Editorial 2Q97
Windows NT in a Real-Time System, are you serious ?
Real-Time Magazine covers the major topics in the real-time market. The evolution of technology in this field pushes us to reconsider a specific topic every 2 years. This time, the Real-Time Operating System issue is launched. Two and four years ago we had to print 2 issues on the subject. We have tried to limit it this time to 1 issue, but again this has not been possible. The major reason is the invasion of Windows NT as a potential candidate for embedded (real-time) applications. The market considers Windows NT as a must in an open application due to the availability of cheap hardware, including device drivers, protocol stacks and other software modules. Although I have no problems at all with this, for a lot of dedicated systems, predictable response is needed. The question is then:
"Can I make a predictable system using Windows NT?"
Two years ago, Microsoft claimed real-time performance for Windows NT. We wondered how true this was and started an evaluation campaign. The resulting article, Windows NT and Real-Time, which you will find in this magazine, points out the problems with Windows NT in this context. It has been available on the WEB for a couple of months now and we received a lot of reactions. Positive from industries who appreciated the independent analysis and negative ones from some vendors being upset: the article was considered as a threat for the potential market that they have in mind. These vendors all showed us how they (will) make Windows NT work in a system that needs predictable response. The extensions they all invented are the simplest demonstration (if you don't accept the argumentation in our article) that Windows NT in itself has severe problems in supporting deadlines. Studying these solutions, I found that
the definition of a Real-Time System (RTS) is still not the same for everybody.
The solutions offered to produce a RTS with a predictable behaviour, seem also to differ a lot from one vendor to another. While some consider that having fast interrupt response is sufficient, others are more serious and rather prefer a multi-threading solution. Therefore, I consider it necessary, not only to publish the articles from the different vendors about their solutions, but also to compare the different solutions. As you will see, the vendor articles do not explain in detail the solutions - are they afraid of competition, or are they not sure of their solutions?
This is the track Real-Time Magazine will follow in the future. We will not only collect valuable information of today's technology, but we will also comment and evaluate the available products. This issue, as you have understood already, is devoted to Windows NT. Here we compare the solutions, which will be available soon. Once available we will evaluate them. This evaluation program has started for the next issues that will cover another category of RTOS. The first evaluation concerns VRTX, the next looks at VxWorks; others will follow.
This magazine claims to be of high quality. We are not just referring to the layout, but also to the value of the information provided. Simply collecting articles from different vendors is common practice, though this is not considered anymore by us as THE way to inform you. All this information is available on Internet. To reduce your surfing time, we created the Real-Time Encyclopedia. To enhance the quality of the systems you design, we provide you with the results of our product evaluation team. Stay tuned.
Dr Martin Timmerman