Developing a Multi-Agent-Based System with Digital Twin Integration for Efficient Resource Allocation to Reduce Carbon Footprint in Data Centers

This digital world of ours has one consumer primarily: data centers. And they consume energy - $\mathbf{1. 5}$ times the entire global economy - and significantly contribute to carbon emissions and the need for sustainability. We propose a multiagent system for the data center management. This data center management system uses agents monitoring loads of the server, energy consumption, and cooling needs. Tasks are balanced with non-preemptive priority-based scheduling, dynamic load balancing through a genetic algorithm, and adaptive cooling via Q-learning based on real-time data. Coupled with a Digital Twin, a real-time virtual model of the data center, predictive analysis and optimization of resource allocations and cooling strategies improve decision-making and efficiency. Realistic data center traces in simulations indicate 13% energy use reduction, 15% fewer carbon emissions, and a cost-saving of $10 \%$, which can point towards the untapped potential intelligent MAS and digital twin integration holds for the sustainable management of data centers.