Scheduling workloads in a network of datacentres to reduce electricity cost and carbon footprint

• We optimize job scheduling in a cloud computing operation.
• The objective function includes both carbon and electricity costs.
• We use empirical data from California, Alberta and Ontario.
• Carbon and electricity costs can be simultaneously reduced by 13.14% and 51.44% respectively.
• We show how the cloud operator can sell carbon credits.

This paper quantifies the extent to which the scheduling of workloads among a network of datacenters can reduce both electricity cost and carbon footprint. Based upon empirical data from California, Alberta and Ontario, it develops an optimization model that quantifies the savings in relation to the price of carbon on carbon markets and in carbon taxes. Combining the electricity cost with the carbon footprint using the price of carbon, results indicate a simultaneous saving of both 8.09% of electricity cost and 11.25% of carbon footprint, when jobs are scheduled in the current time-period. When jobs can be scheduled in future time-periods, a simultaneous saving of both 51.44% of electricity cost and 13.14% of carbon footprint was obtained. These results are shown to be robust with respect to variations in the price of carbon in taxes and markets in the European Emissions Trading System, Australia, British Columbia, California, and Japan, apart from exceptional periods when the carbon price was very low. The paper shows how a cloud operator can demonstrate that these savings are “additional” to business as usual so as to sell carbon credits on a carbon market, and indicates the standards available for certifying and auditing those emissions reductions.

Figure optionsDownload as PowerPoint slide