ABSTRACT
SuperMUC and its supporting building facility have been equally funded by the federal government of Germany and the Free State of Bavaria. SuperMUC uses IBM’s General Parallel File System with 15 Pbyte of capacity and with an aggregated throughput of 350 GB/s. Both phases of SuperMUC use the IBM LoadLeveler as a resource management and scheduling system. SuperMUC distinguishes itself by its versatile usability due to its homogeneity and standard instruction set architecture. SuperMUC specifically addresses the needs for very large applications capable to use a large portion of the configuration or even the full system. The mentioned increased power consumption of high performance computing (HPC) systems does not only translate to high operational costs but also, as discussed by Natalie Bates et al., to high carbon footprints, affecting the environment, and limiting the further expansion of HPC data centers. PowerDAM uses agent-based data communication models for actual sensor data retrieval from the monitored systems/entities.
