Grid computing collects geographically dispersed resources ranging from laptops to supercomputers to compute tasks requested by clients. Grid scheduling, i.e., assigning tasks to resources, is an NPhard problem, and thus, metaheuristic methods are employed to find the optimal solutions. In this paper, we propose a Chemical Reaction Optimization (CRO) algorithm for the grid scheduling problem. CRO is a population-based metaheuristics mimicking the interactions between molecules in a chemical reaction. We compare the CRO approach with four generally acknowledged metaheuristics, and show that CRO performs the best.