Sun Hopes to Open New Doors to Using Microchips

Everybody has a favorite anecdote about futurists--their failed predictions of personal helicopters, say, or their prophesies of food pills. My favorite, however, involves one futurist’s vindication.

Twenty-one years ago, Danny Hillis, the founder of supercomputer manufacturer Thinking Machines Corp., gave a speech at the New York Hilton in which he predicted that someday there will be more computers in the world than people. He was immediately met with ridicule: “What will we do with them?” one man asked, “Put them in doorknobs?”

Today, if you go to the New York Hilton, that’s exactly what you’ll find: card-key locks controlled by a tiny computer on every door.

This anecdote illustrates a vital truth about today’s computer market. Most of the microprocessors sold today don’t go into personal computers or workstations--they go into devices like phones, stereos and, yes, doorknobs.

This fact underscored Sun Microsystems Inc.’s announcement early this month of three new low-cost microprocessors specially designed to support Sun’s Java programming language and scheduled for release in early 1997. Representing a major shift for the Mountain View, Calif.-based workstation manufacturer, these chips--the picoJava, microJava and ultraJava--are aimed squarely at the doorknob market. It’s a market that Sun believes could reach $15 billion by 2000.

Sun is convinced that cheap Java chips will succeed because consumer devices are increasingly communicating with each other over a network. Java, Sun argues, is perfect for such tasks, and hence inexpensive chips that can run Java programs are bound to be popular. Perhaps. But you can’t help wondering: Will Java chips really end up becoming ubiquitous, or will they turn out to be as successful as jet packs?

A quick look at the embedded market (embedded refers to any use of microprocessors outside of what are traditionally considered computers) shows that there is at least a window of opportunity for Sun. Right now, the market is dominated by old chips no longer powerful enough for PCs. Remember Intel’s 286 chip? Now available for just a few dollars, you can find it controlling things like Coke machines and dishwashers. Same with old Motorola 68020 chips.

*

The problem is that embedded applications have become much more complex in just the last two years, and the old microprocessors are no longer up to the task. The software in a cellular phone, for example, has become so complicated that it is very difficult to write using traditional embedded techniques. Compounding the problem, embedded processors such as those found in your laser printer or in your car are usually connected to a network. This is also difficult to do well with standard embedded processors and programming techniques.

*

Java helps out in both of these areas. Java is similar to the popular C++ programming language, but has been carefully designed so that it is easier to use and “network-friendly.” It manages to achieve these goals in a few ways. First, Java is simple. While C++ suffers from a bad case of featuritis (in hopes of appealing to everybody, it offers more features than any one person could ever need), Java offers just the features that matter. Second, Java is safe. That means any Java program can be sent over the network, and you don’t have to worry about whether it contains a virus. Third, Java is portable. A Java program written on an IBM PC, for example, will run on a Macintosh without any modification.

Here’s how these three attributes can be used to advantage in an embedded application. Suppose the phone company adds a new feature to their cellular network. They’d like to be able to update every phone so it supports this new feature, but today that would require every customer to bring in their phone. With Java, however, the phone company could write a simple “applet” and then download it over the cellular network to every phone.

Because Java is safe, nobody needs to worry that the applet will somehow “crash” his or her phone. And even though the phones are made by many different manufacturers, they can all support Java.

Although this scenario illustrates how Java can be used for embedded applications, it doesn’t support the case for special Java chips. After all, part of Java’s appeal is that it can run on any processor.

Well, yes. But there is an important caveat: Any microprocessor can run Java programs, but some can run them much faster. It turns out that Java’s biggest problem right now is that it’s extremely slow. Programs written in Java end up taking much longer to run than equivalent programs written in C or C++.

Sun suggests the solution is to throw hardware at the problem--and conveniently, Sun is in the best position to provide that hardware. By tailoring a microprocessor to Java’s needs, Sun should be able to improve performance by 300% to 500%. And because Sun already makes its own line of chips for workstations, much of the design groundwork has already been laid. If the result is cheap enough, embedded engineers should find the chips irresistible.

When James Gosling of Sun first came up with the idea of Java, he imagined it being used in every light switch so they could communicate with other consumer appliances and even display information about what they control on a small LCD. Right now, that sounds like a fantasy straight out of “The Jetsons.” But in 20 years, Gosling may look as prophetic as Danny Hillis does today.