SPath: An Energy-Efficient DAG Task Offloading Scheme with Space-Ground Cooperative Computing

The fast improvement of satellites in both computation ability and numbers in the network, makes them possible to form as promising edge computing nodes for space-ground cooperative computing. These satellite edges are especially helpful to cover IoT terminals in remote areas (e.g. deserts, forests, and oceans) who have both limited computation and battery resource, as well as short time deadlines. However, the mobility and flexibility of satellites, and the dependence between subtasks of terminal DAG (Directed Acyclic Graph) tasks, pose new challenges to the cooperative offloading scenarios. Moreover, the energy supply for both the ground terminals and the edge satellites are very constrained, and are demanded to be optimized for longer services. To solve this problem, we propose SPath, an energy-efficient satellite-ground cooperative offloading scheme, which can reduce the energy consumption and latency of DAG tasks. Moreover, it transforms the offload performance optimization problem into a shortest path problem for a more direct solution. The simulation results with real StarLink satellite movement data show that SPath manages to reduce the energy consumption of DAG tasks by 40.1% and the delay by 39.5% over the exited methods.