Intel bets on open source future

To stay one step ahead of the competition, Intel is investing heavily in new software strategies of which open source is a fundamental part

At the beginning of this year Intel decided to get behind an open-source project called Xen, the end result of which is to develop a virtualisation engine — or hypervisor — designed to allow multiple operating systems to run on a single server. The project came along at the perfect time for Intel — the chipmaker is building Virtualization Technology (VT) into chips beginning this year, but Windows won't support the feature for two or three years.

Participants in the Xen project — also backed by Sun, HP, Novell, Red Hat, AMD, Voltaire and others — claim Intel's contribution has been substantial. "The [open source] community is very grateful, and the market too — since Xen can exploit both existing and new Intel equipment to deliver a unified virtualisation solution," says Simon Crosby, vice-president of marketing at XenSource, a start-up that develops and supports the software.

Xen isn't an isolated example. Intel is beginning to take an increasingly active role in helping open source projects such as Xen and GNU/Linux to flourish. The chip-maker may not be angling to become an open source hero along the lines of IBM, which a couple of years back trumpeted a $1bn ($600m) investment in Linux, but the developments amount to a recognition that open source software is as critical to its business as proprietary systems such as Windows.

In the bigger picture, software of all sorts is becoming fundamental to Intel's business. This has become even more the case with Intel's recent shift from focusing on individual components to selling entire platforms, and with the introduction of new features such as Active Management and VT that rely on software.

In the meantime, Intel has become one of the world's biggest software developers, employing more than 5,000 software engineers. "Increasingly things are more about software than hardware for Intel. It's not an addition to the platform, it's an increasing part of the value of that platform," says Gartner analyst Brian Gammage. "And they know they are not going to see that value delivered if they restrict the support of their utilities to one Microsoft operating system."

Developing markets
Linux is becoming increasingly big business for Intel, both on enterprise servers — where it has become an entrenched presence — and potentially on the desktop as well.

But despite the momentum that companies such as Sun and Novell have tried to instil in the concept of Linux as a desktop OS, some experts claim that in mature PC markets, such as Western Europe and North America, it still isn't a serious proposition. But they could be missing the real market.

Linux workstations present a far more attractive option in emerging markets such as China, where the labour costs of migration are lower, Windows isn't so entrenched and price is more of an issue. "There is a growing realisation that the future growth of the market is moving away from the mature markets, into the emerging markets," says Gartner analyst Brian Gammage. "With the new users there, cost is absolutely determinant. And labour costs make Linux migration more appealing in emerging markets."

Gammage sees this trend as the motivator behind Intel's recent moves to begin treating Linux on a par with Windows where it comes to equipment makers. For example, the chipmaker recently centralised its Linux efforts into the Linux Program Office, part of the Software and Solutions Group, the same division that coordinates Intel's contributions to other open source projects. Intel also recently hired Dirk Hohndel, a well-known figure in the Linux world, as director of its Linux and open source strategy.

On the sales channel side, Intel recently introduced a kit designed to make it easier for manufacturers to put together Linux desktop systems. The Quick Start Kit for Linux mirrors similar efforts that have long been in place for Windows, and is aimed at China, India, Latin America and other emerging markets.

The Software and Solutions Group's Channel Software Operation, which is behind the Linux kit, has a Linux expert of its own in the form of Danese Cooper. In March Cooper was hired away from Sun, where she had served as the company's resident Linux advocate; she is now director of open source strategy for the Channel Software Operation.

Intel and industry observers stress that such moves aren't a radical shift for Intel — rather, they're a continuation of Intel's long-standing relationship with the open source community. And they don't mean Intel is showing any special preference for Linux, either — it's all about sales. "This is all in the context of Intel's need to sell as much hardware as possible," Gammage says.

New technologies
Software has always been important to Intel, but it is right at the core of Intel's latest big strategies, its shift away from components toward platforms, making computers easier to manage and virtualisation. The success of these strategies is a life-or-death matter for Intel — in an age of constantly dropping prices and saturated markets, it is relying on these new directions to keep buyers interested.

Intel announced in the first part of this year it would switch its emphasis from individual chips to platforms such as Centrino, the integrated mobile computing set of chips. The move is significant for Intel, which for years relied on raw clock-speed figures to sell its processors. This so-called "platformisation" shifts the sales strategy away from ever-increasing performance to how well integrated and easy to use the system is — something particularly important in mobile computing, where battery life and system set-up are big issues. And that integration depends on software.

An integrated hardware platform makes things more complex for software developers, because instead of interacting directly with a chip they're interacting with the embedded software that controls the platform. This extra level of complexity has made Intel's own software efforts more important, since Intel needs to work with software makers more actively to ensure that there's broad support for the hardware features.

With Centrino, for example, Intel took a more active role in developing Linux drivers than it ever had before, collaborating with open source developers and publishing the Centrino Wi-Fi drivers under an open source licence. Though the Linux drivers were completed long after Windows drivers had arrived, the move showed Intel was willing to cooperate.

Previously, open source developers had been left to do the detective work themselves. Such collaboration made sense from a business point of view, because it meant those running Linux laptops could be added to Windows users as potential Centrino customers.

The first of the new generation of platforms is the excitingly named Professional Business Platform (PBP), announced in May, which introduces two technologies that are key to Intel's future: Active Management Technology and VT. Like Centrino, PBP is a blueprint for a PC based on a bundle of Intel chips that have been pre-tested to work together, the idea being to eliminate any potential incompatibilities and highlight the way the bundle works as a whole, rather than the performance of individual parts.

Where Centrino emphasised power management and integrated Wi-Fi, PBP is designed to make desktops easier for enterprises to manage, and to wring more performance out of the silicon. Active Management, embedded in the chipset, allows administrators to remotely manage desktop machines over the corporate network, and quickly remove them from the network if they're infected by a virus, for example. VT allows a system to function effectively as two separate machines, running two separate operating systems, with little or no loss of performance.

Future plans for PBP include adding new features such as wireless chips, and a possible Centrino-like rebrand and marketing campaign in 2006. Intel is also planning further platforms aimed at business and home desktops.

The payoff for Intel: if businesses are spending less money on maintaining their systems, that frees up budget for capital expenditures like new desktop hardware. The new technology also gives companies a reason to invest in new hardware, in a saturated market where performance increases have started to seem less relevant.

Much of Intel's work with open source is geared towards making this strategy work by ensuring the new technology is supported by Linux — indeed, features like VT are likely to see Linux support long before they show up in Windows.

That's where projects such as Xen come in. Xen started off as an obscure project at Cambridge University, designed to allow multiple operating systems to run simultaneously on a single computer, a field dominated in high-end servers by EMC's subsidiary VMWare. VMWare's software runs each operating system in a separate virtual machine, a complete hardware simulation; this allows the operating systems to run unmodified, but at the cost of some performance.

Xen uses a process called paravirtualisation that eliminates the virtual machine, allowing much better performance — typically around 2 percent compared with a typical 20 percent performance penalty with VMWare. However, paravirtualisation requires that the operating system be ported to run on Xen, something that isn't possible with proprietary operating systems like Windows. Application-level software doesn't need to be modified.

Intel's contributions to Xen mean that Xen running on VT-enabled Intel chips will be able to run operating systems without modifications; they won't run as fast as modified software, but should still be significantly faster than those running in VMWare. (Xen is also working on support for AMD's virtualisation scheme, code-named Pacifica, which is due next year.)

"Intel has worked extensively with our team to develop and integrate the VT extensions for Xen," says Xen's Crosby. "As such, Xen will be the industry's first supported code base for VT, when VT processors ship."

Intel's attitude towards open source has been "extremely positive", with contributions including support for EM64T — Intel's 64-bit extensions for x86 — and porting Xen to IA64. "Intel's work on VT in particular was a major contribution to the [open source] community, since VT itself has not even shipped, and the feature set is close to Intel's heart," says Crosby.

Without Xen, Intel might have to wait another three years for VT support to appear in Windows, says Gartner's Gammage. "VT technology is arriving imminently, and Pacifica is on the way early next year. Both Intel and AMD are keen to see their developments producing some return, and that's not going to happen without software support. We're unlikely to see it in Microsoft until 2008," he says.

Xen is already supported in some form in SuSE Linux Professional 9.3, Debian and NetBSD, and Red Hat's Fedora Core 4 is to ship with Xen included. Xen version 3, which began testing in June, will add multiprocessor support.

Hardware security
The flip side of platformisation is that some of Intel's planned systems put an emphasis on hardware-supported security systems that would be difficult or impossible for open-source software to support. For several years open source developers have been keeping an uneasy eye on Intel's LaGrande hardware security system, the counterpart to Microsoft's widely criticised Palladium plan (now known as Next Generation Secure Computing Base).

Now aspects of LaGrande, such as the use of Trusted Platform Modules to enforce copy restriction, or digital rights management (DRM), are to be integrated into an upcoming consumer-oriented desktop platform known as East Fork.

Such developments have been met with dismay by the open source community. "We're wary of oncoming developments like 'East Fork', which we consider bad for consumers and dangerous for independent developers," says Eric Raymond, a noted open-source developer. "Universal DRM would be a disaster for consumers; they would no longer control their own machines, and the competitive pressure that Linux puts on Microsoft and the big media companies would end." However, Raymond notes that so far Intel has put such technology on the market without trying to force people to use it, and may continue down the same path.

Until now, Intel's main contribution to the open source world has been to commoditise 32-bit hardware, which gave the open source community an inexpensive, open standards platform to build on, Raymond says. He thinks the company's recent, more active approach should be welcomed, though with a few reservations. "I think we'd be quite happy with Intel as a neutral, simply supplying chips and open specifications," he says.

"Mostly what we want from them is that hardware prices keep dropping, and that they not collude with bad actors like Microsoft or the MPAA, RIAA and DVDCCA, who want to make open source impossible."