Toward resilient green cloud computing: Joint operations of energy storage and spatial task allocation

In recent years, the surge in advanced internet computing workloads in data centers, has intensified the challenge of ensuring energy efficiency while maintaining stable and resilient computational services. In practice, widely deployed integrated centralized work-scheduling and energy-management systems are designed to optimize power distribution, renewable-energy utilization, and computational efficiency. However, it faces significant sociotechnical challenges that hinder their effective implementation. To address these issues, we integrate reserve regulation and task allocation to further optimize the operations of multiple data centers in a computational resource-sharing platform. We model the problem as a two-stage stochastic integer program. In the first stage, we optimize backup battery capacity for each data center, and in the second, we introduce a non-preemptive M/M/1 queue with task priorities. We then analyze stationary processing times and apply second-order cone programming for improved computational tractability. We develop an outer-approximation algorithm to improve computational efficiency for large-scale problems, while our integrated strategy balances cost efficiency, environmental sustainability, and resilience at minimal costs. A case study demonstrates that the integrated strategy for multiple data centers reduces total costs by 10.39% and 10.88% compared to task allocation and backup-battery reserve regulation strategies, respectively. Task priorities save 3%–4% in operational costs, while the strategy ensures stable operations and stronger resilience during power interruptions. The outer-approximation algorithm outperforms commercial solvers by 30%, and task replication improves renewable-energy utilization and energy efficiency in smaller data centers. These findings highlight the potential of our strategy to enhance data center efficiency, sustainability, and resilience.