In order to implement big data access and process with high efficiency, an algorithm of nodes location was proposed according to the state of computable resources. In this paper, we first describe and map the computational resource with javascript object notation(JSON) in P2P network system. Regarding the computational nodes as spatial points, then we present a generalized euclid distance(GED) model using the method of spatial points clustering. Through this model, the computational nodes can partition into multiple sub-group upon the characteristic attributes. After that, we calculate the spatial distance and attribute distance by spatial geometric model of global network positioning(GNP), ultimately implement the computable nodes location with efficiency, to provide the basis of load balance, especially in cloud computing. Experimental results show that our method not only can significantly improve system performance, also in accuracy of nodes location.