[comp.realtime] My IBM rep says that the RS/6000/w AIX is realtime, is this true?

ssw@cica.cica.indiana.edu (Steve Wallace) (08/24/90)

Our IBM rep is telling us that the RS/6000 running AIX will do
realtime applications.  Specifically, the RS/6000 will be able to
time user response time with a given accuracy.  The local IBM
tech type says just adjusting the priority on the process will
guaranty a maximum latency between an interrupt (keypress) and
the process waking up and recording the time.

Does anyone know if the RS/6000 & AIX have this capabilities, or
is IBM bluffing?


Thanks,

Steven Wallace
Indiana University
wallaces@ucs.indiana.edu

ejp@icd.ab.com (Ed Prochak) (08/25/90)

In article <4350@cica.cica.indiana.edu>, ssw@cica.cica.indiana.edu
(Steve Wallace) writes:
> Our IBM rep is telling us that the RS/6000 running AIX will do
> realtime applications.  Specifically, the RS/6000 will be able to
> time user response time with a given accuracy.  The local IBM
> tech type says just adjusting the priority on the process will
> guaranty a maximum latency between an interrupt (keypress) and
> the process waking up and recording the time.
> 
> Does anyone know if the RS/6000 & AIX have this capabilities, or
> is IBM bluffing?
> 
> 
> Thanks,
> 
> Steven Wallace
> Indiana University
> wallaces@ucs.indiana.edu

What's the value for the maximum latency?
(milliseconds?  seconds?  minutes?)

Does the priority ever get readjusted by the OS?
(as I understand it, unix priorities may change
 to provide a "fair" distribution of cpu time)

Do you or the process have to have special (root)
privileges to change the priority to the value which
is associated with the guarantee?

I have never used AIX, so I'm just asking very
general questions here. Maybe someone else has
some answers??


Edward J. Prochak   Voice: work-(216)646-4663  home-(216)349-1821
               Email: {cwjcc,pyramid,decvax,uunet}!ejp@icd.ab.com
USmail: Allen-Bradley, 747 Alpha Drive, Highland Heights,OH 44143
Wellington: ENGINEERING is "the ability to do for one dollar,
                            what any damn fool can do for two."

skip@rafos.UUCP (Skip Carter) (08/29/90)

In article <1793@abvax.UUCP> ejp@icd.ab.com (Ed Prochak) writes:
>In article <4350@cica.cica.indiana.edu>, ssw@cica.cica.indiana.edu
>(Steve Wallace) writes:
>> Our IBM rep is telling us that the RS/6000 running AIX will do
>> realtime applications.  Specifically, the RS/6000 will be able to
>> time user response time with a given accuracy.  The local IBM
>> tech type says just adjusting the priority on the process will
>> guaranty a maximum latency between an interrupt (keypress) and
>> the process waking up and recording the time.
>> 
>> Does anyone know if the RS/6000 & AIX have this capabilities, or
>> is IBM bluffing?
>> 
>> 

	Several months ago (around April) I got a chance to demo an
RS/6000 for a couple weeks.  When the IBM rep said the words "real-time"
my ears perked up.  I asked him if it was really real-time or just fast,
he got a little nervous and eventually said something to the effect of
"... well if you mean real-time like say, a Masscomp, no its not real-time,
but ITS REALLY FAST SO WHATS THE DIFFERENCE ?"

	He also told me that the system was BOTH SYS V and BSD compatible,
but he couldn't tell me how that was accomplished.


	I remain skeptical.

-- 
Skip Carter			UUCP:	uunet!rafos!skip
School of Oceanography	    INTERNET:   skip@rafos.gso.uri.edu
University of Rhode Island     Phone:   401-792-6519
Narragansett, RI 02882

jburnes@crash.cts.com (Jim Burnes) (08/31/90)

Skip:

About your question about the RS/6000 and AIX, I really don't think the
thing has what most sentient engineers mean when they say real-time.
As far as I know AIX is really a descendent of the sysv series of unix
kernels which are definately not realtime.  Also never take an IBM reps
word for anything.  The real question that picks the realtime machines
from the ones that 'up your statistical priority' is that realtime kernels
have (almost always) preemtive priority multitasking.  In other words,
realtime kernels are deterministic, not stochastic.  Easy example:

    Astronaut Aldrin is aboard a shuttle and is playing chess with the
onboard computer.  The computer is really crunching away trying to 
find the best move to put Aldrin in check when the task scheduler on
the system sees that its time to do a three second orbital burn.
That burn MUST commence in 37 milliseconds.  On your basic unix box
the astronaut would most likely be stranded in space for ever.  With
the miracles of a realtime kernel the chess program having a lower
priority than the orbital manuevering program (hopefully) means that
the chess program gets swapped out NOW every time this situation occurs.
On most flavors of unix, you would have pumped the orbital program up
to the highest priority and chess would be running at normal or low
priority.  When the orbital burn timer tells the system its time to 
swap in the manuevering program the manuevering program will be swapped in
when the unix kernel thinks the chess program has had enough time
allocated to it.  Now by the OS's computations, the chess program 
deserves another 43 milliseconds.  When the orbital manuvering program
finalllxy gets swapped in and running, its too late to do the orbital
burn and the shuttle and Aldrin skip off the atmosphere, doomed to
play out endless series of chess matches with each other. 

Hope I didnt insult your intelligence or mine with this little explanation.

Have a good one.

paw@eleazar.dartmouth.edu (Pat Wilson) (09/03/90)

It may be true that AIX is part of a real-time system.

When it first came out for the PC/RT, AIX was just one of the processes
run by a small real-time kernel.  Others included device drivers.
There were hooks for real-time scheduling.

So the UN*X end of it is descended from SysV, but there may be more to
it than that.

jat@xavax.com (John Tamplin) (09/06/90)

In article <4350@cica.cica.indiana.edu>, ssw@cica.cica.indiana.edu
(Steve Wallace) writes:
> Our IBM rep is telling us that the RS/6000 running AIX will do
> realtime applications.  Specifically, the RS/6000 will be able to
> time user response time with a given accuracy.  The local IBM
> tech type says just adjusting the priority on the process will
> guaranty a maximum latency between an interrupt (keypress) and
> the process waking up and recording the time.
> 
> Does anyone know if the RS/6000 & AIX have this capabilities, or
> is IBM bluffing?

I was a consultant to IBM AWD while they were developing AIXv3 and I
think I can answer this.  There are two classes of process priorities,
fixed and variable (normal Unix style).  The scheduling is done preemptively
with the highest priority runnable process getting the CPU.  Every so often,
variable priority processes are adjusted based on the amount of CPU they
have received, but are always lower priority than fixed priority processes.
At each fixed priority level, there is an instruction count associated with
that level.  Any process running at that level should run for no more than
that many instructions.  Interrupt levels and off-level interrupt processing
are treated as the highest priority level of those that are fixed.  Given
these constraints, AIXv3 is indeed a "soft" real-time system.  Soft, since
there is no guarantee of meeting a dead-line, but it is predictible if you
know what set of real-time processes are running.  Of course, if any device
drivers or real time processes violate the instruction count, all bets are
off.

The kernel is rebuilt from the ground up, and has very little to do with
stock SysV or BSD code.
-- 
John Tamplin						Xavax
jat@xavax.COM						2104 West Ferry Way
...!uunet!xavax!jat					Huntsville, AL 35801

skip@rafos.UUCP (Skip Carter) (09/07/90)

In article <4192@crash.cts.com> jburnes@crash.cts.com (Jim Burnes) writes:
>
>Skip:
>
>About your question about the RS/6000 and AIX, I really don't think the
>thing has what most sentient engineers mean when they say real-time.
>
>(stuff deleted ...)
>
>Hope I didnt insult your intelligence or mine with this little explanation.


	I guess I sounded a bit flip,  I DID understand that the RS/6000
wasn't real-time, in fact I had a good time making the IBM guys squirm when
forcing them to admit it !

	Not insulted at all (thick skin).

-- 
Skip Carter			UUCP:	uunet!rafos!skip
School of Oceanography	    INTERNET:   skip@rafos.gso.uri.edu
University of Rhode Island     Phone:   401-792-6519
Narragansett, RI 02882