Batch job scheduling for Reducing Water footprints in data center

The number and scale of data centers explode with the dramatically surging demand for cloud computing services, raising serious sustainability concerns. While reducing energy consumption is undoubtedly essential for sustainability, the enormity of water consumption (both directly by cooling systems and indirectly by electricity generation) has been long-neglected despite its emergence as a critical consideration in light of the growing conflicts between water supply and demand. In this paper, we take the first step towards the data center water sustainability. Specifically, we exploit the temporal diversity of data center water efficiency and propose an online algorithm, called BREW (Batch job scheduling for REducing Water footprint), which dynamically schedules batch jobs to minimize the water-power cost (quantified in terms of a weighted sum of water consumption and electricity cost) while bounding the maximum job queue backlog. It is formally proved that BREW achieves a close-to-minimum water-power cost compared to the optimal offline algorithm with future information. We also perform a trace-based simulation study and the result validates our analysis: compared to the state-of-the-art solution that solely minimize the electricity cost, BREW can significantly reduce the water-power cost while incurring a negligible delay increase (e.g., over 24% water-power cost saving but only 30 minute delay increase).