Under the background of big service

more and more resources are released and utilized in the form of services. Meanwhile

the users' requirements grow in complexity

which leads to the exponential growth of the service composition calculation scale. In this paper

a task-granulating algorithm called TgA is proposed to solve large-scale service composition optimization problem quickly and effectively. Firstly

we build a hierarchical service composition model based on task granulation

and analyze its computational complexity. Secondly

we analyze the feasibility theoretically during the task-granulating according to the calculation of existing QoS attributes. Finally

a large number of simulation experimental results show that

compared to the classical particle swarm optimization (PSO) algorithm

the proposed algorithm can improve the service composition optimization performance by 4 to 7 times and increase the optimization accuracy by more than 10%.