The execution time of shortest path computing determines the efficiency and quality of dynamic traffic assignment. This paper focuses on the design of a parallel calculation system for the shortest path in traffic network aims to reduce the execution time of shortest path computing in dynamic traffic assignment. Here we mainly focus on the process of algorithm parallelization and network decomposition. As the core of the system, a two-queue parallel algorithm is designed for the shortest path in the traffic network which the recursive spectral bisection decomposition approach is employed to separate the network into several parts and the algorithm acts on every sub-network on each processor. The proposed system are experimented on real traffic network and a set of policies are employed that the number of processors is designed in each policy in the study case, and the performances of the parallel calculation system are discussed, the proposed algorithm is proved to be the efficient and effective.