Celebrating Commercial Software

The growth of commercial real-time operating systems (RTOSes) has enabled the revolution we've all enjoyed in snazzy leading-edge consumer electronics products, informative and entertaining automotive electronics and sophisticated, life-saving medical devices. Leading manufacturers like HP, Casio, Sharp, Sony, Bose, Philips and Welch Allyn are enhancing our quality of life with product after product that use modern commercial microprocessor and software technology.

Years ago, such products were less common and often ill-fated, as the electronics industry – particularly the software sector – tinkered with home-grown solutions and do-it-yourself (DIY) development tools. The result was unreliable products, overdue development schedules and costly staffing that reinvented the wheel over and over again.

Fortunately, over the last 20 years, software development has made dramatic advances that have skyrocketed productivity and propelled the electronics industry into a key role in everyday life. Paramount among these advances are the tools available today for software developers that enable them to be much more productive than in the "dark ages," during which they routinely crafted the tools and software they used in the products they were building. Since the early '90s, managers would be fired if they endorsed do-it-yourself development for software or hardware that could be purchased off-the-shelf for less cost and significantly less resource consumption. In today's economy where development costs are under such close scrutiny and project schedules are more and more demanding, DIY development is even more a thing of the past.

Despite this advancement in thinking with respect to software development tools, embedded developers continue to be divided as far as the best way to satisfy their applications' needs for real-time control, scheduling and interrupt management. Although commercial RTOS products are plentiful and support dozens of microprocessors, a dialog continues encouraging developers to do-it-yourself.

Interestingly, although the dialog continues, commercial RTOSes have been used in the most successful and ubiquitous electronic products. The call for developers to shun commercial systems appears to come from a strong DIY advocacy of independent developers who are (justifiably) proud of their work and who want to share it with the community. Obviously, there will always be some engineers who prefer to do it themselves, as opposed to licensing a commercial product, just because they can. Typically, however, such engineers are not engaged in commercial product development, but are academics, consultants or part-time RTOS hobbyists.

While DIY makes sense in an academic setting where new paradigms are being explored and examined, it is less effective than a commercial RTOS in leading to successful product development and production. According to industry analysts at Embedded Market Forecasters (EMF, www.embeddedforecast.com), who have surveyed hundreds of embedded developers for the past several years, developers using a commercial RTOS have a greater likelihood of finishing their development project on time than those using a DIY approach.

DIY proponents raise concerns regarding commercial offerings, including that such RTOSes are too big, too expensive and not suitable for many applications.

Are commercial RTOSes unsuitable for any applications?

If commercial RTOSes were unsuitable for any significant number of applications, we'd see DIY RTOSes used in those applications resulting in successful products brought to market, but we don't. Commercial RTOSes (let's include commercial Linux distributions in this category, making it actually "non-Roll-Your-Own RTOS") dominate most commercial electronic products developed in the last 10 years, which proves their suitability for any application.

There's also no denying the ubiquitous nature of the commercial RTOS as sufficient proof of its superiority over DIY RTOS solutions. According to EMF, in 1995, only 33 percent of applications included a commercial RTOS. Today, that number has grown to more than 76 percent.

Do commercial RTOSes cost too much?

Analysts have examined the cost of commercial vs. DIY solutions. From 2003 to present, EMF has consistently found commercial RTOSes dominate ROI results with open-source trailing. DIY fares even worse than open source, to the limited extent that anyone reports using a DIY RTOS for a real project.

In the recent white paper "Shoot-Out at the RTOS Corral," EMF shows the ROI advantages of using certain RTOSes over others. According to the EMF survey, only 22 percent of respondents reported using an "In-House" RTOS on their last project. Commercial RTOSes better enable developers to meet schedule and get products to market on time, with 60 percent of projects completed on time or ahead of schedule vs. 54 percent among those who chose DIY.

Notably, not all commercial RTOSes faired the same. EMF surveyed hundreds of embedded developers asking, "Was your last project completed on-time, on-schedule, or behind-schedule?" and "What operating system did you use?" The chart below portrays the results indicating those using commercial operating systems completed their projects on time or ahead of schedule more often than the industry average.

Year after year, EMF data from 2006-2009 showed developers who chose ThreadX completed their designs on or ahead of schedule 75 percent of the time – a benchmark that analyst Jerry Krasner believes is due to "feature glut" on the part of some RTOS vendors. In its white paper, EMF observes "using an RTOS that is overqualified for the application may have a negative impact on time-to-market, design cost and product support ... because additional capabilities, beyond what is required by the needs of the application, make these RTOSes more complicated and harder to use." Readers are encouraged to read the white paper.

In all instances, the killer for DIY and open source lies in the amount of support, including product enhancement and ports to new microprocessors. Both these areas require substantial investment of manpower from an in-house development team. Commercial RTOS suppliers also invest heavily in these areas, but their man-years are amortized across many customers, with each one bearing a small fraction of the cost.

Are commercial RTOSes just too large?

These days, commercial RTOSes are available in sizes as small as 2 kilobytes. Very few applications have such an extreme size constraint, particularly when memory costs have fallen so significantly and processors boast much faster clock rates. As Linux developers are well aware, scaling the open source or DIY RTOS down to this extent is a monumental task. 

Today, commercial and open-source RTOSes are available as complete products, ready-to-use and off-the-shelf, with features and technology far beyond what even the best developer could produce in-house. For example, in the development tools area, we now have open-source compilers from GNU, as well as commercial products from leading vendors like ARM, IAR, Intel and Green Hills. While the open-source GNU compilers produce good code and are freely available, the compilers from ARM, IAR, Intel and Green Hills go far beyond GNU with advanced optimization techniques that consistently produce smaller code and faster execution speed.

In the operating systems area, we see open-source examples such as Linux, as well as commercial products like Wind River's VxWorks or Express Logic's ThreadX. Again, while Linux is useful in many products, research demonstrates the commercial Linux offerings consistently experience development costs that are 16.1 percent less than the DIY versions. That commercial RTOSes from Intel (Wind River), Express Logic and others are found in literally hundreds of millions of products, from cell phones and ink-jet printers to commercial airliners and FDA-certified medical electronics, demonstrates their ability to operate within the size constraints of a wide range of embedded system products.

DIY development delivers great satisfaction once the OS is finally clicking along – making all those nights and weekends of struggles with timing and interrupts seem worth it. Such is the plight of companies just getting started with no funding that have no other choice until they can get the prototype running well enough to invite the investment of a backer, who then insists they migrate into tools that simplify their development and reduce overall labor costs. Obviously, very few developers fall into this category.

Conclusion

In this day and age, the "wheel" is a well-established technology. Even when it comes to RTOS development, there's really no excuse for developers reinventing the wheel in light of the number of commercial RTOS offerings available in all sizes, shapes and flavors. Every embedded system of reasonable complexity requires an RTOS, and developers know it's more efficient to spend their time designing their application than re-inventing the RTOS wheel.

Based on trends, it appears DIY developers are a small though enthusiastic minority and embedded developers – those working on cell phones, cameras, defibrillators and GUI-driven dashboards – are facing project timetables that no longer allow them to take the risks associated with internal RTOS development.

While DIY will continue to thrive among academics, hobbyists and unfunded enterprises, the majority of professional developers will continue to enjoy the many benefits of commercial software and the result will be a continued emergence of cool new electronic gadgets and life-saving technology.

So when you hear someone ponder the possibility of writing their own RTOS, ask them what products they use every day that use a home-grown RTOS. Maybe, just maybe, that will give them cause to consider using a commercial RTOS product.