Re: What is "real-time"?

[munnari.oz.au!metro!seagoon.newcastle.edu.au!wombat.newcastle.edu.au!eepj m@uunet.uu.net (Peter Moylan)]

In article <6095@ucsbcsl.ucsb.edu>, hossein@nu.Berkeley.EDU (Hossein Moiin) wri
tes:
> 
> What is real about real-time? What does the word real imply in the
> definition of real-time systems? I have wondered about this and I think
> that I saw a paper of the same title some time ago. I was wondering if 
> the real in real-time means that time-axis is treated as a real axis.

In the control systems community, there is a tradition of distinguishing
between real time and, for example, simulated time.  This ties in with
the distinction between on-line and off-line algorithms.  An off-line
calculation is one which can be done before the system starts running,
e.g. at the design stage.  An on-line calculation is one which has to
be done while the system is running because it uses data from the
running system.  Obviously, there's a relationship between "meeting
timing constraints" and "on-line", so to many control people think of
"on-line" and "real-time" as synonymous.

This is not precisely the way the real-time community thinks of things
these days, but I think it does explain the etymology.  "Real time"
simply means genuine, actual, real time as measured by a clock.  It
has nothing to do with the meaning of "real" when talking about
complex numbers.

I think that we're just now beginning to realise that the modern
definitions of "hard real-time" are broken, or at least that they're
irrelevant to real-time control.  Any practical real-time system must
have some reasonable robustness properties, i.e. it should degrade
gracefully under conditions like a short-duration processor failure.
>From this it seems to follow that a real-time system should not contain
any hard real-time deadlines.

Of course this does not destroy the usefulness of "hard real-time"
as a concept.  We can continue to design our scheduling methods while
maintaining the fiction that deadlines can be met.  We do, however,
have to recognise that these deadlines will not be met out in the
real world of power surges, sensor failures, and the like.
-- 
Peter Moylan                      eepjm@wombat.newcastle.edu.au