bwGRiD Cluster Heidelberg (High Performance Cluster)
This high performance cluster can be used for large compute jobs, for example extensive model calculations. The cluster is designed for batch operation, i.e. the compute jobs are executed non-interactive via execution of a script written by the user. Interactive access is possible too, e.g. for software development (tests, debugging) and for work with interactive programs. The machine can be used for
- distributed memory parallel computing (Each compute node has its own memory. Other nodes cannot read from and write to this part of memory. 140 compute nodes with all together 1120 cores can be used.)
- shared memory parallel computing (up to 8 cores share a common memory space, i.e. each core can read/write to this common part of memory, SMP Symmetric Multi-Processing)
- scalar computing (Only one core is working on the problem.)
The state-universities of Baden-Württemberg have proposed to build a regional grid infrastructure (bwGRiD). The proposal was headed by the High Performance Computing Center Stuttgart (HLRS) and was funded with cluster hardware by the German Federal Ministry of Education and Research. Additional manpower and storage infrastructure is funded by the state government of Baden-Württemberg. Our parallel computer bwgrid.uni-heidelberg.de is one of these high performance clusters of the bwGRiD Initiative.
Interconnection with bwGRiD Cluster Mannheim
Since July 2009, the bwGRiD clusters in Heidelberg and Mannheim have been combined to a single cluster. The system operates with a common user administration and a common batch system. Alltogether, it provides the computing power of 2240 cores to the members of the both Universities.