Desktops

Sun's next goal: A Linux ecosystem

A top exec says Sun is "dead serious" about an open-source push for its Sparc chips. Critics say it's trying to save an endangered species.

Sun Microsystems' ambitions have grown another size larger.

The server and software company launched its servers based on its own UltraSparc T1 "Niagara" chips in December, a major part of a drive to restore its lost luster and financial strength.

But alongside the hardware launch came a more quiet software push: an attempt to make the Linux and BSD Unix open-source operating systems a serious option for buyers of Sparc-based computers. To promote the technology combination, Sun is trying to coax an accompanying software business into existence.

Sun has had some experience building such software "ecosystems." For example, it's in the process of resurrecting a version of its own Solaris operating system for computers with x86 processors such as Advanced Micro Devices' Opteron. But Sun, which already has several irons in the fire, faces formidable challenges in the Linux and BSD effort.

"The time for Linux on Sparc as any kind of major market phenomenon has come and gone--over five years ago now, maybe longer," Illuminata analyst Jonathan Eunice said. "It just serves to split the available development resources."

Through projects such as UltraLinux and Aurora Linux, Linux and some BSD variants can already run on Sparc processors. But the products are not commercially relevant for most potential customers. The two major Linux sellers have already pulled back--Red Hat dropped its Sparc support in 2000, and Novell Suse's last supported version was released in 2002.

Still, Sun has no shortage of gumption. "Linux on Sparc is dead serious," President Jonathan Schwartz said in an e-mail interview. "I'm personally talking to leaders in the community. BSD, too."

The effort is part of Sun's attempt to restore its relevance and financial fortunes by shaking its image as a proprietary technology company. That legacy from the '90s hurt the Santa Clara, Calif.-based company when it missed out on two major growth trends that spanned the rest of the server industry: machines built with x86 processors such as Opteron and Intel's Xeon, and the open-source Linux operating system.

Now, one 180-degree turn later, Sun is making its Solaris an open-source project and plans to do the same with its UltraSparc processor. "To be successful, Solaris has to go beyond Sparc. But also to be successful, Sparc has to go beyond Solaris," said David Yen, who as executive vice president of Sun's Sparc server group is trying to make the chip family "the new industry standard."

A "daunting task"
To make Linux or BSD on Sparc a real option, Sun must hook in outside open-source programmers. More challenging, it must also convince major software companies such as Oracle that there's enough customer interest to make it worth their while supporting the operating system-chip combination.

One company that knows just how hard it is to build new software ecosystems around a particular chip-operating system combination is IBM. Big Blue hired dozens of programmers to make Linux usable on computers using the company's Power processors--work that extended to IBM's software group as well.

"Whether Sun is the lead, or a supporter, is less relevant than answering the question 'Is there customer interest?'"
President Jonathan Schwartz, Sun Microsystems

IBM labels its Linux on Power effort a success, as it attracted some customers and partnerships with several major software companies after years of work. However, it does also have a keen appreciation of the difficulties involved.

"If (software companies) stand to gain additional business, and if customers are feeling a pain that the solution addresses, then the ecosystem occurs naturally," said Karl Freund, vice president of IBM's pSeries Unix systems. "If the market doesn't really want the stack, it could cost about $50 million to $100 million a year."

More skeptics are at Hewlett-Packard, another major Sun rival that has learned from its own ecosystem-building experiences. Faced with difficulties getting Intel's Sparc rival, Itanium, to catch on, HP is underwriting a large portion of a five-year, $10 billion effort to popularize the chip family.

"They're late, they don't have a particular price-performance advantage or any particular reason the Linux market would move there," Don Jenkins, vice president of marketing for HP's Business Critical Server group, said of Sun's move. "It strikes me as a pretty daunting task."

But Schwartz said he's not worried about the investment required to build two new software ecosystems in addition to the current Solaris-x86 work. Sun's Niagara chip will do much of the proselytization work for the company, he argued.

"That's 'old think,' " he said when asked if Sun had enough energy and resources to build Linux and BSD ecosystems for Sparc. "Open-source communities are a much bigger player today than vendors in creating ecosystems. So whether Sun is the lead or a supporter is less relevant than answering the question, 'Is there customer interest?' Given the 5-to-1 price/performance benefit of running Web loads on Niagara versus Xeon, interest exists from a broad variety of customers to migrate existing Linux-Xeon deployments over to Linux or BSD on Niagara."

Potential partners cool
The gateway to commercial Linux today is through Red Hat and Novell, which will both need to be convinced of Sparc's merits.

"We have no plans to port Red Hat Enterprise Linux or Fedora to Niagara," Red Hat said in a statement.

Novell's statement was equally cool on the idea: "We have no current plans for a Suse Linux port to Sparc. We stopped building it on SPARC after version 7.3, due to decreasing customer demand."

Schwartz is realistic about the partnerships. "Novell and Red Hat are businesses--the onus is on Sun to make a port to Niagara a compelling value proposition," he said. However, other Linux distributors are showing more interest: Discussions are "going far better with Debian and the CentOS community," he said.

Oracle, arguably Sun's most important software partner, didn't rule out support, but it said in a statement that customer demand is a prerequisite.

"Oracle continues to deliver its products on a variety of operating systems, including Linux, Windows and various flavors of Unix. Our decision to support existing and/or new operating environments will always be based on demand from customers," the business software maker said.

Seeding the market
Although Sun doesn't plan to work on the operating systems itself, it is trying to seed the market. The company is giving hardware to David Miller, leader of the Linux on Sparc effort and a Red Hat employee, and a handful of others, said Mike Splain, chief technology officer of Sun's Sparc server group.

"With the Linux community, we are farther along. With BSD, we're still more feeling it out," Splain said.

Niagara should be a powerful draw, he said. The chip has eight processing engines, called cores, and each one can execute four simultaneous instruction sequences, called threads. That 32-thread total contrasts with a maximum of four in other processors today, but other chip designers are moving toward designs with more threads.

"We believe the Linux community has lots to gain by getting on the Niagara-style computing bandwagon," Splain said. "There is no doubt all the other processor road maps on the planet are all going to adopt multicore, multithread architectures. If they can support Niagara, it's good for their business in the long run."

There should be no doubt about which operating system Sun prefers, though. One objective of the Linux on Sparc effort is to let customers "realize the benefits of Niagara and slowly make the transition from Linux to Solaris," Splain said.

But Linux also is a good fit for the UltraSparc T1, he said. It's widely used on lower-end servers that handle routine Internet and Web chores, the same area where Sun aims its new T1000 and T2000 systems. That difference separates Sun's effort from IBM's Linux on Power effort, he said.

"I think the T1 is much more a high-volume processor than the Power processor," Splain said. "My expectation is we'll see higher penetration of Linux and a quicker adoption rate."

Splain said skeptics should look at Sun's track record. "Solaris x64 demonstrates it can happen," he said. "The thing you have to have is realistic time estimates. You can't just throw the switch and say, 'Yeah, everyone is going to come over."