Manufacturing service composition (MSC) is an essential issue in cloud manufacturing, which streamlines complex manufacturing tasks into manageable subtasks and integrates distributed services to enhance task completion. Existing studies allocate services for subtasks with maximizing quality of service (QoS) simultaneously, assuming that all subtasks are of equal importance. However, different subtasks hold varied significance and priorities. One rational method is to prioritize the allocation of premium or scarce services to important subtasks. Therefore, this study proposes a two-phase, subtask priority-based approach for the hierarchical allocation of the MSC. The initial phase applies a multi-attribute decision-making method based on complex networks, the Enhanced Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS-EK), to assess subtask importance. The TOPSIS-EK method ascertains subtask importance, delineating the subtasks into Key Manufacturing Subtasks (KMTs) and Ordinary Manufacturing Subtasks (OMTs). The second phase uses bilevel optimization for the hierarchical allocation of the MSC to KMTs and OMTs, respectively. A hybrid Particle Swarm Optimization and Genetic Algorithm with Chaos-sequence and Inheritance (PSOGA-CI) is developed to solve the model. The proposed approach is validated with a case on the production of an airplane engine turbine rotor blade.