Networking and Internetworking - Guide to Cloud Computing for Business and Technology Managers - page 37

Metcalfe and his colleagues felt that even the Data General network was
too expensive and not flexible enough to work in an office setting, where
one may want to connect or disconnect machines frequently. He also felt it
was not robust enough—the network's operation depended on a few critical
pieces not failing. He recalled a network he saw in Hawaii that used radio
signals to link computers among the Hawaiian islands, called ALOHAnet.
With this system, files were broken up into packets , no longer than 1000 bits
long, with an address of the intended recipient attached to the head of each.
Other computers on the net were tuned to the UHF frequency and listened
for the packets, accepting the ones that were addressed to it and ignoring all
the others.
What made this system attractive for Metcalfe was that the medium—in
this case radio—was passive. It simply carried the signals, with the com-
puters at each node doing the processing, queuing, and routing work. The
offices at Xerox PARC were not separated by water, but the concept was per-
fectly suited for a suite of offices in a single building. Metcalfe proposed sub-
stituting a cheap coaxial cable for the ether that carried ALOHAnet's signals.
A new computer could be added to the Ethernet simply by tapping into the
cable. To send data, a computer first listened to make sure there were no
packets already on the line; if not, it sent out its own. If two computers hap-
pened to transmit at the same time, each would back off for a random inter-
val and try again. If such collisions started to occur frequently, the computers
themselves would back off and not transmit so often. By careful mathemati-
cal analysis, Metcalfe showed that such a system could handle a lot of traf-
fic without becoming overloaded. He wrote a description of it in May 1973
and recruited David Boggs to help build it. They had a fairly large network
running by the following year. Metcalfe recalled that its speed, around three
million bits per second, was unheard of at the time, when “the 50-kilobit-
per-second (Kbps) telephone circuits of the ARPANET were considered fast.”
Those speeds fundamentally altered the relationship between small and
large computers. Clusters of small computers now, finally, provided an alter-
native to the classic model of a large central system that was timeshared and
accessed through dumb terminals. Ethernet would have its biggest impact
on the workstation, and later PC, market, but its first success came in 1979,
when Digital Equipment Corporation, Intel, and Xerox joined to establish
it as a standard, with DEC using it for the VAX. UNIX-based workstations
nearly all adopted Ethernet, although token ring and a few alternate schemes
are also used.
DOS-based personal computers were late in getting networked. Neither
the Intel processors they used nor DOS was well suited for it.
Purchasers of PCs and PC software were driven by personal, not corpo-
rate, needs. The PC and DOS standards led to commercial software that
was not only inexpensive but also better than what came with centralized
systems. By the mid-1980s, it was clear that no amount of corporate pol-
icy directives could keep the PC out of the office, especially among those