Issues and directions in scalable parallel computing
PODC '93 Proceedings of the twelfth annual ACM symposium on Principles of distributed computing
| Hi-index | 0.00 |
The Computing Surface is a flexible, extensible computing system which is constructed from a choice of intelligent building blocks. A variety of building blocks are available with a range of computational or peripheral capabilities. The general application area in which the Computing Surface is going to be used influences the number and mixture of building blocks chosen to construct a particular machine. Each building block contains a 10Mips, 1 Mflops processor with four autonomous message channels providing communications bandwidth of 10Mbytes per second. The Computing Surface contains an electronic switching mechanism which allows these message channels to be connected in a topology defined by the software of individual applications as they are loaded to be run. Because the message passing capability is 'point-to-point', total system communication bandwidth increases in proportion to the number of building blocks, and therefore in proportion to the processing performance of the system. The architecture is therefore truly extensible, allowing supercomputer class machines to be constructed.