Optimal Energy and IT Service Emergency Schedule for Internet Data Center
Zhongqi Zhao, Lei Fan, Zhu Han · 2024
In the era of Artificial intelligence (AI), inter-net data centers (IDCs) play a crucial role in supporting the global information infrastructure. However, large language model (LLM) training consumes a lot of energy, which poses unique challenges to IDC, especially in power emergencies. In order to prevent power outages from causing huge economic losses to IDC, a good energy management system is essential. In this study, we propose a new mixed-integer linear programming (MILP) model of an IDC during an electricity emergency to maximize the profit while minimize operation cost. On power side, we considered renewable energy, traditional thermal power generators, uninterruptible power supply (UPS), heating, ventilation, and air conditioning (HVAC) system. On IT job scheduling side, we allow jobs to be done, held, discarded and transferred to other IDC. Finally, we analyze and visualize the computational results and verify the correctness of the proposed model.