High-performance computing resources

Scientists and students at the German Research School for Simulation Sciences have privileged access to the high-performance computing facilities at the Jülich Supercomputing Centre (JSC) and the Center for Computing and Communication of RWTH Aachen University.

No less than three supercomputers for European research are available in Jülich: JUQUEEN, JUROPA and HPC-FF. Two of them, JUQUEEN and JUROPA, are intensively used by the scientists of the German Research School for Simulation Sciences.

JUQUEEN is currently the fastest computer in Europe with a computing power of 5.033 petaflop/s. This is equal to the computing power of more than 90,000 modern PCs. This supercomputer is equipped with 393,216 processors housed in 24 cabinets, each around the size of a phone booth, in the computer room in JSC. Its memory capacity is around 393 terabytes. Together with the other Jülich supercomputers, JUQUEEN have access to around 6 petabytes of hard-drive storage. This corresponds to the storage capacity of more than a million of DVDs.
		The second supercomputer JUROPA is based on a cluster configuration of NovaScale servers from the French computer specialist Bull, and on blade servers from the American company Sun with Intel Nehalem processors. It consists of 2208 computing nodes. Each node is equipped with two Nehalem EP Quad Core processors reaching a total computing capacity of 207 teraflop/s. The total of 17664 water-cooled processors have a clock rate of 2.93 GHz each and they will be able to access approximately 24 terabytes of total main memory.
Source: Forschungszentrum Jülich
		The Center for Computing and Communication of RWTH Aachen University operates a new compute cluster consisting of two main partitions, the MPI partition with 1350 2-socket nodes with Intel Westmere EP processors and the SMP partition with 358 4-socket modules with Intel Nehalem EX processors. Either two or four of these 4-socket modules can be coupled by proprietary BCS chips to build 8- or 16-socket systems, respectively. This has been recently performed making available 16-socket nodes. These very large compute nodes with 16 8-core processors each will provide up to 1 TeraByte of main memory providing an opportunity for a new class of applications.
Furthermore 16 of these 4-socket modules will be coupled by the vSMP software of the company ScaleMP to generate a singles system with 512 processor cores accessing 4 TeraByte of shared memory (minus overhead). The efficient usage of such a system requires a sensible way of NUMA-aware programming. Nodes are interconnected with a fast QDR Infiniband network. The file server infrastructure consists of a highly available NetApp file server and also a Lustre file server for fast parallel file access. The total compute power is over 300 TeraFlop/s. The total main memory capacity is about 92 TeraByte. The detailed configuration is listed here. Currently 95% of compute nodes are operated with Scientific Linux distribution and 5% with Windows 2008 Server.
Source: Center for Computing and Communication of RWTH Aachen University
The utilization of those resources provides a cornerstone for the German Research School of Simulation Sciences. Only with the necessary resources at hand development, testing and production runs of highly scalable high-performance codes become possible.