
Signature
Kenneth P. Turvey <kt-usenet@squeakydolphin.com>
On Oct 29, 7:53 pm, "Kenneth P. Turvey" <kt-use...@squeakydolphin.com>
wrote:
> What you probably want is for a thread to use no more than 10% of the CPU
> over some time interval. I don't think there is a way to do this in Java.
that's bad
> You are probably writing the code you are trying to limit, so this doesn't
> really apply to you.
Actually, I don't write the programs I'd like to limit. As part of
the EU project MUSIC (www.music-ist.eu), I'm developing a method,
called divide and conquer (D&C), to distribute (parts of) applications
onto an ensemble of devices. As a result, otherwise unrelated, third-
party applications can end up on the same device, thus competing for
memory and CPU. (Check out Publications > Publications related to
MUSIC > Divide and Conquer on the MUSIC portal).
D&C will assign resources to applications. Because control is better
than confidence, it is desirable that D&C can limit the resources and
is able to change these limits later on. Of course, for the purpose
of research it is OK to simply assign the limits and assume that the
applications will follow them. But I'd prefer to have at least one
demonstrator with actual control over the resources, even if this
would work only for one specific hardware/software combination.
Any suggestions welcome.
Ulrich
Ingo Menger - 30 Oct 2007 10:58 GMT
> Any suggestions welcome.
Since you write "any": I remember having read about research concerned
with the fitness of Java for real time applications. Perhaps the
search engine of your choice finds something for +java +real +time.
(Mine does).
Ulrich Scholz - 31 Oct 2007 11:15 GMT
> > Any suggestions welcome.
>
> Since you write "any": I remember having read about research concerned
> with the fitness of Java for real time applications. Perhaps the
> search engine of your choice finds something for +java +real +time.
> (Mine does).
Thanks for pointing me to Java SE Real Time. That seems to be what I
need. I started a new thread with questions about Java SE Real Time
but feel free to point me to other information within this thread.
Thank you,
Ulrich
Ingo Menger - 31 Oct 2007 11:39 GMT
> Thanks for pointing me to Java SE Real Time. That seems to be what I
> need. I started a new thread with questions about Java SE Real Time
> but feel free to point me to other information within this thread.
Unfortunately I can't. The only thing I know is that something like
"Java SE Real Time" exists (though I forgot the exact name) and that
"real time" somehow implies management of CPU ressources. So I thought
I could suggest you research in that direction. Hope you find
something useful.
Lew - 31 Oct 2007 14:26 GMT
> Unfortunately I can't. The only thing I know is that something like
> "Java SE Real Time" exists (though I forgot the exact name) and that
> "real time" somehow implies management of CPU ressources. So I thought
> I could suggest you research in that direction. Hope you find
> something useful.
Just to add to the store of knowledge, strictly speaking a "real-time" system
is one that makes guarantees about service time. Thus you know that a context
switch can take no more than some time epsilon, for example. That epsilon
could be large as long as it's guaranteed. Naturally the shorter the epsilon,
the more responsive and generally useful is the real-time system.
The key feature of real-time systems is determinism. For non-real-time
systems, no matter how long the wait, something can take longer. You just
cannot be sure. With real-time systems you are sure.

Signature
Lew