[Stephane Pierre Bordas]
| I don't think this would be a good idea to use Windows with such small
| durations, but I am facing the reluctancy of my partners and would like to
| know your advice on that.
I'm not sure if I'm adding anything valuable into the discussion, but
I have been more or less a victim of wrong decision making about these
things during the last months, so I thought I could share some
thoughts.
The company I work with have been doing some experimental seismic
surveys using low cost aquisition systems such as PC's. The recording
system was designed by external people and the decision to use Windows
95 and NT was made because drvers was already available, and I
couldn't guarantee that I would be able to make RT-Linux drivers in
time before the survey was going to take place.
Therefore, we used two PC's equipped with Windows 95 and two DAS6402
aquisition cards. A third PC with NT and a DAS1402 card was used to
record the gun signature using a hydrophone mounted on the air gun we
use as a seismic source.
The result was this:
* We lost samples on the NT box, on 3-4% of the traces recorded, the
system did not respond when it got the trigger signals, but waited
up to 20 ms before it started sampling. Thus, we could not assume
that any of the traces was correctly recorded, so we can't use those
datasets for anything.
* On the two Win95 PC's at the seabed, we logged the daytime of each
time the system were trigged. We know that the systems started
sampling when they got the trigger signal, but the time stamps were
wildly off. The standard deviation of the difference between the
two machines was as high as 3-4 seconds. Fortunately, this was not
enough to make the data unusable, but it made me sitting two weeks
swearing while writing ugly perl-scripts that were supposed to clean
up the mess.
Luckily, we endend up with good data, but I had to waste two weeks
trying to find out what had happened.
The next time all systems will be based on RT-Linux. We will not use
Windows in any time critical systems again, using Windows for such
tasks is plain wrong. Right now I am doing some investigations on
what will happen when the jiffie-counter goes full, as it will after
497 days since last reboot. If nobody cuts the power, uptimes like
that will be reached in a project I am working on, and we must figure
out how we should deal with this thing. Besides hardware problems,
this is the only problem I am expecting regarding unstability in
RT-Linux.
Good luck with your project.
--
Øystein Svendsen
--- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
----
For more information on Real-Time Linux see:
http://www.rtlinux.org/~rtlinux/
