DESIGN OF AN ITERATIVE PREDICTIVE CARBON-AWARE RESOURCE SCHEDULING METHOD FOR CLOUD–FOG–EDGE ECOSYSTEMS THROUGH MULTI-STAGE ENERGY AND THERMAL OPTIMIZATIONS

Abstract

Growing computational demand across cloud, fog, and edge infrastructures is intensifying energy consumption and carbon emissions, yet existing scheduling frameworks typically treat power draw as a static, short-term metric. This narrow view struggles with the fluid geography of modern workloads bursty, migratory, and thermally entangled leading to inefficient energy use and weak carbon accountability. To confront these gaps, we introduce a five-stage resource–energy orchestration model that explicitly intertwines spatio-temporal energy prediction, quantum Inspired optimization, live task migration, thermal dynamics, and carbon-economic feedback. The pipeline begins with Multi-Modal Spatio-Temporal Energy Profiler (MSTEP), which continuously profiles heterogeneous nodes through tensor decomposition and graph-temporal convolution, forecasting per-node energy use over 5-second horizons and improving power-capping accuracy by about 12%. Its predictive map drives the Energy-Aware Quantum Inspired Resource Orchestrator (EQUIRO), a classical Hamiltonian optimizer borrowing from quantum annealing to escape local minima, cutting energy consumption by roughly 18% while lowering latency. The resulting plan is enacted by Reinforcement-Driven Adaptive Task Migrator (R-ATM), where policy-gradient agents treat migration as a continuous-time control problem, reducing unnecessary moves by ~20%. To prevent thermal hotspots created by such migrations, Cross-Layer Thermal-Aware Cooling Optimizer (CLTACO) applies physics Informed neural networks to couple micro-scale thermal diffusion with macro cooling strategy, yielding around 15% better power usage effectiveness. Finally, Carbon Impact Feedback and Economic Optimizer (CIFEO) close the loop by translating operational data into carbon-weighted pricing and deferral schedules, achieving near-neutral or positive margins with an estimated 25% cut in carbon footprint. This integrated architecture demonstrates how predictive, cross-layer intelligence can transform cloud-fog-edge scheduling from reactive energy management into proactive carbon-aware economics, pointing toward greener and more economically resilient distributed computing.