Raising the workload concentrations in both cloud and edge infrastructure has compounded the probability of thermal hotspots, which have caused more rack power expenditure, increased hardware depreciation, and increased operational expenses. To overcome this issue, the current paper will suggest a thermal-conscious resource management framework that coordinates prediction, scheduling, and adaptive control into one orchestration strategy. The methodology makes use of hybrid RC-based thermal models alongside on-line learning in order to obtain calibrated temperature forecasts. This prediction helps to make decisions concerning the location of workloads, dynamic voltage and frequency scaling (DVFS), and selective migration to maintain safe thermal operation without impacting performance. Experimental analysis indicates that the framework can cut the peak temperature by a maximum of 12°C, radically decrease the cooling power by half, and cut the workload latency by three-fifths in connection with typical scheduling. And long-term reliability is better 20% after 10 hours, but the cost activity always reduces by 20%. Collectively, this evidence implies that thermal-aware orchestration can substantially improve energy performance, reliability, and sustainability through providing greener and more resilient cloud-edge ecosystems.
Xiaochun Cheng (Sun,) studied this question.