Back to article
Choices in Embedded Linux: Commercial or Roll Your Own
Making the Choice
February 27, 2004
Over the past week, Red Hat and Wind River's upcoming embedded Linux OS has generated considerable industry buzz. Yet there are many other commercial solutions already on the market in the embedded Linux space. There's also the "roll your own" approach, relying on "free" open source components. So what's the best route for a developer to take?
The answer is highly individual, industry observers say. Factors coming into play can include the company's financial budget and inhouse technical resources, as well as desired time to market, and the reasons for creating the embedded device in the first place.
In fact, this sort of flexibility is one of the key drivers behind embedded Linux, according to Tim Riker of Texas Instruments (TI). Device-makers can move between packaged solutions and open source componentry. while avoiding getting locked in to a single brand.
A related advantage is that "Linux covers so many CPUs and chips," said David Mandala, also of TI, speaking at the recent Consumer Electronics Show (CES). Linux has been tested for hardware ranging from Intel to MIPS, StrongArm, and Sparc, Mandala noted.
Other frequently cited benefits of embedded Linux, in comparison to other embedded OS, include lower costs, greater reliability, and better capabilities for product differentiation.
However, embedded Linux is a complex development environment, and this complexity can bring hidden costs, others say. "Free software" does not mean that "bits just fall like great manna from heaven," cautioned Bill Weinberg, director of product and technical marketing for MontaVista Software.
Embedded Linux encompasses "a whole lot of code," agreed Rob Lamphier of Real Networks. With its Helix DNA code for desktop and embedded environments, RealNetworks is "striving to (provide) a productive experience," Lamphier maintained.
Many do-it-yourselfers are simply "looking for the cheap way," echoed Craig Hollabaugh, author of the book Embedded Linux: Hardware, Software, and Interfacing. "They see free as being zero dollars."
Yet a "collaborative" approach can also be quite wise, suggested TI's Riker. Companies can "leverage what the (open source) community has done," while also giving back to the community. At the same time, they can "focus their dollars on the pieces that are differentiators for them."
Some contend that, to add to the complexity, the embedded Linux market remains much more fragmented that desktop Linux, which is already dominated by two main distributions: Red Hat and SuSE Novell.
Weinstein pointed to several factors that can prompt a decision for or against embedded Linux: management imperatives; internal Linux advocates on the development side, or, at the other extreme, embedded Linux "horror stories."
Large companies with management imperatives sometimes build embedded Linux "skunkworks" projects, using evaluation packages or older code, Weinstein said. After these projects turn out to be serviceable, "then, that's where we come in."
As one horror story, Weinstein mentioned a large telecom company which has already spent $1.6 million on in-house Linux development without creating a deployable solution.
Yet under other sets of circumstances, "roll your own can be good," Weinstein acknowledged.