With the increasing availability of large quantity of geospatial data, and the growing complexity of spatial algorithms, they are the most compelling challenges to the GIS computing efficiency. Faced the problems of computing and I/O intensive, it?s an appropriate approach that involves the techniques such as distributed or parallel computing into GIS context. Some researches concerned into the parallelization algorithms for most all parts of GIS respectively, but few involved in the parallel architecture from integrating the specified overall GIS package into cluster. Fortunately, we discuss Multi-Uses Parallelization Model (MUPM) for GRASS GIS from making full use of its character of Multi-users, in which many GRASS users, located in different nodes, can work concurrently on subtasks. Under MUPM, the running function module maybe in sequential or parallel, i.e., most modules of GRASS can work in cluster with high performance without the need for redesigning and re-implementing existing algorithms. According to this, the framework, approach and key technologies etc. with Cluster-Based Parallel GIS are introduced. Finally, some tentative results are also available for illustration.