To solve the cloud service composition optimal-selection (CSCOS) problem in cloud manufacturing (CMfg), based on the deep analysis of the difficulties and shortcomings in current researches, an optimization model for multi-task scheduling with limited resources problem (MLSR) is proposed. Considering the interest conflicts of service demanders and important quality of service (QoS) indicators, the multi-task scheduling problem with limited resources is converted into multiple static non-cooperative game, and the service demander (SD), different execution manufacturing path of each task and comprehensive service level (CSL) are considered as game player, game strategy and game payoff of a game, respectively. On this basis, the process to seek the pure strategy Nash Equilibrium (PSNE) solution is compiled into an algorithm, whose solutions are the final manufacturing execution paths for each task. Finally, simulation results show the feasibility, effectiveness of both the model and algorithm.