While the robotics community agrees that the benchmarking is of high importance to objectively compare different solutions, there are only few and limited tools to support it. To address this issue in the context of multi-robot systems, we have defined a benchmarking process based on experimental designs, which aimed at improving the reproducibility of experiments by making explicit all elements of a benchmark such as parameters, measurements and metrics. We have also developed a ROS (Robot Operating System)-based testbed with the goal of making it easy for users to validate, benchmark, and compare different algorithms including coordination strategies. Our testbed uses the MORSE (Modular OpenRobots Simulation Engine) simulator for realistic simulation and a computer cluster for decentralized computation. In this paper, we present our testbed in details with the architecture and infrastructure, the issues encountered in implementing the infrastructure, and the automation of the deployment. We also report a series of experiments on multi-robot exploration, in order to demonstrate the capabilities of our testbed.