ATM Switches

One can identify three developments of the 1980s that impelled the development of high-speed switching.
The first was the continuing mismatch between computer channel speeds and the public transmission network. Twenty years ago it was MIPS talking to Kbps; now, it is close to GIPS talking to Mbps. Although the transmission networks could, by means of several tiers of multiplexing and demultiplexing, partially accommodate this torrent of zeros and ones, it was neither economical nor did it make sense intuitively to multiplex and demultiplex bit streams whose end use was a high-bandwidth device. To provide an example, a Cray supercomputer with a 100-Mbyte/s channel (800 Mbps) driving a graphics application would have had to be muxed into multiple DS-3 (45 Mbps) streams and reassembled. The costs of this sort of mismatch simply ensured that for such applications, the cheapest route was usually to move the application to the supercomputer or the supercomputer to the application.
The second development was the widespread use of graphical representation. Widely used as a means of visualization and/or data reduction, the bit streams (even when compressed) again starkly outlined the mismatch between many graphical applications and an economic means to remotely distribute them. For example, a megapixel display (800 1200) using just 16-bit color graphics and employing only 15 updates per second generates a 230.4-Mbps bit stream.