UNIX has become such a pervasive component in the world of computers and networks that it would be difficult conceive of a computer industry without this open source operating system. And the prime factor that has propelled it to this status is precisely the open source nature of its development. Originating at Bell Labs when it was part of the officially sanctioned telephone monopoly, UNIX was not introduced as a proprietary product and was therefore open for development by a broad cross section of the programming community. Despite much technical criticism over the years and the presence of many proprietary competitors, the momentum of open source development proved to be the critical factor in UNIX's dominance.
This scenario is starting to repeat itself with the sudden rise of Linux, again an open source operating system that has been adopted by the software engineering community for much the same reasons as UNIX. Named after Linus Trovalds, who wrote the initial kernel, Linux is rapidly becoming a defacto standard in networking, particularly in server design.
But while Linux has achieved wide acceptance in the system design community, its move into the thorny area of real-time embedded systems is proving to be more problematic. A commonly cited reason for its slow progress in that arena is the fact that the Linux kernel was not designed for real-time applications to begin with and therefore suffers inherent performance disadvantages.
One industry veteran of embedded system design, Jerry Fiddler, co-founder of Wind River Systems and a contributor to this week's In Focus, believes that the problem is much deeper than that — stemming instead from several fundamental aspects of embedded systems."The calling card for open source software is low entry cost--you don't have to pay for the operating system software," Fiddler said. "But it turns out that most of the cost of software is in subsequent development, not in the initial price of the operating system." Because Linux is not owned by a single company dedicated to its success in the market, finding the right tools and support for developing applications could turn out to be a costly process in itself. No matter how the initial operating system software was acquired, a software company will still have to bear the cost of software integration, development, training and maintenance, Fiddler pointed out.
Embedded system design, by its nature, only makes the problem worse since these software systems are tailored closely to the demands of the hardware. In contrast, desktop or server software can usually make use of most of the development work from previous efforts. Because computer systems in general use only a few different processors all of which carry fairly standard feature sets, software development costs have not retarded the progress of Linux in this major area.
In fact, the progress of Linux in the server market has been unprecedented for an open source system, Fiddler observed. The operating system has displaced offerings from major companies such as Sun Microsystems, IBM Corp., Hewlett-Packard and Microsoft Corp. There are signs that it will be a success in both the desktop and high-end scientific computing arenas as well. Windowing interfaces for desktop Linux systems are appearing and supercomputer vendors are starting to develop clustered processor systems that are tied together with Linux.
Enthusiastic backers
While the world of embedded systems is much more fragmented than the server market, the growing enthusiasm of embedded software developers may be enough to push Linux into the lead, according to Kevin Morgan, vice president of engineering at Monta Vista. "Linux, as an embedded operating system, is going to become the unifying and dominant OS in a way that has not happened in this industry ever," Morgan said. "Classically, the embedded world has been very fragmented when it comes to operating systems--about 50 percent of embedded operating systems are built in-house and of the remaining 50 percent, you have extreme fragmentation with the dominant vendor — Wind River — holding 15 percent market share overall. Well, that is not dominant." Morgan sees Linux becoming truly dominant in the next few years, capturing at least 50 percent of the market and eventually getting to 75 percent market share.
His reason for believing in this type of momentum is the availability of all aspects of the OS. Embedded software developers can get their hands on source code allowing them to optimize and customize the system from within. "When you can't do that kind of thing in embedded development, you are at a competitive disadvantage," he observed. In addition, there is a large pool of programmers out there who are already familiar with the system. "Kids coming out college know this operating system and they know it inside and out. In contrast, with proprietary OSes, it is difficult to find someone with that kind of inside knowledge," Morgan said.
Another factor is the pervasive availability of Linux on leading edge hardware. "All new hardware coming out today from any vendor has to have Linux enabled on it," he said. "Linux is too key an operating system — no matter what market segment you are targeting not to have Linux enablement." In the past, embedded developers had their hands tied while waiting for proprietary operating systems to appear on new hardware.
Contributors to this week's In Focus offer very different views on Linux — with discussions ranging to its benefits and opportunities for embedded developers to the concerns of Telecommunications Equipment Manufacturers (TEMs) about building and deploying on Linux to the hidden costs of GNU public license (GPL) code and whether Linux is in fact the right answer.
