Online Low-Carbon Workload, Energy, and Temperature Management of Distributed Data Centers

TL;DR

This work tackles low-carbon management for geo-distributed data centers under uncertainty by introducing a Lyapunov-optimization-based online coordination that jointly optimizes workload, energy, and temperature without requiring predictive models. A parametric online algorithm with virtual queues solves a per-slot LP to make real-time decisions, while an LP-based procedure tunes the horizon-wide parameters $V$ and $\theta$ to bound the optimality gap $F^l(V,\theta) - F^m$ and ensure feasibility under an emission cap $C^E$. The approach explicitly accounts for cooling-driven energy, carbon intensity, and price uncertainty, achieving emission-constrained operation with competitive costs, as demonstrated in case studies against offline and greedy baselines. The method offers robust, scalable, and prediction-free operation for data centers, enabling practical deployment of low-carbon, temperature-aware workload management in distributed settings.

Abstract

Data centers have become one of the major energy consumers, making their low-carbon operations critical to achieving global carbon neutrality. Although distributed data centers have the potential to reduce costs and emissions through cooperation, they are facing challenges due to uncertainties. This paper proposes an online approach to co-optimize the workload, energy, and temperature strategies across distributed data centers, targeting minimal total cost, controlled carbon emissions, and adherence to operational constraints. Lyapunov optimization technique is adopted to derive a parametric real-time strategy that accommodates uncertainties in workload demands, ambient temperature, electricity prices, and carbon intensities, without requiring prior knowledge of their distributions. A theoretical upper bound for the optimality gap is derived, based on which a linear programming problem is proposed to optimize the strategy parameters, enhancing performance while ensuring operational constraints. Case studies and method comparison validate the proposed method's effectiveness in reducing costs and carbon emissions.

Figures (5)

• Figure 1: The structure of the distributed data center system.
• Figure 2: The value of $Q^E$ in the iteration process.
• Figure 3: The simulated queues $q^F$, $q^B$, $e^S$, and $\tau^H$ in time slots 5,500--5,600. The bounds are represented by red dashed lines. All the queues are within the bounds.
• Figure 4: Test results under different $Q^E$.
• Figure 5: Test results under different $C^E$.

Theorems & Definitions (4)

• Lemma 1
• Lemma 2
• Theorem 1
• Theorem 2