Abstract: In order to solve the traditional constellation system where satellite controllable times are short, no on-orbit scheduling capabilities, real-time processing capabilities are poor, and mission execution benefits are low, a constellation-oriented cloud computing system (C3S, constellation cloud computing system) is proposed. The system uses the characteristics of high and low orbit satellites to design the architecture, realize the real-time controllability of multiple satellites, uses intelligent scheduling strategy, realize the effective on-orbit scheduling of multi-satellite cooperative earth observation tasks and multi-satellite cooperative target detection tasks. for multi-satellite cooperative earth observation tasks, through heuristic initial solution generation strategy, roulette selection algorithm based on Boltzmann's selection probability, adaptive crossover operator and mutation operator based on the degree of population dispersion, stopping criterion based on population convergence coefficient, to improve the genetic algorithm. for the multi-satellite cooperative target detection task, a two-stage scheduling strategy considering the data transmission time and real-time inter-satellite distance is designed. Finally, on the ground demonstration and verification system formed by the customized development of kubernetes, the rationality of the architecture design and the effectiveness of the on-orbit scheduling algorithm were verified, compared with the traditional constellation management and control mode, the average controllable time of the low-orbit constellation node within 24 h is increased by 17.7 h, the average time delay of the multi-satellite cooperative target detection task is reduced by 50%, and the average benefits of the multi-satellite cooperative earth observation task is increased by 2.58 %.