Task-Aware Offloading for Cloud Computing In Sky-Based Aerial Data Centers

Data Center-Enabled High Altitude Platforms (DC-HAPs) present a promising solution for reducing terrestrial data centers' energy consumption through natural cooling and solar power harvesting. This paper investigates task-aware computational offloading strategies to further minimize terrestrial data center energy consumption. We first transform the energy consumption minimization objective to computational utilization maximization for HAPs. By modeling both transmission and computation as queuing systems, we derive closed-form expressions for maximum achievable computational utilization under both latency-constrained and unconstrained scenarios for single-type tasks, revealing the critical relationship between task instruction lengths and bit lengths. For multi-type task offloading, we formulate an optimization problem and develop a heuristic sequential algorithm guided by theoretical insights, alongside a novel feasibility-preserving transformation that significantly enhances traditional optimization approaches. Simulations demonstrate that our heuristic algorithm maintains consistent performance regardless of problem complexity, outperforming conventional methods while requiring substantially less computational time. This work provides a foundation for practical DC-HAP deployment in next-generation green computing infrastructure.