Will the Real Embedded Linux Please Stand Up?

Many vendors of proprietary operating systems are jumping on the Linux bandwagon. Which, if any, of the offerings out there are really embedded Linux?

The statistics make it fairly clear that the embedded market for computing devices is much larger than the standard desktop or server PC market: fewer than 10% of microprocessors are used in "regular" computers. Nor would there be much argument that one of the biggest up-and-coming technologies is Linux. Notebook-toting techies and star-struck analysts alike are eager to explain the potential of Open Source software to break the lock that Microsoft has had on corporate information systems. Indeed, if the number of favorable articles dedicated to Linux products and services is any indication, all our cell phones and microwaves may soon be running embedded Linux.

Given these facts, it can come as no surprise that leading manufacturers of traditional (read proprietary) embedded operating systems are rushing to associate themselves with Linux in an effort to capitalize on the buzzword factor. The reasoning is simple: "If our embedded market is huge, and Linux is the next big thing, we better use Linux for our embedded products." The fact that embedded device manufacturers seem to feel the same way only encourages this type of thought.

Hence the embedded systems news has been filled with recent Linux-centered announcements such as these:

• QNX, a manufacturer of a proprietary real-time operating system, announces its version of Open Source code plus a "high level of Linux compatibility"
• Wind River Systems announces that its proprietary operating system will soon include Linux development capabilities
• MontaVista Software announces improvements in its Hard Hat version of Linux to support soft real-time embedded devices
• Red Hat Software ships its embedded developer's kit
• And in the most brazen attempt of all to capture the Linux buzz, Lynx Real Time Systems changes its name to LynuxWorks.

Of course, Linux really is a great platform for embedded systems. It offers more power at a lower cost than anything else on the market. Consider:

• The Linux kernel is true Open Source software, meaning that you can modify anything in the operating system to fit the needs of your project
• Linux is based on standards like POSIX, making it easy to port standard applications
• Linux is perceived as having no cost or a very low cost
• Linux is a complete system, allowing faster time to market than creating your own embedded solution from scratch
• Companies like Lineo have enhanced Linux with embedded features such as hard (genuine) real-time operating system capabilities.

Behind all of this excitement and flurry of press releases, however, lay two fundamental problems with Linux in the embedded market segment—problems that these companies would rather that no one consider.

1. Hybrid systems that attempt to add Linux to an existing product risk giving up their existing markets without achieving all the benefits that Linux has to offer.
2. Pure Linux-based companies that are rooted entirely in the free software world risk having no viable business model, thus leaving customers without a long-term partner to develop and maintain their solutions.

Traditional embedded systems companies such as QNX and Wind River Systems see stiff competition from Linux on the horizon, so adding Linux compatibility or some Open Source aspects to their product lines is a natural reaction. But these companies can't truly open their source code as Linux does—doing so would give away all their intellectual property and destroy their revenue stream. Yet if they don't attempt to follow the Linux model to some degree, they lose all the network benefits that have propelled Linux to its current position.

The natural consequence of this is that traditional embedded software companies must continue to charge fees based on their proprietary solutions. Those fees may be driven down by the emergence of Linux, but competing on price with Linux is a recipe for disaster for an old-fashioned software company.

Even the idea of Linux compatibility should raise eyebrows among the technically competent. Developing on a Linux host (as Wind River Systems proposes) does nothing to remove the per-device royalty of their proprietary operating system for the target embedded device. Likewise, adding Linux binary compatibility (the ability to truly run Linux programs) to an embedded operating system such as QNX or LynxOS seems like a wasted effort if this application support isn't coupled with some of Linux's greatest strengths, including:

• Broad support for peripheral devices
• Protocol stacks that provide exceptional performance and functionality
• The recognized reliability and stability of the Linux kernel and supported kernel modules
• Support for an increasing number of diverse hardware platforms, many of which are ideal for embedded systems.

Traditional Linux providers such as Red Hat Software also face challenges as they enter the embedded systems market. Linux as originally provided by such companies occupies hundreds of megabytes. Whether they have the technical expertise or the product capabilities required by an embedded device remains to be seen. Red Hat Software, as an example, acquired the eCos operating system as part of its purchase of Cygnus. This non-Linux operating system showed promise, but is not even included in the new Red Hat embedded developer's kit. Red Hat makes no indication of support for popular non-Intel embedded microprocessors, nor of devices such as flash storage and wireless systems that are key to many embedded projects.

Because Linux is rightly recognized as a powerful technology base for embedded systems, developers may be tempted to use a standard Linux version such as Red Hat and pare it down for their embedded device. Such logic, however, ignores a significant advantage that Linux is supposed to provide, and that embedded developers must keep in mind: Time to market. The time spent creating an embedded Linux system from a standard version of Linux can invalidate any cost savings of Linux compared to the proprietary solutions that the developer was trying to avoid.

But the biggest problem with the free-software fans entering the embedded market is their business model. The idea behind free software is to charge for services, not products (though the license allows for both). Embedded developers, however, must rely on the Linux vendor to track changes in Linux, provide knowledgeable technical support, and assist in developing solutions for future projects. In a company where the free software, no-royalty model prevails, revenues are unlikely to be sufficient to sustain the business. Developers are soon left without any vendor to back them up. Only a business like Lineo, one that bases product sales on relatively low royalties, can provide the benefits of Linux technology to embedded device developers along with a long-term partnership built on a sound business model. The revenue stream then funds continued development of Linux on chips used in new embedded devices—chips that standard free Linux doesn't support; it also provides for quality tools that make every developer's life a lot easier.

We will continue to see Linux hype from everyone involved in the embedded systems market. Existing embedded vendors are too threatened by Linux to dismiss it, and existing Linux vendors are too enthralled by the size of the embedded market to ignore it. But a little wisdom is called for. Vendors of proprietary solutions can't have it both ways: both the technology benefits and low cost of Linux and the revenues of their traditional technologies. Nor can Linux vendors have it both ways: both a pure free software model and a viable long-term business. Only a carefully crafted strategy that combines Linux technology with a workable business model should be considered as a viable embedded solution from the view of both technology and a sound business relationship.