Resiliency in the sustainability of Distributed Green Data Centers

This paper presents a study of the effect of workload distribution on the resiliency of a new paradigm in sustainable data centers which considers small modular units, called “Distributed Green Data Centers” (DGDC), that are distributed over a geographical region and primarily powered by local energy sources, particularly but not limited to renewables. Considering as resiliency metric the probability of system unavailability, it is seen in this work how workload distribution increases resiliency even in challenging scenarios where the DGDCs experience similar patterns of solar energy availability because they are all located in the same geographical area. While workload distribution has a positive effect in increasing the resiliency, it also increases the data center mean response time. Nevertheless, simulation results show that while keeping the relative increase in response time within acceptable ranges, the probability of system unavailability is reduced in relative terms by 11% and 7% for, respectively, the cases of diverse and similar solar energy availability patterns at the DGDCs.