M3C: managing and monitoring multiple clusters

PC clusters running Linux, provide the computational power of supercomputers of just a few years ago at a fraction of the purchase price. The lack of good administration and user level application management tools makes the operation of clusters more difficult than it need be and results in an increased operation cost. This paper describes an ongoing effort at Oak Ridge National Laboratory to develop tools to simplify the administration and use of computation clusters. The two tools integrated in this work include the M3C (Managing and Monitoring Multiple Clusters) and C3 (Cluster Command and Control) tool suite. M3C provides a Web-based graphical user interface for cluster administration. It is designed as an extendable framework that can work with different underlying back-end tools. C3 is an independent project that provides the underlying commands to effect an operation on a cluster. In this paper it serves as an example how a back-end tool can be integrated into the M3C framework