ARCHIVE

Since the beginning of 2007 a Cell Linux cluster is doing it's work at Forschungszentrum Julich in Germany. In this context, a cluster of 12 IBM QS20 blades - 2 BladeCenter chassis with 24 CBE (Cell Broadband Engine) processors in total - were procured to examine the potential of cell processors as a building block for future high-end computing systems. In particular superscalar applications and performance analysis as well as the efficient management of clusters of cell blades will be investigated. Based on Fedora Core-5/6 Linux the MPI stack and management features of ParaStation are used. Combined with ParaStation's GridMonitor the system health and monitoring is integrated into the existent monitoring and administration platform of all other cluster systems. This enables system administrators to have a focused view of all installed cluster systems on a single cluster console, independent of vendor, Linux operating system and installation site. A multiplicity of information from different devices and services from a cluster may be read, evaluated and stored. The GridMonitor provides the administrator with various aspects of the available information, from an overall status of all configured clusters to in-depth details of nodes and devices. Data can be grouped with respect to different aspects and are visualized using a web browser. Furthermore, parameters may constantly be monitored and the administrator may be notified, if required. With this feature of providing a single cluster tool ParaStation and GridMonitor are now supporting x32 and x64 architectures (Intel, AMD) as well as Itanium, Power, PowerPC and Cell based systems. For more informations on the project at Forschungszentrum Julich visit www.fz-juelich.de/zam/juice. ParaStation and GridMonitor are products of ParTec, Germany (www.par-tec.com).