The Center for Information Services and High Performance Computing (ZIH) is a central scientific unit of the Technische Universität Dresden in Germany with a broad spectrum of services and research activities. With its interdisciplinary orientation, ZIH supports other departments and institutions in their research and education for all matters related to information technology and computer science.
In its mission to be a competence centre for scientific computing, parallel programming, and performance optimization, ZIH offers its HPC resources to academic users as well as support for HPC application developers. It also cooperates with other international HPC centers to assist users in enhancing their programs on highly parallel systems with more than 100,000 cores.
ZIH’s own research activities include interactive performance analysis and visualization based on event tracing methods, as well as automatic debugging tools that use a runtime error detection approach. The widely used tool Vampir, which is developed at the ZIH, is one of the leading commercial performance analysis tools. ZIH and other partners jointly develop the scalable measurement environment used by Vampir as an Open Source project. Current research activities for performance analysis focus on pattern processing, GPGPU computing, scalability, and energy aware performance optimization. In the runtime error detection field, ZIH develops the Marmot tool, which automatically detects wide ranges of MPI usage errors. A new 2nd generation runtime error detection called MUST, will provide significant scalability improvements and is jointly developed with Lawrence Livermore National Laboratory.
