| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 1237.1 | real-time is real-unclear | RTOEU::EGAUTHIER | AUA - Another Useful Abbreviation | Wed Mar 12 1997 04:47 | 19 |
|
Typically there are two types of real-time that are discussed;
Soft real-time and Hard real-time.
Unfortunately, often people correlate Hard with Fast. This is
incorrect. Really Hard means deterministic. Hard real-time
applications typically cannot tolerate missing an event or loss of
data.
Soft real-time are little more tolerant of missing an event or data
packet.
In short, the definition depends on the application. Is it data
driven? Event driven? High throughput? Many I/O points? Etc.
This probably doesn't clear things up at all...sorry.
-Eric
|
| 1237.2 | Compared to what? | BBPBV1::WALLACE | john wallace @ bbp. +44 860 675093 | Wed Mar 12 1997 06:20 | 16 |
| If "right answer, too late" is unacceptable, then it's a realtime
system.
Unfortunately that makes everything from (e.g.) aero engine controls,
thru theatre ticket booking systems to (at a stretch) payroll fit into
"realtime". Theatre bookings and payroll are classic VMS applications
too (wrt origins of this discussion).
Realtime systems are often event-driven rather than stored-data
oriented. Oracle => stored-data, LabViews => event-driven.
Some people think being able to access hardware easily is a
pre-requisite (or a characteristic) of a real-time system.
regards
john
|
| 1237.3 | Where'd Doug put his famous paper (in UK IEE journal this month)? | BBPBV1::WALLACE | john wallace @ bbp. +44 860 675093 | Wed Mar 12 1997 06:30 | 20 |
| Is comp.sys.realtime (sp?) still around ? Is "what is realtime" an FAQ?
Is DIGITAL UNIX a realtime OS? It's pretty good on interrupt latency,
even compared with VxWorks. Thread to thread context switches are
pretty good (also compared with VxWorks). It's not so good on
determinism as VxWorks, though V4 is a lot better than v3 (min, mean
and max are much closer together, there are even fewer instances of
long preemption times or poor interrupt latencies).
Is OpenVMS a realtime OS? Who knows... no-one seems to measure this
kind of thing for VMS these days. I suspect mostly it would be not much
worse than UNIX, though process-related stuff would be slower as a
process has more context than on UNIX.
Is Windows NT a realtime OS? Not by most people's definition - but then
nor were DOS and OS/2 and there are lots of factory floors being
controlled in "real time" by those two.
enough for now
john
|
| 1237.4 | | RTOEU::EGAUTHIER | AUA - Another Useful Abbreviation | Thu Mar 13 1997 09:50 | 8 |
| > Is OpenVMS a realtime OS? Who knows... no-one seems to measure this
If you measure it by installed base, yes. Many, many people have
deployed real-time systems using VMS, but that was in the days of the
VAX.
-Eric
|
| 1237.5 | thre is RealRT and JustRT | NAMIX::jpt | FIS and Chips | Wed Mar 19 1997 08:42 | 30 |
|
See:
http://www.realtime-info.be/encyc/techno/publi/faq/rtfaq.htm
or
news:comp.realtime
This is one of the most common topics in the news group.
Anyway, personally I do not accept the idea that there is
"no definition" for realtime, and the previous answers state
pretty well in practise what the meaning of real time in real
life is.
Anyway, the key idea is deterministic and predictible behaviour of
system. Idea of application "knowing" that under no circumstances
certain latencies between actions or events will not be exceeded.
The "hardest" real time systems use typically High Priority hardware
based Watch Dog to watch system sanity and activate recovery actions
in case where for some reason events will be blocked longer than is
acceptable. In good RT system these safe guards/watch dog actions
never fire as the system (hw and sw) exceed the needs of application.
Well, this would be worth of a book ;-) There are some fun war-stories
from years when I used to design RT hardware and software for another
company, and those projects teached me that there certainly is
difference between Real Time and "real time".
-jari
|