The system consists of homogeneous computing nodes, including: one or several central processors (usually RISC), local memory (direct access to the memory of other nodes is impossible), a communication processor or network adapter, hard drives and / or other I / O devices added special I / O nodes and control nodes. Nodes are connected through some communication medium (high-speed network, switch, etc.)
IBM RS / 6000 SP2, Intel PARAGON / ASCI Red, SGI / CRAY T3E, Hitachi SR8000, Parsytec.SP2T3ESR8000Parsytec transputer systems. Scalability
(Scalability is an opportunity to increase the number and capacity of processors, RAM and external memory and other computing system resources. Scalability should be provided by computer architecture and design, as well as appropriate software tools.) The total number of processors in real systems reaches several thousand (ASCI Red , Blue Mountain).
There are two main options: A full-fledged OS only works on the controlling machine (front-end), a strongly stripped-down version of the OS runs on each node, ensuring only the work of the branch of the parallel application located in it. Example: Cray T3E. Each node runs a full-fledged UNIX-like OS (an option close to the cluster approach). Example: IBM RS / 6000 SP + OS AIX, installed separately on each node. Cluster
Programming within the message transfer model (MPI, PVM, BSPlib) MPIPVMBSPlib