Looking for online information about the old POSB mainframe/branchterminal system

Anybody know where I might find an online repository, or anything really,
with information about the old mainframe and terminal system that was
used by the OLD POSB until it was replaced when purchased by the ANZ?

I am especially interested in finding info about the old Olivetti
electromechanical terminals that were in use in the branches around the
country until they were replaced with the Phillips terminals.

I've had a hard look around the Internet and I can't seem to find
anything about it.

 

Well, I was one of the network engineers at Postbank / ANZ and was
intimately involved with the cut over from the Olivetti's to the Philips
WS11 controllers. Those were the days, 1200bps of analogue madness (one
1200bps "line" connected to between 8-12 branches) connected to ICL
DRS40 mainframe front end processors. The memory is a little fuzzy now
but what do you want to know ?

 

Pretty much everything. )

I knew that the Olivettis connected directly to the regional mainframes
which in turn connected to the central one.

What I'm interested to know is what OS the mainframes were running, and
how the network was managed and specifically how the Olivettis were set
up.

I used to use them back in the '80s, and so I think I sorta might still
know how to key in transactions with the help of the Post Office Manual
- red bar, green bar, yellow validation slips, etc.

I'm interested to know how those Olivettis worked, and how they
communicated with the mainframes.

Chur - didn't expect to actually see someone posting here who was
involved in actually setting up/fixing the stuff. )

I reckon that there should be a website set up somewhere specifically to
remember how the POSB banking computer system was set up - because it
was, IIRC, the first fully online realtime banking system in NZ.

Those Phillips terminals were hopeless. A quick teller could key in
transactions considerably faster than the Phillips terminals could
process the keyed in information, and it only had a very tiny buffer and
spat the dummy if you went too fast.

The Olivettis just kept on going, no matter how fast you were. o)

 

The actual mainframes were in Newmarket in Auckland (25 Nugent Street).
There were front end processors (The ICL DRS20
http://en.wikipedia.org/wiki/International_Computers_Limited#DRS_Range )
in Auckland, Wellington and Christchurch. I remember the
Auckland-Wellington sites were linked via ethernet bridges (ethernet was
10base5 in those days http://en.wikipedia.org/wiki/10BASE5) with 2 x 48k
digital links). I think Christchurch was connected off wellington with a
19.2k but can't quite remember).
The funny thing is the network actually had a bigger mainframe running
it than the actual banking application for a long time.

I'm trying to remember but I believe the mainframes were 2x3900 ICL
running the network and 2 x 29 series (I think this is a picture of a 29
series in the background http://www.tnmoc.org/images/2966-1.jpg) running
the application. Eventually the 39 series was moved to running
everything. The o/s was VME. I rememberer that system with great
fondness, it was fast and very reliable. Eventually ANZ decided to make
us all redundant and move (backwards IMHO) all the applications onto the
IBM s390's in Melbourne and thats where it stands today. They even still
run the old ICL dot matrix passbook printers in the branches that I bought !

Olivettis were great old terminals, its just the mechanics were getting
a bit old. They were connected via 4 wire analogue asynchronous 1200bps
multi-drop lines to the ak/wn/ch. When the cutover to the philips ws11
was done I think we upgraded to 9600bps digital multi-drop lines from
memory.

and I'm not even that old ! sometimes the younger set do look at me
strangely when they complain about their several megabit DSL links and I
explain my first real computer job included feeding the ticker tape from
the 75baud machine into the 50baud one !

It was, and it was a fantastic system until ANZ killed it off. We bought
up the network at , 7am in the morning from memory and shut it down at
6pm I think. It was real time during those hours, and then ran all the
batch processing outside of those hours.

Yes, a lot of the tellers didn't want to change, but the Philips
provided a better solution overall. The Olivetti was only good for one
terminal in the branch where as the Philips could run several (8?)
terminals and was "digital" (got to love those 8 inch floppies). From
memory I think we bought a whole lot of second hand ones to effect the
change. BUT, in the mean time an entire replacement system based on PC's
and windows 3.11 was locally developed, we even rolled it out to the
backoffice center at 277 Broadway in Newmarket, and were preparing to
roll it out to all the branches when ANZ killed the project, made
everyone redundant and outsourced everything to Databank (later to
become EDS, where I moved to).

 

I remember one thing about it as a customer, it would bounce POSB
cheques in real time. The teller would 'clear' POSB cheques you were
depositing via the terminal and if there was insufficient funds it
bounced there and then, and the teller would give it back to you.

 

From a Teller's perspective the Philips terminals were very much slower -
especially if you hadn't memorized the new processes for each
transaction, and you had to keep looking up at the monitor instead of
merely glancing from between the passbook and the keyboard.

I agree about the reliability of that system. It was rare for it to fall
back, and very rare indeed for it to go down completely. I think that all
of us who used it and had some understanding of what the Trading Banks
did knew that we had a very good system. )

I didn't realise that the Olivettis were analogue machines.

Thank you very much for the links, and the info - appreciated.

I still think that there should be an online repository of information
regarding the POSB banking system should be created - specifically
because it was such a good banking system. )

 

Yes - because we treated cheques drawn on our own branch as if they were
withdrawal slips, and we could check the current balance of any a/c there
and then.

A person could do a "No Book" deposit in a branch in Auckland, and
immediately afterwards the a/c holder could walk into a branch in Dunedin
(or anywhere) and update their book and then withdraw the money
immediately.

That was one of the features of the POSB system - transactions were done
immediately - as soon as the Teller had confirmation that the transaction
was accepted the a/c would have already been updated.

 

The Olivettis had a 10 character keyboard buffer... implemented in hardware.
Real mechanical hardware(*). I was a telegraph tech in the Post Office, and
used to maintain the previous model Olivetti terminals used for Telex. The
Telex machines were TE300 series, mostly TE315, and the POSB machines were
TE400 series.

(*) The TE300 and TE400 series machines were "digital mechanical" machines.
They had digital logic implemented in mechanical hardware. They had
monostable and bistable clutches, for example. The aforementioned keyboard
buffer was a set of bistable sliders. Pressing a key would push or pull a
set of sliders into a pattern corresponding to the 1s and 0s of the
character to be transmitted. There were 10 slides for each bit, arranged in
a ring, with as many rings as needed for the code used (Baudot code used 5
bits, but there were spaces for up to 8 bits). The buffer "writer" would go
around the rings setting the bits as each key was pressed, and the buffer
"reader" would go around the other end of the slides reading their
positions.

 

Cool!

They were amazing machines. They were, IMHO, much easier for tellers to
use, once we had learnt the keying process for each different transaction.

And, once you'd chosen the transaction type experienced tellers could
complete the entire transaction without needing to look at anything other
than the transaction slip and the passbook. )

The additional space for 4 characters in the buffer obviously made all
the difference when a fast teller keyed in the transactions. )

Can you give any details about the electrics in one of e POSB Olivettis?

I know it had an electric motor in the back that drove all the mechanical
bits in the top part of the terminal, but underneath it seemed to be a
whole lot of circuit boards

Do you know if any of them survive in a museum somewhere here in NZ?

 

This was possible because the branches were all effectively connected.
The mainframe itself in Auckland (there were actually two mainframes on
either side of the main computer room sharing common disk via a 50Mb/s
fibre lan, the front end processors were connected to the mainframes by
3 x thicknet 10meg ethernets), did not really know that Dunedin and
Whangarei were not in the building next door.
There was also the "Tandem" (now HP/Compaq) mainframe that ran in
paralel that ran the ATM's and eftpos.
(http://www.partnersremarketing.com/tandem.htm picture down the bottom)
It would do real time transactions to the ICL mainframe during working
hours and run via a local database overnight when the system was offline
doing its batch jobs. Before teller start up in the morning the offline
Tandem database would sync with the ICL updating it with all the
overnight transactions. The Tandem connected to all the other bank
systems (9.6k or 19.2k links normally) to allow the use of other banks
cards on the ATM eftpos network as it could communicate using many
different protocols running its "base24" operating system. Eftpos is
still done in a similar way using HP non-stop machines.

 

I didn't get to open any of the POSB Olivettis. They were just coming in
when I left the Post Office.

Not in NZ as afar as I know, but Google shows a few in museums overseas.

 

Thanks very much for that additional info. )

It's all interesting stuff, IMHO. )

 

What was housed in Herd St, Wellington?

That is where I had thought both the Wellington front-end AND central
mainframes were.

Also, what generally had happened when the system went into "fallback".

I knew when it was in fallback because we wouldn't be able to do passbook
updates, and the acceptance code on the transaction slips were different.

My memory may be faulty here, but IIRC the acceptance codes used to
alternate between /1 and /2, and in fallback they would be /3 & /4.

 

I can't remember the address of the Wellington Network site, but it
could well have been there.

Nope, only the DRS20 Front end processors. They had an ethernet
connected by "Isolan" spanning tree bridges to Nugent St where the
mainframes were and 8x async serial ports that connected to the branch
multi-drop lines. There were also helpdesk staff and cheque processing
machines from memory.

Hmm, Its been a long time, but I think what you might be referring to
was when the strange situation existed where there were actually four
manframes running things. Two (the more powerful ones, 39 series that
eventually took over everything) ran the network and had their own
"offline" database. They were connected by a dozen or so (from memory)
19.2k links (modem eliminators) to the 29 series mainframes
(primary/backup) that ran the real time system. From memory it was
possible to "disconnect" the 39's from the production system and run in
a fallback mode.

I remember distinctly one day during heavy rain that we had a situation
where a bird had nested on the roof of the building and clogged a
central gutter. This caused a cascade of water to make its way down
three floors and ended up pouring into one of the disk packs and 29
series mainframes. This caused obvious panic and a disconnect in the
middle of the day, Much to our amazement though within 2 minutes the
system automatically moved production to the secondary mainframe and
disks on the other side of the room without any human intervention. This
shows how advanced the ICL (Fujitsu) mainframes were in those days.

I don't know enough about the teller operation to know if this is what
you are referring to, my job was running the network (we had a team of
four from memory). It was a great company to work for at the time.

 

Yeah - that would be it. )

Fallback hardly ever happened, but when it did we couldn't update the
passbooks, but we could do deposits & withdrawals based on the printed
balance in the book. I think it worked because a part of keying in the
transaction was also keying in the current passbook balance.

Sounds like some very clever and on-to-it programmers had set it up in a
very elegant manner. )

You might remember this...

One day before the POSB turned into PostBank there was an outage that
lasted for most of the day - not just fallback but a complete outage, and
the urban myth that I'd heard from a colleague at the time was that a
large ship had come into Wellington Harbour and had forgotten to turn off
its deep sea radar and that had knocked out the computer in Herd St.

It was, as I said, such an unusual event to have any such outage let
alone a protracted outage like this one was, so I'm wondering if you
might remember it. )

 

Given the overall reliability of the entire banking system as it was It
sounds like the system with the two 29 series mainframes had also been
engineered to provide the same level of reliability in the event of a
disaster. )

And, based on your account, it lived up to expectations. )

 

Haha... no sorry don't recall that one, but someone sure got that one
wrong. Firstly radar could never effect those machines (I trained as an
avionics (radio/radar) tech before joining PostBank), and secondly the
computers werent even in Wellington. Most likley whatever happened would
have been before I joined as I joined just after it turned into Postbank.

 

Yes, in many ways going from the ICL gear to the IBM mainframes was a
big step backwards. It took IBM many more years to catch up on even the
concept of LAN's (then they did token ring !).

 

Good Grief - that was a lot of catching up.

 

Were _all_ the DRS20s up in Auckland?