) celebrates the release of its much delayed update of Windows, called Vista, it will probably mark the end of the road for Windows as an all-in-one operating system. Projects on the scale of Vista -- updating and writing tens of millions of lines of interlocking code -- are becoming impossible to debug fully. While Windows will be around and making money for Microsoft for a long time, there's a better way to build such software. But getting beyond the complexities of Vista will mean changing the Windows operating system in fundamental ways. The challenge facing Microsoft is not simply its massive size but the fact that its pieces interact in ways that are beyond human comprehension. My installation of Windows XP Professional, for example, includes 1,600 "dynamic link libraries," a type of file that's particularly likely to cause troublesome unanticipated interactions.
The cure for this complexity lies in a new paradigm for PC software. Technology exists that can divide a large and complex operating system into a number of smaller, simpler units that run on one computer but function independently of each other. To the user, it will look much like today's software, but it will be less prone to glitches, crashes, and attacks.THE IDEA OF SPLITTING THE SOFTWARE BRAIN of a single, physical computer into a number of software-based "virtual machines" has been around since IBM (IBM
) made such software available on mainframes in the late 1960s. In recent years the technique has been used extensively on the big computers that power corporate networks and the Internet. For example, a single computer might be split into three "virtual servers" -- one to handle Web pages, one to process e-mail, and the third to run a database. An immediate benefit is improved reliability, because a software crash on any one virtual machine does not affect the others.
A scenario of how this might work on PCs was laid out for me by Gregory Bryant, general manager of Intel's (INTC
) digital office platform division. One virtual machine might handle ordinary applications, such as Microsoft Word or TurboTax. A second could be optimized to handle digital media: music, videos, or photos. Both of these systems would link to the network (and Internet) through a third virtual machine that would handle the actual connections. This division of labor could make PCs safer, since the communications module would be solely dedicated to secure networking and need to be updated only to fend off viruses and other malware.
Except for the networking part, which will be supported on new Intel chips due out later this year, this approach could be achieved today using software from Microsoft or VMware. But each virtual machine would have to run its own copy of Windows, making the whole system spectacularly inefficient.
The post-Vista computer will probably use a far more streamlined operating system that loads only the components needed by each virtual machine. Linux has this sort of modularity today, but Windows does not. What's more, running several virtual machines on an operating system will require a big increase in processing power, preferably from multiple processors. The switch is already under way. Nearly all the chips produced by Intel and Advanced Micro Devices (AMD
) will have at least two processors. The Sony (SNE
) PlayStation 3 is built around a nine-processor IBM chip, and Intel has plans for chips packed with thousands of processors.
The operating system that makes this all work won't be Vista, and probably won't arrive until 2012. Microsoft could be the pioneer, but it will require major cultural changes, given that the company has spent the past 20 years making Windows ever bigger and richer in features. The fact that Microsoft currently has a monopoly is no guarantee, as we are talking about a truly disruptive change. The post-Vista world could see the first real competition for the desktop since Windows 95 cemented Microsoft's dominance a decade ago.For past columns and online-only reviews, go to Tech Maven at www.businessweek.com/technology/wildstrom.htm By Stephen H. Wildstrom