Adaptive Multi-Objective Firefly Optimization for Energy-Efficient and QoS-Aware Scheduling in Distributed Green Data Centers

Green data centers (GDCs) are increasingly deployed worldwide to power digital infrastructure sustainably. These centers integrate renewable energy sources, such as solar and wind, to reduce dependence on grid electricity and lower operational costs. When distributed geographically, GDCs face considerable challenges due to spatial variations in renewable energy availability, electricity pricing, and bandwidth costs. This paper addresses the joint optimization of operational cost and service quality for delay-sensitive applications scheduled across distributed green data centers (GDDCs). We formulate a multi-objective optimization problem that minimizes total operational costs while reducing the Average Task Loss Probability (ATLP), a key Quality of Service (QoS) metric. To solve this, we propose an Adaptive Firefly-Based Bi-Objective Optimization (AFBO) algorithm that introduces multiple adaptive mechanisms to improve convergence and diversity. The minimum Manhattan distance method is adopted to select a representative knee solution from each algorithm’s Pareto front, determining optimal task service rates and ISP task splits into each time slot. AFBO is evaluated using real-world trace-driven simulations and compared against benchmark multi-objective algorithms, including multi-objective particle swarm optimization (MOPSO), simulated annealing-based bi-objective differential evolution (SBDE), and the baseline Multi-Objective Firefly Algorithm (MOFA). The results show that AFBO achieves up to 64-fold reductions in operational cost and produces an extremely low ATLP value ( 1.875 × 10 − 7 ) that is nearly two orders of magnitude lower than SBDE and MOFA and several orders better than MOPSO. These findings confirm AFBO’s superior capability to balance energy cost savings and Quality of Service (QoS), outperforming existing methods in both solution quality and convergence speed.