GreenColo: A novel incentive mechanism for minimizing carbon footprint in colocation data center

As an integral part of our digital economy, data centers are growing rapidly, devouring a formidable amount of energy and leaving a huge carbon footprint. While owner-operated data centers (e.g., Google) can implement various power management techniques to reduce energy consumption, colocation data centers, which offer a "halfway" solution to users who do not want to build their own data centers or completely resort to public clouds, suffer from "split incentive" that creates a barrier for greenness: colocation operator desires greenness but does not have control over tenants' servers; tenants who own the servers may not be willing to manage their servers for greenness unless they are properly incentivized. In this paper, we aim at minimizing the carbon footprint of a colocation data center while satisfying the colocation operator's long-term budget constraint. To break the split-incentive barrier and satisfy the budget constraint, we develop a dynamic incentive framework, called GreenColo, in which tenants can voluntarily submit energy reduction bids along with their desired payment and, if accepted, will be financially rewarded for energy reduction. GreenColo can be implemented online based on the currently available information (e.g., tenants' bids and current carbon efficiency) and dynamically select winning bids to minimize carbon footprint. We demonstrate the effectiveness of GreenColo both analytically and empirically. Our trace-based simulation results show that GreenColo can reduce carbon footprint by 18%, while the colocation operator does not incur any additional cost (compared to the no-incentive baseline case) and tenants may save up to 25% of their colocation costs.