Benchmarking for educational purposes in the context of computer science can be hindered by the low number and the homogeneity of machines to be assessed, and the inaccuracy of the benchmarks to represent specic workloads. Thus, this paper proposes a benchmarking tool developed within a service-oriented grid in order to allow students to benchmark multiple workloads in machines that may belong to several educational institutions. This tool has been validated in a real educational scenario within a course on Computer Architecture.