The workstation market has always been about putting high-performance computing systems into individual users' hands: When a geophysicist needs to analyze terabytes of data to find pockets of oil underground, or an engineer wants to calculate the effects of wind shear on an airplane's tailfin at 40,000 feet, a workstation is the best, most cost-effective way to get the job done. But even more complicated tasks, crunching through colossal datasets, call for even more power -- which has traditionally meant leasing time on a supercomputer or otherwise having to wait for time on a system shared by multiple users.

But now a Santa Clara, Calif., startup company, Orion MultiSystems, is offering businesses the next best thing to a supercomputer on every desktop. "Our systems are meant to fill the gap between the high-end PC and the back-room supercomputer," explains CEO Colin Hunter.

Actually, the Orion MultiSystems workstations share the cluster architecture of many Linux supercomputers, but instead of offering multiple machines linked by Gigabit Ethernet within one room, Orion installs multiple processors linked by Gigabit Ethernet inside one case. The result, the company says, is a system that delivers the same performance as a data center, but is easier for users to access and for companies to maintain.

Orion's compact clusters caught the attention of Stacey Quandt, an analyst for the Robert Francis Group, an IT research firm. In her report The Cluster Workstation: A New Wave of Disruptive Technology, Quandt wrote, "Orion is avoiding the conventional wisdom of the 1980s, which was to target high-priced niches and shun broader, more cost-sensitive opportunities. Individuals or small groups of programmers can leverage an Orion workstation cluster for application development and porting, API and library consolidation, and standard binary operation potentially faster and at a lower cost than other current alternatives."

Many Hands Make Light Work

The release of Orion's workstations builds on a steep increase in the use of clustered systems to perform computing-intensive tasks. Over the last decade, supercomputers have evolved from huge, proprietary systems to collections of relatively small, standard ones, usually using some form of the open-source Linux operating system. Thomas Sterling and Don Becker at the Center of Excellence in Space Data and Information Sciences (CESDIS), a part of the U.S. National Aeronautics and Space Administration (NASA), originally developed this concept -- often called a Beowulf cluster -- in the early '90s. Since then, Beowulf clusters have sprung up in countless college computer science departments, as well as becoming popular for demanding corporate projects.

Still, Beowulf clusters have their drawbacks. For one thing, they tend to be cobbled together from a variety of sources. While this keeps initial costs down and enables nearly any collection of hardware to generate multigigaflop performance, such clusters can be difficult to build and maintain. In addition, these networked systems are almost always shared resources among multiple users.

"The problem with any shared solution is scheduling access," says Hunter. This makes traditional clusters impractical for use by animators, engineers, or architects who need access to massive computing power without having to schedule an appointment.

This is where Orion comes in: it takes cluster computing out of the back room and put it on -- or at least on the floor next to -- the professional user's desk. The company is currently offering two cluster workstation models, the 12-processor DT-12 desktop and the 96-processor "deskside" model DS-96. Each works much like a conventional high-end PC -- plug it into a single, ordinary AC outlet, hit the switch, and the system boots in under 90 seconds -- but provides a lot more power.