A dynamic pricing model for carbon-aware compute clusters

The rising demand for compute services has led to a proliferation of data centres (computer clusters), resulting in a significant environmental footprint as over 1.5% of the global power supply is consumed. In this work, we propose an incentive-compatible pricing mechanism for cloud computing providers that automatically promotes customers to submit jobs in a carbon-aware fashion. To demonstrate the efficacy of the mechanism, we have developed a power model and a job submission model to simulate the operation of compute clusters. We also employ auction theory to incorporate cluster parameterisation into a pricing mechanism. With this mechanism, strong utilisation of renewable energy and high utility for customers can be achieved by appropriately allocating high/low-priority jobs, both for synthetic as well as real-life workloads. Our work provides a possible solution, with environmental and financial benefits, for cloud computing providers who seek to minimise their carbon footprint while gaining a competitive advantage by passing the energy savings to their customers.