Editorial 1Q98
Hypes in the dedicated systems world
After the breakthrough of PCI technology during the last two years and the overwhelming interest in the Windows NT Real-Time Extensions in 1997, we see two other hypes dominating the major conferences and tradeshows in Europe and North America: Real-Time Java and Windows CE.
Java here, Java there, Java everywhere …
The title of our technology evaluation report on the use of Java in real-time (RT) systems refers to the pre-eminent sales argument of Java support on nearly every booth, conference paper and product datasheet. One will never argue about this, Java is in: people want Java, thus the vendors support Java, very simple indeed. But is it that simple? Not quite! While preparing this issue, we discovered how difficult it is to get a vendor to make a clear statement about the use of Java in RT systems. While you find Java support everywhere, only a few companies dare to approach the topic in detail. Shouldn't we all stop dreaming? At this level of the technology, we clearly do not back the use of Java technology in RT environments. We see it used properly only in embedded web servers, also covered in this issue.
Windows CE, not Hard Real-Time, but not for long anymore…
In last
year's fourth quarter issue, I mentioned the major news of Microsoft
entering the dedicated systems market with release 2.0 of Windows CE. After
having been omnipresent at the
Embedded Systems Conference
West in San Jose last year, CE was for the first time dominantly visible
at the successful
Embedded
Systems '98 event in Germany. Over 6 third-party companies were
promoting Windows CE, while Microsoft themselves will exhibit for the first
time in Europe at the new
Embedded
Systems Conference Europe in September later this year.
Real-Time Magazine has already studied Windows CE in detail. Based on a
theoretical evaluation (a detailed technical evaluation will follow) presented
at the Real-Time Systems Conference in Paris earlier this year, we can say
that Windows CE cannot be used "as is" for hard RT applications. This is
because of its low number of available priorities, its unclear interrupt
pre-emption (via IST), the lack of nested interrupts and the interference
with clock routine (which has been clarified by testing), to name some of
the drawbacks.
You are still not convinced? OK, if you still want to use it, you may use
it only for simple soft RT systems with limited external stimuli (limited
IST levels), while keeping the CPU load low and while the critical jobs are
done at ISR level… but why then use CE?
What the "regular" RTOS should worry about more, is Microsoft's ability to
adapt or even rewrite the kernel in minimum time. Indeed, with Windows CE
2.0, Microsoft demonstrated its ability to rewrite the kernel from scratch
in a 12-month timeframe!!! They took into account the real-time specifications
of the
comp.realtime
FAQ, which you also find in html, format on
Real-Time Encyclopaedia.
When comparing the list of what is still to be done to make Windows NT hard
real-time (see our
previous
articles on Windows NT) and what CE still needs to be hard real-time,
we see that the list has shortened! Just add many more thread priorities,
make nested interrupts possible and improve the knowledge of system metrics
in general and you have
Windows
CE for hard real-time systems! Not really difficult for Microsoft…
Windows CE hard real-time may be expected within one, maximum 2 years! In
the meantime we refer to our detailed Windows CE evaluation report, available
soon in the magazine and on our website.
Dr Martin Timmerman