In the first part of the overview, the authors present the main architectural classes of high performance computing systems: shared-memory single-instruction-stream, multiple-data-stream (SIMD); distributed-memory SIMD; shared-memory multiple-instruction-stream, multiple-data-stream (MIMD); distributed-memory MIMD; and cache-coherent non-uniform memory access (CC-NUMA). For each of these classes, a concise description of the main architectural features is presented. Next, the authors present several examples of current high performance computing systems.
They present the main characteristics of each machine, including the clock cycle, theoretical peak performance, main memory capacity, number of processors, and communication bandwidth. In addition, several remarks concerning the architecture and the software are given.
The overview concludes with a discussion of several systems under development that may lead to commercial systems that could be introduced to the market in the near future. The overview is an excellent reference for researchers interested in high performance computing.