Most people will agree that having access to source code is a valuable asset. Even better - royalty free source code. Companies just love it - no development costs, and nothing to eat into your margins, no matter how much you sell. Academia embrace it with open arms. Students can delve deep and understand essential features of an RTOS. You can experiment to see what happens when you incorporate new techniques or features. It provides fertile ground for research projects.

Indeed, the two main initiatives to create a real-time extension of Linux originate from the academic R&D environment: RT-Linux from the New Mexico Institute of Technology, and RTAI from the Politecnico di Milano.

For a company with money to make or lose, and a reputation to make or break, what is vitally important is to have a robust, well-documented RTOS that can be used - and re-used - with every confidence in each new product. Achieving significant levels of confidence in the underlying RTOS allows the company to reduce new product development costs, enhances the trust of their clients that the product they buy will perform, and in so doing, reduces new business development costs. Nothing is worth more than customer loyalty.

A much sought-after advantage of open source software is that the RTOS is free from run-time royalties. However - you know the old adage - "there's no such thing as a free lunch". Beware. The use of such a "free" RTOS can cost you dear in other ways. You are likely to incur extra development costs as you attempt to unravel the inner (darker) secrets of something where serious productisation would have got in the way of serious R&D. Be warned - one must pay close attention to ensuring that what will be saved in royalties will not be drowned under remedial engineering and development costs and worse - lost customer loyalty.

Dedicated Systems Experts has been able to carry out a comparative evaluation of real-time Linux variants against RTOS products already established on the commercial market. The evaluations clearly show that both RT-Linux and RTAI lack several fundamental features required to qualify as a "real" RTOS.

For example, both real-time extensions of Linux use a scheduler with a single linear queue. Tasks that enter the ready state are added at the end of the ready queue, regardless of their priority. Consequently, when the scheduler has to fetch the next task from this queue, it must traverse the complete queue, looking for the task with the highest priority. Such a mechanism obviously results in a task switch latency that is not bounded - and therefore lacks a crucial key factor - predictability.

In view of this and other results - which we will publish in a forthcoming issue - it is astonishing to observe that several RTOS support companies are switching to RT-Linux as their primary product.

You can wager good money that their first clients will pay the price, through incurring, in all likelihood, significant remedial development costs for adapting the chosen real-time Linux to their requirements.

When the resulting code is again made available to the open source community, real-time royalty-free Linux may in the end evolve into a viable alternative to royalty-bearing RTOS. Until then however, companies adopting real-time Linux for the development of new products, risk paying a heavy price as significant remedial engineering overhead is incurred - and as windows of market opportunity sail right on by.

As of today, we are unable to classify this code as a commercially viable product. There are no clear signs that it is evolving in any way comparable with accepted professional practice of software design, implementation, test and release. Where is the source code control, where is quality assurance, where is …. ..? Think twice before you start using this. No. Strike that. Don't even think about it.

Dr Martin Timmerman
Contact

Dedicated Systems Magazine in general