Encyclopedia of Computer Science and Technology

supercomputer 461by the 1990s the growing power of regular desktop PCs wasreducing the need for special-purpose workstations. As theWeb grew starting in the 1990s, Sun’s line of multiprocessingWeb servers became quite successful, though the “dotbust”of the early 2000s cut revenues.One of Sun’s founders was a key developer of UNIXsoftware (see Joy, Bill). Sun developed its own version ofUNIX (SunOS) for its workstations in the 1980s, and thenjoined with AT&T to develop the widely used UNIX SystemV Release 4, which in turn became the basis for Sun’s newoperating system, Solaris. (Sun has also supported the useof Linux on its hardware.)Sun’s biggest impact on software development, however,has been its development of the Java language and platformsince the early 1990s. Although newer languages such asPython, PHP, and Ruby have come along to challenge it,Java, with its ability to run via “virtual machines” on allmajor platforms, is widely used and has a rich set of libraryroutines and programming frameworks.Scott McNealy, one of the company’s founders, remainsits chairman. Sun had $13.87 billion revenue in 2007 ($473million net income), and employs about 36,400 people.Further ReadingBoyous, Jon. “Java Technology: The Early Years.” Sun DeveloperNetwork. Available online. URL: http://java.sun.com/features/1998/05/birthday.html. Accessed November 18, 2007.Southwick, Karen. High Noon: The Inside Story of Scott McNealyand the Rise of Sun Microsystems. New York: Wiley, 1999.Sun Microsystems. Available online. URL: http://www.sun.com/.Accessed November 18, 2007.Sun Multimedia Center [videos]. Available online. URL: http://sunfeedroom.sun.com. Accessed November 18, 2007.Sun Wikis. Available online. URL: http://wikis.sun.com. AccessedNovember 18, 2007.supercomputerThe term supercomputer is not really an absolute termdescribing a unique type of computer. Rather, it has beenused through successive generations of computer designto describe the fastest, most powerful computers availableat a given time. However, what makes these machines thefastest is usually their adoption of a new technology orcomputer architecture that later finds its way into standardcomputers.The first supercomputer is generally considered to bethe Control Data CDC 6600, designed by Seymour Crayin 1964. The speed of this machine came from its use ofthe new, faster silicon (rather than germanium) transistorsand its ability to run at a clock speed of 10 MHz (a speedthat would be achieved by personal computers by the mid-1980s). Even with transistors, these machines generatedso much heat that they had to be cooled by a Freon-basedrefrigeration system.Cray then left CDC to form Cray Research. He designedthe Cray 1 in 1976, the first of a highly successful seriesof supercomputers. The Cray 1 took advantage of a newtechnology, integrated circuits, and new architecture: vectorprocessing, in which a single instruction can be appliedA Cray 190 A supercomputer. Seymour Cray’s leading-edgemachines defined supercomputing for many years. (NASA photo)to an entire series (or array) of numbers simultaneously.This innovation marked the use of parallel processing asone of the distinguishing features of supercomputers. Themachine’s monolithic appearance gave it a definite air ofscience fiction, and the first one built was installed at thesecretive Los Alamos National Laboratory.The next generation, the Cray X-MP, carried parallelismfurther by incorporating multiple processors (the successor,Cray Y-MP, had 8 processors, which together could performa billion floating-point operations per second [1 gigaflop]).Soon Cray no longer had the supercomputer field toitself, and other companies (particularly the Japanese manufacturersNEC and Fujitsu) entered the market. The numberof processors in supercomputers increased to as manyas 1,024 (in the 1998 Cray SV1), which can exceed 1 trillionfloating-point operations per second (1 teraflop).Meanwhile, processors for desktop computers (such asthe Intel Pentium) also continued to increase in power, andit became possible to build supercomputers by combininglarge numbers of these readily available (and relatively lowcost)processors.The ultimate in multiprocessing is the series of ConnectionMachines built by Thinking Machines Inc. (TMI)and designed by Daniel Hillis. These machines have upto 65,000 very simple processors that run simultaneously,and can form connections dynamically, somewhat likethe process in the human brain. These “massively parallel”machines are thus attractive for artificial intelligenceresearch. It is also possible to achieve supercomputerlikepower by having many computers on a network divide thework of, for example, cracking a code or analyzing radiotelescope data for signs of intelligent signals.Programs for supercomputers must be written usingspecial languages (or libraries for standard languages) thatare designed to provide for many processes to run at thesame time and that allow for communication and coordinationbetween processing (see multiprocessing).