> I one had the responsibility of choosing a CPU, OS and language
> (well to be truthful, the choices were made - I needed to figure
> out if the choices were OK) to ensure that a READER (scanner
> device at one end) connected via a microwave link to a WRITER
> (laser plate maker at the other end which produced the actual
> plate to print a single newspaper page - about a mile away in the
> printing press building). As many plates as were required were
> produced by the 6 WRITERs (a computer guided laser printer which
> produced metal plates). This was being done back in 1981 and there
> was no reasonable device to hold the scanned image from the
> READER.
:) Nice Job.
Reminds me of an application where I had some part in the
hardware maintanance, around 1979 - At the RBG (Rechenzentrale
bayerischer Genossenschaftsbanken - Computercentere for
Bavarian Coop-Banks), we had an OBL (Oprischer Belegleser -
'Optical Form Reader') for scanning bank cheques (OCR), sort
them and store the result (*). The host was a /370ish system
from SIEMENS running under BS1000, a non-VM OS. The system had
a distance of some 60 cm between the reader and the first turnout
to decide the proper turnout to take. At a speed of >5 m/s
this will leave way less than 100ms for two complete blocked
I/O cycles _and_ a decision. Also the host had to store the
readed information for further processing on disk _and_ tape.
This result in random inserted disk/tape I/Os to be handled
without any chance to stop the OBL - even if it would have
been possible, the reader would have send some more cheque-
informations - not to speak of the loss of time, since it took
almost 15 seconds to fire him up.
The time frame was prety tough, especialy, since the CPU wasn't
a real runner (760 kOps), but the task was well done, even some
terminal-I/O was possible for controll status, etc. Around 1978
the SIEMENS sales people did succesfull place BS2000 our (back
then new) timeshareing VM system - with one notable exeption:
The OBL host did further run with BS1000 1.3. Even a machine
more than twice as fast (2.000 kOps) and a changed reader timing
couldn't handle the job it in time - The OS was designed to
handle resources as flexible as possible to as many users as
possible, but with an big backdraw when it has come to _all_time_
guaranteed reaction times in sub second dimensions.
It took them some 5 years until they could replace the OS (in fact,
the BS1000 OS development was stoped official in 1982, but as late
as 1984, SIEMENS was forced to offer new releses to some few but
importent customers (like Deutsche Bank), to support new hardware
- even a limited VM system was implemented (while keeping the OS
still small)).
This (and similar) Task helped the BS2000 development a lot, since
the OS crew was forced to leave the ivory tower of multi user OSes
and learn that the customers did run real world tasks on our machines :)
Nowadays the OS is a real fine thing - and almost dead from the
companies perspective - if it wheren't for some customer insisting
that there is no real replacement :)) History repeats itself.
> Do you begin to understand why a real time OS must be able to
> guarantee that each "task" be scheduled within a specified time
> limit so that the command to the READER to start sending could
> be completed within the window?
What I want to tell is just the fact that the label has nothing
to saay, it is always task dependant - in fact, all OSes are
real time, if they finish their work within the needed timeframe
(and if not, they are not non-RT, they are just not usable :).
Gruss
H.
(*): a phantasic machine - it could transport a used 100 Mark
note and store it properly into the 'trash' register without
any problems - and fast, ubelivable - more than 15 cheques
per second - so fast, the paper seamed to be an continuos
band - just the justage of the turnout was kind of a nicghtmare,
but I loved it.
--
Stimm gegen SPAM: http://www.politik-digital.de/spam/de/
Vote against SPAM: http://www.politik-digital.de/spam/en/
Votez contre le SPAM: http://www.politik-digital.de/spam/fr/
Ich denke, also bin ich, also gut
HRK
Received on Mon Apr 19 1999 - 07:48:01 BST