Spatial Information Grid is an ideal infrastructure to handle the data-intensive and computing-intensive geo-spatial processing. In order that each agency could ad hoc connect to this computing environment and make autonomous decision, we build a Geospatial Data Grid in peer-to-peer way. The query processor module in each peer can decompose the user?s query into sub-queries that executed in different nodes. One problem in the parallel spatial join query optimization is how to determine an appropriate node group to disseminate the sub-queries. Especially, if there is more than one node sharing the same area of interest, there is a dilemma: on the one hand, the task scheduler tends to decompose this query into sub-queries and disseminate them to as many as possible nodes so that they could process the user?s query in parallel; on the other hand, recruiting too many nodes will also bring in overhead in repetitive computing, redundant data transmission, and the result merging. Based on the study of trade-off between increasing parallelism and reducing redundancy using the Utility Theory in economics, we put forward a fast node selection algorithm for the parallel spatial join query dissemination. The test in our system shows this strategy could balance the above two conflict demands and is appropriate for use in Data Grid.