High Performance Computing: PALMA II
The ZIV provides services in the field of scientific high performance computing. That way high-capacity hardware can be used for numerically solving complex problems, which standard desktop PCs are insufficient for by far. On the one hand, high performance computers include high-parallel supercomputers, which combine numerous processors and a very extensive main memory. On such SMP systems (“Symmetric Multiprocessing”) operations can run on several processors simultaneously while accessing a shared main memory. Therefore, this is called “Shared Memory Programming”. On the other hand, it is possible to achieve an enhancement in processing power using computer clusters. Here various individual computers are connected via a fast communication network (e.g. Myrinet, Infiniband). For simultaneous appliance the main memory is distributed to several systems.
PALMA stands for “Paralleles Linux-System für Münsteraner Anwender” (“Parallel Linux System for Münster’s Users”). In contrast to normal PCs, this is a computer cluster with over 18,000 processor cores, the interaction of which enables even the most complex tasks in science and research to be solved. The new cluster has been in the service of research since the beginning of July 2018. In the previous test operation, a total performance of 800 TFlops could be determined in the Linpack benchmark, which means a placement on rank 418 of the Top 500.
- more than 18,000 processor cores
- computing power of more than 800 TFlops
- 80 TB main memory
- 800 TB of hard disk space
- funded by the German Research Foundation (Deutsche Forschungsgesellschaft, DFG), the State of North Rhine-Westphalia and contributions from the Rectorate, the It Center (ZIV), the Departments of Physics, Chemistry, Medicine and Economics, and the Helmholtz Institute