Realtime VM on non realtime OS

Hi,

searching about Java Realtime I have a big question:

How is it possible to run a RT-Java VM on normal, non-realtime Linux S?
I don't understant it

Thanks,
Herbert

Herbert Stiftler wrote:

Hi,
>
searching about Java Realtime I have a big question:
>
How is it possible to run a RT-Java VM on normal, non-realtime Linux S?
I don't understant it
>

Hmm. I don't see any specific constraint on doing what you describe.
What I would expect to see (if you did do this) would be some effect on
the real time performance such as:
- time to switch between tasks
- overhead of activities not managed by the Java VM
and so on.

Are you looking for a specific implementation of Java on Linux or just
want to understand the general concepts?

Depends on how "realtime" is defined. For example, I wouldn't want to be
in an airplane with a fly-by-wire system being controlled by what you've
described -- the timing constraints, and the consequences of a missed
deadline [no pun intended], are far too important but, on the other hand,
I might be very happy with an MP3 player that uses that same configuration
as long as the "dropouts" are small enough that my ear doesn't notice
them.



Tue, 02 Nov 2004 11:11:02 +0100, Herbert Stiftler wrote:

Hi,
>
searching about Java Realtime I have a big question:
>
How is it possible to run a RT-Java VM on normal, non-realtime Linux S?
I don't understant it
>
Thanks,
Herbert

>>Herbert Stiftler <herbert_stiftler@yahoo.dewrote in message >>news:<cm7mfh$1ga3$2@ulysses.news.tiscali.de>
>>
>>Hi,
>>
>>searching about Java Realtime I have a big question:
>>
>>How is it possible to run a RT-Java VM on normal, non-realtime Linux
S?
>>I don't understant it
>>
>>Thanks,
>>Herbert


Realtime "traditionally" meant software that ran/could run in a
"non-timeshared" environment so the situation you describe is
unconventional and probably undesireable. Nowdays the term Realtime is
being used more frequently across a wider range of projects and in a
less clear, heavily contextualised sense.

The application of wide ranging realtime techniques in S design do
not make a general purpose, high volume shrink wrap commercial /S.
That's why Telecoms Switching equipment never ran shrink-wrap
operating systems. Also instead of running a dedicated RT S you could
make a logical progression and throw away the S altogether - it's
only getting in the way (adding to the overhead) of what your
dedicated real-time software task is attempting to do. Finally you
could throw out all the software and build a custom digital logic
circuit in ASIC and run at electron speeds.

In a lot of circumstances realtime just means how quickly something is
gonna happen, and I think that, today you can get away with running a
lot of crappy software in some of the realtime scenarios so long as it
runs on the latest, fastest clocked processor. But again nothing is
for free and this strategy has the caveat of consumming a lot of power
and generating a lot of heat.

dmoore0100@yahoo.co.uk
Independent S Kernel, Compiler, Microarchitecture consultancy.

The problem I had was that i found a java (hard) realtime VM that only
needed a normal linux kernel. But after asking the vendor i found out
that running on this kernel only soft realtime is guaranteed. An info
that should be mentioned on the web-page

Thank you!

Mark H Johnson wrote:
Herbert Stiftler wrote:
>
>Hi,
>>
>searching about Java Realtime I have a big question:
>>
>How is it possible to run a RT-Java VM on normal, non-realtime Linux S?
>I don't understant it
>>
>
Hmm. I don't see any specific constraint on doing what you describe.
What I would expect to see (if you did do this) would be some effect on
the real time performance such as:
- time to switch between tasks
- overhead of activities not managed by the Java VM
and so on.
>
Are you looking for a specific implementation of Java on Linux or just
want to understand the general concepts?
>

>

Herbert Stiftler wrote:
The problem I had was that i found a java (hard) realtime VM that only
needed a normal linux kernel. But after asking the vendor i found out
that running on this kernel only soft realtime is guaranteed. An info
that should be mentioned on the web-page
>
For the most part, that was the point I was trying to make. The RT
performance of the Java runtime cannot be any better than the underlying
operating system.

We can also have endless arguments on what is "hard" and "soft" real
time. Each application has different constraints and if you meet those
constraints, you can have a real time system (independent of what S,
run time, etc.). In practice, I have found that a properly built 2.4
kernel with preempt and low latency patches has an upper bound of about
1 msec in overhead for a maximum priority real time task on the systems
I have tested. That is quite good enough for the application I run but
is not good enough for others. YMMV.