GA-CGD: A Global Energy-Aware Container Placement Strategy for Heterogeneous Data Centers

Data centers are continuously expanding due to AI workload growth, leading to increasingly prominent high-energy-consumption issues. In heterogeneous data centers, the diversity of IT and cooling equipment complicates holistic energy optimization. This paper proposes the Genetic Algorithm with Controlled Gene Diversity (GA-CGD), a container placement strategy that jointly optimizes computing and cooling energy. Unlike prior virtual–machine-focused approaches, GA-CGD introduces a Controlled Gene Diversity mechanism within a genetic algorithm to balance server consolidation and thermal-aware workload distribution. A probabilistic allocation range strategy reduces the active server count, while Dynamic Voltage and Frequency Scaling (DVFS) further lowers computing power. Experimental results show that GA-CGD reduces the total data center energy by 16.93% and 12.83% compared with SABA (a simulated-annealing-based, heat-recirculation-aware placement method) and TSTD (a thermal-aware and DVFS-enabled task scheduling method), respectively.