Let me start by saying I am an NT/W2K bigot and recommend it for many
things. I would not use NT, with or without extensions, for “hard
real-time”. Its interrupt latencies are far too long and unpredictable for
this use and I don’t trust any of the real-time extensions. Period.
Depending upon the requirements, you may consider a client/server style of
architecture. You could use something more suited to real-time (QNX, Wind
River Tornado, iRMX/InTime, Windows CE or even roll-your-own if it’s
small/simple enough) as your real-time engine and use NT as you HMI. This
has been done successfully by several DCS (Distributed Control System)
vendors for a few years now.
Just my $0.02 (US) worth.
Greg
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com]On Behalf Of
xxxxx@linkeng.com
Sent: Tuesday, June 13, 2000 7:57 AM
To: NT Developers Interest List
Subject: [ntdev] [Off-Topic]" Real-Time" extensions to NTHi all,
This topic comes up every now and then – sorry if this has been answered
before.Has anybody had any experience (good or bad) with real-time
“extensions” to
NT? A while back, our company had investigated products such as
HyperKernel, but ended up not using them because the scope of the
projects
we were involved with did not truly require “real-time” performance. We
currently stuff National Instruments’ data acquisition & control
cards in a
PC, and write custom software to interface to them. Not exactly
“real-time” in any sense of the word, but adequate up to now beacuse the
control requirements were not stringent.Now the question of “real-time” has come up again here (more
stringent PID
control requirements), and although I’m a firm believer that
“real-time” is
not possible under NT (or W2K, or any other general-purpose OS), I’m
charged with re-investigating the possibility of using such
extensions (or
finding other solutions) “Other people do real-time control in
Windows” I
hear from higher ups here – but I’m not buying it – not yet anyway.All the information I’ve read suggests that interrupt latency is not
predictable, and built-in timer performance is not predictable.
So without
a predictable stable time-base, stable control (in my opinion) cannot be
assured. I must admit that I’m biased toward using devices that
have their
own microcontrollers (and dedicated software/firmware) on them, but
nonetheless, I’d like to hear from anyone who’s had any
experience with the
NT “extensions” or similar solutions.Thanks in advance for any thoughts or suggestions.
-bill
Bill Christie
Link Engineering Company
www.linkeng.com
You are currently subscribed to ntdev as: xxxxx@pdq.net
To unsubscribe send a blank email to $subst(‘Email.Unsub’)