Univac 418 – Paddington

Univac 418Computer Message Switching.

by Robert Brand

In February 1969, OTC’s automatic message relay system became operational at the Paddington terminal. I was sent there for Field Training in 1971 so it was a very new venture and the first to compete in a global competitive environment. I published this photo on Facebook years ago and got a lot of responses. The comments form part of this story.

As a Trainee, I got to have a play with the offline system – there were two – the production system and the standby system. They sat side by side. They were switching messages around the world, but I had access to the offline system and learned how to program in machine language. Kieth McCredden was in charge at the time.

It was on level 4 Paddington before the floors were renumbered later. The Message Relay Switching Centre (MRSC) shared the top floor with the air conditioning for the terminal. It used Fastran drum drives! More on them later. My programming was a bit “ordinary” on my first attempt at doing something useful and I nearly got it right – nearly. I added a wrong loop in the program. The high speed printer “form fed” reams of paper on the floor before we could stop it! Early computers must have been difficult before they had I/O stuff like high speed printers, but this beast could print a book in a few seconds.

Bob Emanuel said: Ah yes, the Univac 418 computer, where every neon light you can see on the panel was also a bit in a register. Fascinating stuff, I cut my teeth in the international telco game on this machine. It was a FASTRAND Drum. 64 read/write heads, weighed over a ton and had, I think about 3 Mb storage. When they had to be replaced OTC searched high and low and found two in the deep south of the USA. They had to knock a huge hole in the west wall of OTC at Paddo to crane them in.”

Thanks Bob. I also remember a time when they took the goods lift out of service and added a massive winch at the top and slowly winch one up that way. It was a huge operation! Later, modern disc drives were added and the thing changed significantly before finally being retired.

Phil Lennon said: I love this photo. My first field placement in OTC was in the MRSC at Paddo. The Univacs were pride of the fleet back then. I was there for the replacement of the drum storage drives with the DEC platter storage system. The 32 byte core memory was a marvel of embroidery art.”

Yes – tiny ferrite beads with tiny wires snaking through for read and write. They was one on the wall in a display like a 3D picture showing the ferrite matrix and wires. Ferrite is a type of ceramic compound composed of iron oxide (Fe 2 O 3) combined chemically with one or more additional metallic elements. It is easy to magnetise and keep its magnetic domains stable.

As Phil said, the original core was replaced in the early days with solid state units. One old unit was in a frame hanging on the wall. It was a work of art. The read and write wires threaded through a matrix of tiny ferrite beads that could be magnetised and read later. Identical to the current RAM in use by modern computers as far as the layout is concerned. A modern person would recognise the address matrix immediately, but instead of Gigabytes on a computer chip, it was 32 bytes on a board!. I wish I had one of those boards on my wall

If you look closely you will see a knob in the middle of the Univac 418 console – that is the speed control for debugging. You could wind the speed to a step every few seconds – yes a “step” You could even press a button to step the lady through her dance – all punch card programming of course. You can see the other desk on the right – As I said there was production and standby systems. It was switching a variety of “messaging” internationally. It was nice to work with such big toys. The standby was never a simple hot changeover. It had to be loaded if the other side failed, so always ready for early programming on the side.

The paper shot our horizontally from the printer because it had so much momentum unless you loaded the catcher. The drums were massively heavy and once spinning, the heads were lowered to near the surface. The speed of the spinning drum held the heads microns above the magnetic surface. Smoke would be a big problem. The big particles would wedge between the head and the drum and you would get a hit and damage the surface. Oh the joys of modern storage in comparison.

Univac 418

The following from an OTC Veterans publication explains why we ended up with these machines:

It was at this point that we would often become aware that the customer was installing a private line, not to London or New York, or where-ever his main overseas correspondent was, but to Hong Kong. The customer would use this line to transmit all his overseas telex traffic, not just to London or Hong Kong, but to everywhere. OTC had become the  unwitting victim of the Cable and Wireless MSC (Message Switching Centre) in Hong Kong, a battery of Univac 418 mainframe computers, engineered to handle private telegraph networks, customised to the needs of each individual company. Such systems and services were being offered by many international carriers, at that time, but the C & W MSC had embarked upon a campaign to target Australian telex customers, probably because they knew that OTC did not have this capability.

It was apparent that OTC would continue to lose business this way, unless it had a computer-based message switching system to offer its customers, so that a network of private lines, connecting to every major office of any corporation around the world, could communicate, via the Sydney-based switch. This entailed selling the service, not just to our Australian customers, but to the corporations (usually multi-national) whose headquarters could be based anywhere in the world.

After OTC management were made aware of the competitive disadvantage that OTC was suffering, in this situation, it was decided to install a message switching centre for private line networks and it was to be called Interplex. At first the Interplex system comprised a number of small, stand-alone computer systems, allocated on a one-per-customer basis, but this arrangement proved too inflexible to meet all our customer needs, so an arrangement of General Automation (GA16/64) mini-computers, called “Mini-Plus” systems, was installed at Paddington, and these were capable of meeting a much wider range of customer requirements.

From that time on, the competition between OTC’s Interplex service and the C & W MSC was very keen. We won some very good accounts and we lost some important ones to our competitors. It was an area of OTC’s business which was truly engaged in competition for business with an aggressive alternative supplier, an unfamiliar scenario for many who had spent their entire careers employed in monopoly carrier situations.

If you like the story from the OTVA vets, please consider joining the VETS to receive more stories and access to a wide range of OTC publications http://otva.com

Paddington MRSC Univac 418

Paddington MRSC Univac 418

Other comments from non OTC staff from around the world:

Jack Crenshaw
wrote: “In those days, computer time was expensive: $600 per hour. That’s in 1960 dollars, when a Coke cost a nickel, gasoline 25 cents per gallon, and that $600 would pay my salary for six weeks.”

As I recall, it (the runaway printer thing) was caused by a missing character in a Fortran print statement. You were supposed to start every string with a printer control character … 0 for console, 1. If you left out this character in the older printers, nothing much happened, it just did a line feed. But the newer printers had large torque motors to advance the page, and if the control character was missing, they would wind up like a Formula 1 engine, filling the computer room with paper.

David W Galea I used to work on a Univac at Telecom Australia (Telstra) in the early 1980. Damn memory storage cabinet needed repairs using octal.

Paddington MRSC Univac 418

Paddington MRSC Univac 418 – a wider view

Some recent Facebook comments since posting this story:

Wal SakalukWhen a .exe file was a bundle of cards with an elastic band around them……”

Hey Wal, try inserting a virus….   Gee this set of punch cards look bigger than normal…..

Tess Simpson Tess Simpson Stuff like that was good enough for NASA. I had to make a few punchcards to bring over to the mainframe at NASA Ames Research Center, to find out Pioneer 11‘s celestial latitude and longitude.
Advertisements

5 thoughts on “Univac 418 – Paddington

  1. Some comments on the Univac 418s,

    They provided the Message Relay System, for the telegram service (remember that!). The system was quite advanced at the time, being the first message switching system to use visual displays (Keyboard Display Units) for telegram entry with semi-automatic conversion from the domestic TRESS telegram network.

    The only other computer telegram switching system then was COTC, which purchased a Univac 418 system before OTC. They had severe problems on cutover and I think had to switch back to their old manual system.

    As an aside, COTC had severe problems when they cutover their Ericsson ARM telex exchange, where a crossed link resulted in putting all the lines out of service.

    The software for the OTC system started with the same basic architecture of a standby system which mirrored all incoming traffic on the Fastrand Drums. In the event of a failure of the online system, the standby would take over and roll back to the last traffic log from the on-line system. The software for OTC was written by the local Univac staff recruited for the purpose.

    The 418s were installed in 1966 as I recall. The Fastrands were lifted in by crane from Oxford St through a hole in the wall as they could not be tilted more than a few degrees. For the later replacement, the end cabinets(s) were detached so the separate cabinets would fit in the lift. The winch had to be used as it exceeded the lift capacity.

    Programming in those days was a black-art, without the foundations of today’s computer science. The system was programmed in assembler, ART (Assembler for Real Time). Due to the limited memory some of the programming practices (not used today) such as using self-modifying code made it difficult to find bugs as the memory dump from the system crash gave the environment quite some time after the original fault.

    The cutover in 1969 was successful. A number of interesting bugs turned up in service. When the logged traffic first reached the end of the area on the drums the software did not include wrapping around to the start of the storage area, Crash!. When errors started occuring on both Fastrand Drums there was no error handling (well two units will not fail at once, will they?) with the result that old telegrams were resent.

    OTC insisted that they should carry out the maintenance on the 418s, which was unheard of, a customer maintaining their own computers! Univac resisted, but had to agree and train the OTC staff. The OTC staff soon became proficient and after some initial problems in maintaining the Fastrand drums were more than a match for the Univac experts.

    The 418s were very reliable, the software causing most of the system changeovers. With staff shortages, the midnight technical shift had to be dropped, with no significant problems. There was a computer operator in the midnight shift to run the accounting programs on the standby system.

    One 418 hardware fault took some time to find as it had the effect of causing an illegal instruction. This was the mechanism used by the software to cause an interrupt to allow the operating system to gain control. One time that some software knowledge came in handy.

    Booting the system involved manually entering the address of the bootstrap routine, which then loaded the loader for operating system and switching software.

    I’m sure the guys at Paddington at the time could add their own tales.

  2. David, Thank you so much for this. It corrects a few things in my story as I was only a trainee on my visit. I did love the place as it was really early computing on a grand scale, but my interests laid elsewhere in the company. Your memory of the fine details really completes this story. Thanks again.

  3. I was a humble hardware techie working on these in the 70s.
    The PMG followed suit from OTC with its CUDN (Common User Data Network) and chose Univac for their experience in message switching. Processors were located in all the capital cities, plus a development model in Melbourne. This system would handle both telex and high speed data. Early customers were TAA, Dept Health and Met Bureau.
    Like OTC, the PMG made the decision to maintain the hardware and software itself following handover, and Univac provided training in the US for the initial PMG staff, later providing further training themselves. I was in the latter intake.
    The first installation commenced in about 1971 for the development units, and Univac 418-III processors were used. These appear to be much smaller than the original 418 pictured, reflecting advances in technology, although they still used discreet transistors. Magnetic core memory was used, but it also had a small cache memory of a few hundred words of TTL memory, where speed was required. I believe the 418-IIIs had faster memory than earlier versions, at 750ns cycle time. As someone mentioned, the control panel allowed one to view every single bit of every single 18-bit register in that CPU.
    Instead of Fastrand mass storage, we used Univac hard drives with 20-layer disc packs. These machines were the size of washing machines and held 20MB of data. To interface the 8-bit disc machines with the 418’s 18-bit word, a Subsystem Controller (I think that was what it was called) was used, where bytes were packed into registers to fill the lowest common denominator of 18 bits. For archiving data, VI-C tape drives were used. High speed data (up to 9600 bits/s) was handled by the CTMC multiplexer. These were an expensive way of handling telex traffic, so the PMG had a low speed multiplexer designed and manufactured purely for that purpose (for the life of me, I can’t recall its name) and this carried 50 and 75 baud baudot data, with the CTMC carrying ASCII data.
    Two 418-III processors were used; one operational and one on hot standby, which was handling traffic like the main processor, but in the event of a main processor failure, the standby would detect a heartbeat failure and take over, with no traffic loss.
    For punch card input and line printer, a Univac 9200 was employed. Although it was a processor in its own right, it was only used for loading programs and print dumps. Smaller states like WA didn’t have this, just a Tally line printer. The 9200 was interesting in that it didn’t use ferrite core memory, but a matrix of magnetically-plated wire which worked basically the same as core, but was cheaper to manufacture.
    There were significant delays in software development, but the system eventually went online nationally in early 1975, about 3 years late. By this time, many prospective customers had pulled out, as dedicated data lines had dropped in price in the meantime, plus they ended up making other arrangements as they couldn’t wait. The operating overheads could not justify its continuing use for what it was designed with so few customers, and it was closed as CUDN later in 1976, the remaining customers being provided with alternate services. It was not wasted, however, and was retained as an internal switching network for telephone orders (SPAN).
    In later years, the core memory was replaced with solid state RAM, designed in-house.
    I was only a hardware man, and one of my curliest faults was on the model processor, where it would randomly shut down every few days. The shut downs were eventually traced to a renegade spike on the earth of the cabling running to the low speed multiplexer. This spike was not present on other processors and was eventually traced to one of the 3v 100A switching power supplies. These were switched with a giant SCR and the cable lug for the SCR was not correctly bolted to its terminal, causing a slightly resistive joint. Somehow this reflected in spikes being generated from that power supply on the earth lines, although the DC was still perfectly clean. Correctly securing the lug fixed the problem.
    As mentioned earlier, the 418s were pretty reliable. We did have the occasional problem with the hard drives, which used hydraulic actuators to move the flying heads in and out. One morning we came in to work to be greeted with a pool of oil trickling out from one of the drives and a flashing red light on the control panel. It was quite a set-up procedure to fit and re-calibrate the new actuator. I don’t recall we ever had a head crash, despite the horror stories we had been told.

    • Gerry, there is nothing humble about the work you did. State of the art really and it sounds like you knew your stuff. OTC also upgraded to the big disk drives and got rid of the Fastrand Drums. Life was pretty interesting back in teh day when computer technology was new. You actually had to learn about it on the job. Often you would be working somewhere for a long time before you got a course on the thing!

      It was fabulous hearing the different ways that these were deployed and the PMG’s lack of success with customer services, but their success using it for telephone orders. Fabulous stuff.

  4. I did a couple of weeks at the MRSC as a trainee in the late 70’s. It was quite impressive, those Fastrand drums were amazing. I do recall a procedure involving a rubber mallet to unstick mercury wetted relays…..

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s