Grids offer the potential of harnessing vast amounts of computational resources during the execution of demanding computations. These resources are geographically distributed, owned by different organizations and are highly heterogeneous. All these create an uncertain environment in all phases of a Grid Scheduling Process (GSP). In this work, we focus on the resource discovery process during which clients of the grid discover possibly suitable resources available for their computation. We propose a network of resource representatives, which maintain the moreor-less static characteristics of available workers they represent (e.g. OS, CPU type etc.). We show that clustering algorithms is a promising approach that can be used for the efficient discovery of suitable resources for a given task set.