Electricity and Carbon-aware Task Scheduling in Geo-distributed Internet Data Centers
Peng Wang, Wenyu Liu, Ming Cheng, Zhaohao Ding, Yi Wang · 2022 IEEE/IAS Industrial and Commercial Power System Asia (I&CPS Asia) · 2022
Besides the enormous power consumption, carbon emission from data centers is also a concern for cloud service providers (CSPs). Based on cloud computing technology, the spatial-temporal transfer of workloads between geographically distributed internet data centers (IDCs) gives CSPs the potential to reduce their carbon footprint. However, workloads in real-world production are highly heterogeneous, which vary in resource demand, duration et al. Without ignoring the workload heterogeneity, we proposed an integer linear programming formulation of the task scheduling problem for geo-distributed IDCs. Extensive numerical studies are performed using real task information, electricity price, and marginal emission rate. Results show that CSP can better meet emission reduction demand and optimize its electricity expenses by spatial-temporal task scheduling.