Apr 9, 2001, 11:45am By John C. Dvorak
Chet Heath was known as a wild duck at IBM, an endearing term reserved for free-thinking creative IBMers who actually made things happen at the $88 billion company.
Heath, a 30-year veteran of IBM, quit to start OmniCluster to promote an invention he had been working on at IBM since 1989. But IBM refused to market the product. Called the SlotServer 1000, it's a computer on a card that can save a company from having to buy more expensive upgrades.
It can save a data center from having to physically expand its operations, and saves money all around. Because it would obviously hurt sales of current IBM offerings, the invention didn't fit into the corporate strategy.
The minuscule computer/server developed by Heath only requires 10 watts of power and fits on a single half-sized PCI card. This device can inexpensively double the speed at which a server can output Web pages. But the computer-on-a-card, which will initially sell for just over $500, has all sorts of other interesting uses, thanks to software developed by superstar coder Bart Brooks and his team.
For example, the card can be used to create a versatile multiuser machine. It has a built-in USB port, ethernet connection and SVGA graphics card. As a result, you can easily hook a monitor, keyboard and mouse to any SlotServer 1000 card and have a complete independent workstation. According to OmniCluster Chief Executive Christopher Fleck, it can produce as many so-called zero-footprint computers as there are empty slots in any nearby system. And each is immune to the activity of the other "computers," unlike software-based multi-user systems.
While this idea is not entirely new, it appears that OmniCluster has the right price and the right attitude towards universal compatibility. "I learned a lesson from the Microchannel experience at IBM." The Microchannel designs rolled out at IBM with its PS/2 machine in 1987 failed to become standards--no thanks to IBM's onerous licensing policy at the time--and eventually disappeared from the market.
In the company booth at the recent ISPCon trade show in Baltimore for owners and operators of Internet service providers, they had one card stuffed into a Sun Cobalt server. The Cobalt box was running Linux and serving Web pages, but overseeing the system was the SlotServer card running Windows. The card worked independently of the Cobalt, but could also monitor and control the Cobalt while serving pages itself. This device is the best solution I've ever seen for remote access maintenance, since any one of the SlotServer 1000 cards can act as sniffer or spy or control mechanism for the rest of the system and for the other cards.
Heath is an interesting character. His most famous inventions at IBM included the micro-channel architecture (MCA), probably one of the most advanced computer architectures for small machines yet designed. It directly led to the PCI bus, which is common today.
He is also responsible for most of the patents for plug-and-play technology. Furthermore, he may be the only employee at IBM to have ever won the rare IBM Corporate Technology Achievement award twice. He was presented with his second award in June 2000, two months before he left IBM to become chief technology officer for OmniCluster. He packed his bags and moved to Boca Raton, Fla., where the company is headquartered.
Heath's departure from IBM and the subsequent loss of other quality employees may not mean much in the short run for IBM, but it may signal a long-term problem if the company cannot keep other creative talent. Heath left with a cadre of other IBMers such as Fleck, an executive responsible for single-board computers and industrial computers and OmniCluster's CEO. IBM's singleboard engineer Ken Honeycutt also joined the group.
Fleck and Heath negotiated a deal with IBM to obtain the technology that Heath developed. Though IBM wasn't interested in selling it, Fleck was confident he could. IBM got equity in OmniCluster in exchange for the buyout. It also appears that IBM has agreed to a non-compete arrangement as part of the deal. The arrangement, according to the principals, is secret. In addition to IBM, other investors include HIG Ventures, CrossBow and Cenetec.
OmniCluster should do well, and if it impacts IBM, then IBM can always offer to buy them out. "I wouldn't be adverse to it," says Heath. In the meantime, I suspect this once-wild duck will be quacking all the way to the bank.