
50%