I have a question hanging my head since some time ago and I’m sure you can help me.
I work with QNX for about 10 years, since QNX 4.25 times, and I observing the tendencies changes (?) of the OS segment market (I may be wrong).
Since we work with QNX in a distributed environment, the first sentence from QNX Neutrino home page alarm me:
“The QNX Neutrino RTOS provides a comprehensive, integrated set of technologies to help quickly deliver robust, reliable embedded systems.”
We always knew that QNX is a very reliable and scalable SO, but in 4.25 days I thought in “embedded systems” as an implementation “option”.
Now (perhaps it always was) it seems like embedded systems is the only (this is the central point of my question), segment market that QNX’s people are thinking and the option of great scalability became an issue inherited from the original master concept.
I wonder to myself whether QNX remains the operating system we really need in our distributed environment.
I keep thinking that the answer is yes, but perhaps your experience and point of view can contribute to me in order to understand the course of QNX OS and that apologetic phrase about embedded
Thank you very much!
Regards,
Juan Manuel
It is clearly a question that comes after breathing the fresh holiday’s air
In QNX2 days there was nothing like it. With QNX2 and arcnet networking was transparent and deterministic, and performance could be greatly leveraged easily by adding more nodes. As a QNX print ad of the day said, it was the mainframe you build one node at a time. Every system I worked on in those days was implemented as a distributed network. One big system I worked on (a vehicle tracking system) had maybe 20 distributed nodes.
When QNX4 came out other operating systems had improved networking capability, so there was more competition. QNX added ethernet, which was faster but not deterministic. The network was a little less transparent with fixed netmaps. Computing power had increased, reducing somewhat the need for a bunch of nodes. After a port, the tracking system I worked on went from a 20 node network to a single STD bus box with 4 cpu cards networked across the backplane.
As computing power increased that system was eventually moved to a single pentium box around 1997-98.
So in this case increased computing power largely eliminated the need for the network entirely.
I started working as a consultant in about 93 and have not had a customer since that time implement a new system that required a distributed implementation. On the other hand, I have worked on a whole bunch of embedded products.
From my perspective their decision to deemphasize transparent distributed networking was probably correct, but it’s sort of sad to see an outstanding feature like that die out.
Hi kwschumm, thank you very much for your answer, which still leaves me a bit worried.
In our case, it is by nature a distributed environment:
We must control an industrial process from very distant places (HMIs for operators).
We need a redundant architecture acquisition (industrial atmosphere is very hard and nodes tend to fail)
We need a redundant DB architecture for the same reason. We definitely can not depend on a single node
Our networks are distributed and are to be between 6 and 40 nodes running:
→ HMIs
→ DBs
→ Apache and other services
→ mathematical models,
→ expert systems,
→ acquisition from sever PLCs (from 1 to 15 maybe more different vendor’s PLCs!)
→ other acquisition communication services with other instrumentation equipment.
→ etc. etc. etc.
In QNX 4.25 days all this jobs works perfectly and we were proud of our decision of using QNX
We are still on the road migrating drivers and expert systems to Neutrino… and it works just fine but system robustness seems to be affected since old times and this is a bit disappointing for developers, and a question mark for managers.
We still believe that Neutrino remains one of the most viable option for critical applications such as ours,
but sometimes we start to miss those golden days in which the nodes were running and running while they had electricity
I think that the emphasis on embedded systems has always been a marketing led decision. QNX is still just as viable as a distributed system. QNX 4 did support multiple network channels between nodes for added throughput, but QNX 6 added enhanced features.
All the same, the emphasis on embedded systems is not without a cost. If you have a unique problem with the network, you are less likely to see QNX jump to attention.
I hope you are asking these same question on non QNX forums
Yes QNX marketing is aimed at embedded, but if you stick to the technical side of things, QNX6 is more competent as a distributed system then QNX4 was. For example you can distribute across multiple cores in the same CPU which QNX4 couldn’t do.
As for QNX6 being not as stable as QNX4, yes I agree with you but my experience tells me many of the issue are actually cause by the learning curve involved. Just today our customer service department replaced a computer in a system because they though it kept on crashing. The new computer didn’t solve the problem. Hence they called me, I connected to the system and figured out what the problem was in about 45 seconds. That’s because I was familiar with the error messages and what could cause it, had a similar problem happened with a QNX6 system I’m not sure I could have figure out that fast, would probably have taken me a few hours.
We got a new QNX4 driver 6 months ago that solve our crashing issue
Car company often target 4wd SUV at women, trying to make them feel safer and “powerful”, but that doesn’t meant that same 4wd SUV isn’t good for a men living in northern region where there is 3 meters of snow.
Try to think back at the early QNX4 days, I’m sure you had issues. Just 6 months ago, we got a new QNX4 1Gig network driver that fixed our crashing problem.
Hi Mario, thank you for your response and time, your opinion is always welcome. Actually I totally agree with you, talking about that learning curve, and that is why we still recomend QNX for our customers, but I have to confess that I feel a little jealous when I read the word “embedded”, “embedded” in each documentation’s page, whether from an architectural point of view, as you mention, QNX6 can work perfectly (or almost ) in a distributed environment.
We are planning to go on with QNX despite the noise generated by that highlighted tend towards embedded systems.
The only thing I hope is that QNX’s people do not forget us when we need support for a problem in a 20 nodes QNET network. So far that’s not happened, quite the opposite, and that’s why we still go on (at least for now… )
We are those men living in northern region where there is 3 meters of snow
After reading more I thought I could add just a little bit of texture to this discussion. For many years my selection of OS’s has been limited to either QNX or uh, that dreck from Redmond. These days I’m thinking more and more about Apple OSX, and of course Linux has been around a long time, but probably thanks to the original QNX ed or qed I’ve always looked at QNX first. Since setting up a QNX 6 system I’ve done many fewer projects than I ever did under QNX 2 or 4, but that has been an economics problem, not one with QNX. Along with the consulting, I use QNX 6 as a very reliable server, running Apache 2 with a variety of add ons, eg. PHP and MySQL. I also did all the computing for a master’s thesis on my development system. Connected to an X10 device I turn my office air cleaner on at night. I’m tempted to port an old Dialogic telephone card driver to QNX6, but the QNX2 system that runs my answering machine has shown itself to be just too reliable. These days it runs off of a SanDisk flash drive so there are no longer any mechanical parts to wear out. I continue to find new uses. There are plenty of things I can’t do, for example run any decent video software. I remember in the past when there was a half-decent mpg viewer, but nothing that will play a DVD or any other format for that matter. In school I used Mathematica a lot for which there is no QNX replacement.
My point being that QNX is and has always been quite general purpose, its main limitation being available applications.
Snow in the city sucks but my house is in the woods. After moving all that snow ( either mechanically or by hand ) then sitting down looking at all the white snow I feel a deep sense of satisfaction and appreciation, it’s quite beautiful . Plus I save money not having to go to a gym to keep in shape
Hey Guys, what’s harder to live with, the cold or the heat?
Where I live, today we are heading for a top of 42 Celsius(that’s 107 Fahrenheit)
It’s very hot and dry.
cheers from Bendigo, Australia
I grew up in Arizona, where summertime temps are very hot, and I never got used to it. Our house had a pool and in the summer the water was so warm it was more like taking a bath. The day I moved from Phoenix it was 122 degrees F.
It’s been a cold winter so far here in Oregon, had 4+ feet of snow over the holidays. After living here for 15 years I’d say the cold is still better than the heat. At least now we have real seasons and warm clothes aren’t hard to come by.
I’m the opposite to Ken in that I moved from the cold weather areas (Canada) to the hot (Texas, California and now Florida).
I prefer the heat. Extreme cold actually hurts because it’s so bonechilling. Heat never hurts like that at least in the places I’ve lived (unless you have a sunburn).
Also, in warm weather area’s I only need 1 set of clothes all year, I never have to drive in bad weather or shovel snow or de-ice my car etc (Cold weather areas to me seem made for kids who revel in playing in the snow vs being an adult and having it interfere with your daily life). Once or twice a year when I want to ski or do other cold weather activities (like visiting for a White Xmas) I can visit those places for a week and get my fix.
a QNX6 node… this was unthinkable with QNX4!!
)…