High Performance Computing

High Performance Computing

The essential feature of high performance computing is the alignment to an architecture suited for parallel computing. High performance computing on the one hand are high-parallel supercomputer on the other hand computer cluster, organized distributed ( distributed computing, grid- computing) or within a local group. Also other computing systems with specialized architecture, for example organized with a very largely memory, falls into the class of high performance computers. At present one ranks the computer of the teraflops performance class and above as high performance computer. Companies like IBM, Cray, SGI, Hitachi , Fujitsu adressing this lucrative and future oriented market.

High performance computing is particularly in scientific computing increasingly of importance and serves as aid to the computation, modelling and the simulation of complex systems and to the processing of enormous measuring data quantities. Such applications are used practically in all ranges of nature- and technical science, typical applications are for instance meteorology and climatology, astro and particle physics, system biology, genetics, quantum chemistry and fluid mechanics. Many of it are scientific origin (e.g. weather forecast, crash test simulation, flow simulation in the aircraft construction) in addition, applications are given without a scientific character, e.g. with the generation of animated films.

FPGA High Performance Computing (FHPC)

Since a few years FPGA based high performance computing is offered and is moving in the meantime for many applications to mainstream. FPGA devices with the latest technology are used instead of CPUs. The advantage against CPU Cluster is the fast computing time (x – times acceleration factor) , low power dissipation, low investment cost and low cost of operation. Obstacles, like the fact that the C algorithm needed to be transfered by hand to FPGA technology, the essential need for FPGA design now how, low level external support, are history. EDA companies developed C to FPGA synthesis tools which automates the conversion from C to FPGAs, FHPC turn key solutions are available and also companies are offering a broad range of services helping to map the algorithm in the parallel structures of FPGAs to gain the necessary acceleration speed.

Options for FHPC usage

FHPC cannot replace the established HPC solutions but they can definitely be used for appropriate applications and can be used in coexistence with CPU Cluster. For the future they are more than an alternative for the growing demand and variety of applications.

• Complete application will be accelerated on the FPGA Cluster
• Part of the application will be accelerated in coexistence with the CPU Host Cluster

SC 3 Cluster

FPGA High Performance Computing Applications

• Medical imaging
• Advanced radar processing
• 3D image rendering
• National security
• Video transcoding
• Battlefield analysis
• Biomedical computing
• Scientific computing
• Financial modeling and analytics
• Earth sciences