The distributed-memory parallel computer has become the most popular platform for the parallel acceleration of numerical simulations in scientific computing area. However, complex system characteristics and large-scale time-varying data produced on such environment present tremendous challenges for convenient data analyzing. A visualization system framework is introduced in this paper, which can be easily used to visual analyze the remote simulations. It utilizes a hybrid computational coupling strategy adapting to the parallel computing environment flexibly without disturbing the simulation code structures. Experimental results show the proposed framework can provide good interactivity during steering remote parallel simulations.