Linux Poised for Desktop, High-End Surge

By Alexander Wolfe

(Back to article)

This week's LinuxWorld Expo in New York touted the advance of the open-source operating system into the sweet spot of mainstream enterprise applications.

But consensus also emerged that Linux is also mounting a pincer attack that will see it ride into increasing service at the low-end in standard desktops and at the high-end in large multiprocessing systems.

The former could steal some thunder from the dominant Windows system, while the latter has long-term significance for the IT community because the highest performing computer hardware typically trickles down to the enterprise in a few years' time.

"We are now starting to see real, measurable growth in desktop Linux," Jeremy White, chairman of the Desktop Linux Consortium (DLC), told an expo session.

White, who is also CEO of software vendor Codeweavers, made his case with statistics he said Google keeps about the computers that hit its search engine. Where the proportion of Linux machines accessing the site was 0.6 percent a year ago, that figure has recently risen to 1.4 percent. "That doesn't sound high until you consider that there's hundreds of millions of computers," White said.

The biggest stumbling block that's been overcome is ease of use. "A few years ago, you installed Linux and you'd spend weeks messing with your fonts," White said. Today, that's a thing of the past and desktop installations are solid.

Additionally, vendors long focused on the enterprise are embracing the smaller-scale world. White said it's his opinion that Red Hat is no longer "in denial" about Linux on the desktop. "They are now genuinely working with their customers," he added.

Bruce Perens, executive director of the Desktop Linux Consortium, sees a similar shift. "2004 is the year we think there will be serious, large scale deployments on the desktop," he said in a separate session. "Eighty percent of people use their desktops [just] for office apps, web access, and e-mail. Those people can use Linux."

However, while the road to the desktop is clear, 2004 will not be the year of the laptop.

"We are still a year away from having a kernel with good ACPI capability," Perens said. (ACPI is the advanced configuration and power management specification through which an operating system controls a laptop's hardware to extract maximum battery life.)

Even while Linux proceeds apace at the low end, the computer scientists who put together the world's fastest hardware are embracing the OS for their multiprocessing servers and supercomputers. While such hardware is largely used to scientific and engineering applications such as oil exploration, drug design, and crash analysis, it provides a picture into the kind of computers IT managers could be working with in as little as two years.

Most applicable may be the huge database apps now seeing service in the high-performance computing arena. Lessons learned there are already impacting emerging e-commerce applications.

Indeed, for IBM, Linux is seen as a potential step toward grid computing, where large conglomerations of computers cooperate over networks. Dave Safford, manager of global security at IBM Research, sees grids enabling what IBM characterizes as "e-business on demand."

Getting a handle on high-end databases is important, Dave Parry, senior vice president and general manager of products and platforms at Silicon Graphics, told LinuxWorld attendees, because "data sets are growing exponentially."

Parry characterizes this as the "data explosion" problem. On the plus side, as techniques emerge to manage data, the result will be that users begin to get "pervasive access" to all angles of e-commerce transactions and reporting functions.

In an early test toward such goals, SGI built a prototype 512-processor system out of Intel 64-bit Itanium processor, running Linux. Parry said the system's performance beats traditional vector-scalar supercomputers. "We really believe Linux can and does scale and we can do things with it you can't do with other systems," Parry said.

Interestingly, Parry doesn't necessarily see Linux at the high end as something that will steal thunder from Microsoft. "The biggest losers may not be Windows," he predicted, but rather proprietary Unix implementations like Sun's Solaris, Hewlett-Packard's HP-UX and IBM's AIX.