An online incentive mechanism for emergency demand response in geo-distributed colocation data centers

Abstract

Deferring batch workload in data centers is promising for demand response to enhance the efficiency and reliability of a power grid. Yet operators of multi-tenant colocation data centers still resort to eco-unfriendly diesel generators for demand response, because tenants lack incentives to defer their workloads. This work proposes an online auction mechanism for emergency demand response (EDR) in geo-distributed colocation data centers, which incentivizes tenants to delay and shuffle their workload across multiple data centers by providing monetary rewards. The mechanism, called BatchEDR, decides the tenants' workload deferment/reduction and diesel usage in each data center upon receiving an EDR signal, for cost minimization throughout the entire EDR event, considering that only a limited amount of batch workloads can be deferred throughout EDR as well as across multiple data centers. Without future information, BatchEDR achieves a good competitive ratio compared to an omniscient offline optimal algorithm, while ensuring truthfulness and individual rationality over the auction process. Trace-driven experiments show that BatchEDR outperforms the existing mechanisms and achieves good social cost.