Load Shifting Prediction Types:
Demand Shifting: Carbon aware workload shifting which finds the greenest start time across all data centers permitted in the search. This method provides the best carbon reduction results out of the three shifting types.
Geographic Shifting: Find a new data center for compute instances for cross region shifting to reduce the resource's carbon footprint. This shifting type assumes an immediate start time for the compute activity.
Time Shifting: Search for the greenest starting time in at a specified data center.
- Run Duration is the expected duration of the compute. Picking a window size will return the start-time with the lowest average emissions rate for the given window size. Without a window size selected, it will return the single time of minimum emissions instead.
- GPU SKU is the desired SKU to be used for the compute. This limits regions in geographic shifting to data centers that can offer the selection.
- Protected Data implies that the workspace contains data subject to international migration and privacy laws. Selecting this option limits any potenial workload shift to data centers that would be compliant and not violate laws.