• Calling All C++ & AI Game...
• SAS Analyst
• Technical Lead
• Business Development Manager...
• 1st / 2nd level Help Desk -...

Linux breaking out in clusters

Dan Neel and Ed Scannell, Information Age

12/10/2001 17:12:27

Open-source Linux operating systems have graduated to take on mission-critical tasks within computer networks faster than any other operating system. Fuelled by a worldwide community of developers, experts predict that by 2005 continued progress will fine-tune Linux into just another mainstream OS option for the enterprise.

Clustering and availability

For Linux to grow from adolescence into adulthood in the corporate world, the open-source development community must deliver more capabilities that centre around high availability and the clustering of multiple Linux computers together as a stable system that can be managed as a single computer.

Proof of Linux's rapid maturation in the enterprise is reflected in a recent survey by IDC. According to IDC, more users deployed more Linux-based mission-critical applications in 2000 than the previous year. For instance, 20 per cent of users surveyed said they had deployed Linux databases in 2000, up from 10 per cent in 1999. Ten per cent of users also said they were running at least one mission-critical application on Linux such as an ERP, CRM, or HR application last year, up from 4 per cent in 1999.

Steve Prather, vice president of network services at ViaWest, an Internet service company, has been using Linux successfully for more than three years.

"We've seen very phenomenal reliability and very strong application acceptance and interoperability," Prather says. "There are so many companies out there that are writing Web-enabled content and/or applications to run on different systems, and Linux allows us to have a very broad knife when it comes to people being able to run customer-unique applications. It works very, very well for the multitude of different platforms that are out there."

Although Linux currently appears more at the front end of a network than in the application or database layers, 20 per cent of the respondents to IDC's Linux survey said they were supporting database software on their Linux system.

"Our surveys show that Linux [on servers] has achieved mainstream status in a couple of markets, but not in most. Where it is successful typically involves work being done by engineers, scientists, analysts of some kind, or Web-related applications," says Dan Kuznetzky, vice president in charge of infrastructure and platforms at IDC.

Kuznetzky says Linux falling short in that database has little to do with the availability of Linux database software.

"If you want Oracle [Linux], it's there. If you want Sybase, it's there. If you want Informix, it's there. If you want [IBM's] DB2, it's there," Kuznetzky says. "There are also three totally open-source database products that are used quite heavily, including MySQL and PostgreSQL."

Certain clustering technologies that provide for system fail-over are also available, but beyond that lies the current challenge for the Linux community.

"There are a few things out there like TurboLinux's Cluster Server and the Beowulf APIs, but there is almost nothing in the cluster management or monitoring area. Linux needs something that can make multiple servers appear as a single image," Kuznetzky says.

It takes a community

To speed the progress of Linux clustering, Compaq last June released its SST (Single System Technology) into the open-source community.

"The SST project is a single-image clustering technology that allows you to manage all these multiple [compute] nodes as one server," explains Judy Chavis, manager of Compaq's corporatewide Linux program office.

Sporting technology with a 30-year lineage of running computer clusters from Tandem as well as Compaq's own True64 Alpha clusters, the open-sourced SST is free to be picked up and developed into a product by anyone. But the necessary time needed to develop an actual clustering product from SST leaves any completion date open-ended, Chavis says.

"Every customer has a set of applications - it may be their database application, their transaction application, their ERP application, their Java development tools - and they're looking at that entire stack and they're trying to figure out in that stack how many pieces are really ready for Linux," Chavis says. "We all want to equate Linux to Solaris, or AIX or True64, which are grown 40-year-old men compared to a 10-year-old Linux operating system, and expect it to have the exact same functionality in its infancy. But there are still pieces that the community, Compaq, and bigger ISVs are addressing."

Commitment

What figures to further fuel Linux clustering fortunes is IBM's commitment to it. As part of the $US1 billion IBM promised to invest this year alone in its own Linux products and those of the community, Big Blue is investing generously in clustering technologies and products.

Much of IBM's muscle-flexing in the clustering arena has been contained to university and other research centres including the Universities of New Mexico and Illinois and the Maui Supercomputer Center. But the company's clustering products are now starting to creep into commercial markets.

Earlier this year IBM announced a major corporate win with Shell Oil, in which the company will install a 1000-processor Linux cluster. The company has scored other smaller wins in the oil market, signing deals to install clustering solutions with Chevron and Western Gecco.

"Clustering is one of the two top areas where we are going to make additional investments because we are starting to see the technology expand significantly," says Steve Solazzo, vice president in charge of Linux products at IBM.

By year's end IBM intends to deliver a set of "predesigned, preintegrated, factory-installed clustering systems," according to Solazzo, that will be aimed at corporate users.

"What we are really able to do now is build supercomputer-class machines with merchant [clustering] technology," Solazzo says.

Open for business

An unexpected marketing opportunity arose for Linux earlier this year when Microsoft announced a revamped volume-licensing plan. Some corporate users pushed back against tactics forcing them to upgrade to current versions of Microsoft OS and applications so as to benefit from the new volume-pricing models.

"I think Microsoft's attempt to clamp down on its users with the new licensing model and the registration of products has made Linux more attractive to us from a cost standpoint," says David Orin, a database administrator at a large Nebraska-based food supplier.

"Hard times are good for open source because IT managers are looking for lower price points. And being able to deploy a whole lot of systems without having to buy a licence for each appeals to them," says Bruce Perens, senior Linux and open-source strategist at Hewlett-Packard.

"People used to working with Unix can sit down at Linux, and it is much less costly both for the tools and the operating system."

What Linux can do for the corporate bottom line has contributed to the changing demographic of those making the decisions to deploy it.

"The previous level of adopters were largely technical people with technical titles," says Dan Kuznetzky, vice president in charge of infrastructure and platforms at IDC. "But we're now seeing Linux appear in all sizes of companies - small, medium and large - in all vertical markets that we're tracking, and more and more the people bringing in Linux have titles like director of IT, CTO, and other management titles."

Racing for dominance

By Ana Orubeondo

Vendors are making headway, but Linux still faces a long road ahead to success in the mobile marketAfter a long- and finally successful - struggle to be taken seriously as a heavyweight enterprise operating system, Linux may now be ready to stake its claim in embedded-market territory. After all, Linux is free, well-accepted in the development community, full of stable development tools and drivers, and relatively compact in size - a huge consideration when it comes to mobile devices with limited constraints.

Moreover, a dominant OS has yet to emerge in the embedded market, and several vendors, including Microsoft, Palm and Symbian, offer highly effective mobile products and development tools. As vendors duke it out for dominance, IT leaders and their users must in the meantime commit to proprietary solutions that do not inter-operate very easily.

As both the embedded market and the travelling work force continue to grow, this lack of standards could take a serious toll on companies struggling to keep their workers connected. In a best-case scenario, non-interoperability between devices necessitates inconvenient file reformatting, similar to the process of preparing a Macintosh document for use in a Windows environment. But in a worst-case scenario, this inability to work together entirely precludes any communication between operating systems, making it difficult - if not impossible - for employees to collaborate from disparate locations.

Complicating matters further, the simultaneous adoption of multiple operating systems requires programmers to learn several different sets of APIs and forces employees to master many interfaces. And guess who gets to pay the bill for all this training?

Enter Linux, the humble open-source operating system that may still emerge as the hero of this standardisation crisis among embedded systems. Because it allows programmers to view - and alter - its application code, Linux grants ultimate power to users instead of vendors. Better yet, it enables developers to create customised versions of the operating system. And of course, Linux has no licensing fees, which removes yet another brick from the wall of proprietary domination.

Embedded hero?

Although eliminating costs and exposing code may level the playing field a bit, Linux must actually become the market favourite to become a true standard. After all, if most enterprises still pay for proprietary systems, what difference will it make that Linux is free and forthcoming with code?

Until recently, Linux's lack of RTOS (real-time operating-system) capabilities has blocked its path to mass acceptance. An RTOS, which immediately incorporates new data into active applications, is frequently used for programs that require quick responses to external events, such as live audio.

An RTOS can also prioritise tasks, interrupting functions of secondary importance to direct energy toward more critical processes. Because it helps prevent system overloads and failures, this prioritising is crucial for mobile devices, which are not powerful enough to withstand repeated crashes. When desktops fail, they usually require a simple, albeit inconvenient, reboot. But the malfunctioning of a mobile OS can render data unusable or even destroy the system altogether.

Because of this, most embedded OS vendors offer RTOS capabilities, a fact which has previously been bad news for open-source languages. Linux is built to perform commanded tasks from start to finish without allowing interruptions or accepting new commands midstream, so it has been unable to offer the benefits of real-time computing.

Until recently, the only way to couple Linux with RTOS features was through the dual-kernel approach, which uses a general-purpose, nonreal-time OS running as a task under an RTOS. The general-purpose system handles communications, disk read-and-write operations, serial and parallel I/O, and memory management. Meanwhile, the RTOS handles just those functions associated with quick response, such as interrupt management. The dual-kernel method provides both compatibility with standard Linux and the capability of continually incorporating new information.

Although this method is effective, it demands a great deal of engineering expertise and time. And because it requires two OSes, it is often easier for companies to simply implement the RTOS alone, foregoing Linux compatibility altogether.

Keeping it real

Companies such as LynxOS RTOS and QNX Software Systems Ltd. have recently begun producing real-time versions of Linux, which offer all the benefits of the dual-kernel approach but without the headache of layering two OSes on top of each other. In short, these companies have placed Linux squarely in the running for the embedded-market title.

And the competition has taken heed, occasionally going so far as to emulate the open-source upstart. For example, Microsoft recently allowed free downloads of its OS products by giving developers free access to its Windows CE OS for 60 days and reducing the licensing cost.

Although Linux is now a legitimate contender in the embedded-systems world, it still has a long road ahead. Symbian, Palm, and Microsoft each boast considerable shares of vendor support and processing power, and Linux does not offer the same range of applications as its competitors.

Only one thing is certain: In a few years, a dominant player will rise to the top of the embedded market and will be near-impossible to overthrow. If Linux (or any other OS) is to become to mobile devices what Microsoft Windows is to desktop machines, the time to take the lead is now.

THE BOTTOM LINE

Linux in embedded systems

Executive summary: Thanks to its open-source nature, Linux has the potential to standardise the highly fractured embedded systems market. Moreover, Linux has proven itself as an enterprise-worthy operating system for server-based applications. But despite the apparent benefits, Linux's current lack of applications for mobile devices may delay adoption in the corporate world.

Test centre perspective: Linux's changeable code allows for customisation, but its RTOS capabilities are still not mature enough to warrant serious consideration as an alternative to established handheld operating systems such as Windows CE and the Palm OS.

.