摘要：Research upon Computer Science, especially in large scale distributed systems like Grids, P2P systems and High Performance Computing (HPC) areas, has to deal with issues related to increasingly complex sys- tems. Theoretical analysis, simulation and even em- ulation seem not adequate enough for a complete study of these systems. Hence the need for new generation of scientific instruments capable for the observation of complex distributed systems running at real scale and under reproducible experimental conditions has arisen. Grid5000 has been designed to answer to this need. It provides a real-life experimental research tool for computer scientists. In this paper we discuss the importance of an experimental grid platform dedicated to Computer Science research. We present the design choices of Grid5000 architecture and we analyze the key components of the platform which is OAR Resource and Job Management System, Kadeploy reconfiguration toolkit along with all the monitoring and experiments steering tools.
