Abstract

It is becoming evident that the switched-networking of multiple general-purpose processors is an efficient and cost-effective path to achieving high-performance computing engines. In such distributed computing environments, the interconnection network typically represents a fundamental performance bottleneck, as electrical interconnect technologies struggle to keep pace with increasing demand for more connectivity, throughput, and rapidly diminishing processor clock-cycle times. The demands made on the data network include (i) high throughput communication, (ii) low contention and (iii) simple physical-layer communication protocols. Additional features of low latency data transport and packet switched operation are desirable.

© 1997 Optical Society of America