Monitoring of Large-scale Cluster Computers - Organized Approaches, Identification of Performance Issues and Minimization of Downtime

To monitor the state of a cluster computer with hundreds of computing nodes is not as simple as monitoring the state of one's personal computer at home. Especially when handling such valuable systems where essential computations for scientific or important business purposes are being executed, it is essential to be up-to-date about the systems functions. This book presents a classification of the wide and vaguely used term «monitoring» for computer clusters. In addition to that a solution is developed to perform scaleable monitoring of clusters with an InfiniBand network interconnection. Therefore, extensive analyses to determine the real impact of the monitoring on the computing performance of the cluster are presented, for which partially the monitoring suite Nagios is used. The book is directed to professionals, researchers and other persons, that have to deal with the management and monitoring of an InfiniBand network, as well as with the issue how much the monitoring process influences computations of the cluster (CPU and network impact) and how to minimize this.