Talking about a revolution


30 Mar 2004

Forty years ago next week, a computing revolution happened. Its effects, in the IT world and beyond, were far-reaching. On 7 April 1964 IBM announced its System 360 mainframe computer (partly pictured). Mainframes themselves were not new – they had been around for ten years or so by that stage – but this was different.

What the company did forty years ago this year fundamentally changed the way computers were made. Before the S/360, big businesses bought large computers to process data and when these needs outgrew the capability of the machine, they had to throw out the lot and start again from scratch.

There was no question of upgrading the machines’ hardware by adding memory, processing power or storage space. In addition, programs written for one computer could not be simply ported across to work on another machine.

The work involved to produce the S/360 made it one of the most expensive projects undertaken in that decade, second only to the Apollo moon exploration programme (which itself used IBM computers). At current prices, IBM’s outlay was not far off $30bn, whereas the company’s revenue at the time was the same as $19.2bn today. Commentators at the time did not consider the move a sure thing.

IBM’s gambit was to produce a computer designed with compatibility in mind. It’s a word we take for granted now in IT: we expect things to work together. Back then, the technology culture was all about proprietary systems intended to wed customers to their suppliers rather than solving long-term business problems. It didn’t hurt, of course, that IBM’s market dominance was such that competitors ended up having to adopt similar strategies. (The lead engineer on S/360 was a certain Gene Amdahl, who would later go on to found his own company selling IBM-compatible mainframes.)

Many of the changes that have happened to the computer in the meantime are the constants of the industry: greater speed in a smaller box. The first S/360s to roll off the production line left little space over from an air-conditioned room the size of a football field. The amount of instructions that the current version can process is degrees of magnitude faster than its older sibling.

Such is the rate of progress that a new S/360 will roll off the production line around two years after the last one, capable of twice the processing power of its predecessor. In size the current models are not much bigger than a fridge freezer. Some continuity remains: according to IBM, code written in the Sixties for the System 360 mainframe would be able to run on current machines.

Over the years the S/360 was followed by the compatible System/370 range, the 3090 and later the System/390 which was subsequently rebranded as the z/Series. All of the new manufacturing assembly of machines sold outside the US takes place in Ireland at IBM’s technology campus in Mulhuddart.

Times have changed; when IBM announced the S/360 it was delivered two and a half years later. Now a more demanding business climate dictates matters. Former IBM chief Lou Gerstner once said that if you can’t develop, announce and deliver a product in six months, your market’s not going to be there. New versions of the system are usually available a month to six weeks after launch.

At its fastest, the Mulhuddart facility can fulfil a customer order within a week, from the order management process through to building, testing and finally distribution, all the way along the quarter-mile-long production line. The standard turnaround time is closer to a fortnight. “By definition all the orders we deal with are different,” explains Fernand Sanchez, campus vice president, during a tour of the facility. Although all of the components are standard, customers can have their computers configured to order, with greater processor power, higher amounts of memory or more input/output (IO) connectivity as their needs dictate.

Improvements in the hardware mean that the new mainframes have an even greater level of robustness and reliability; the z in z/Series stands for ‘zero fault’. “I have to reboot my PC every day,” Sanchez says, “you never reboot a zSeries.”

Hard to believe it was once widely held that mainframes, the so-called ‘big irons’, wouldn’t live to see the 21st century. They live to compute another day, still to be found wherever intensive computing power is needed. The humour isn’t lost on IBM, which has taken to giving the latest models nicknames as an in-joke. Last year’s version is the T-Rex, but the S/360, now called zSeries, is one ‘dinosaur’ that doesn’t have to fear extinction.

IT milestones

1969 ARPANET
Forerunner of the modern internet, the Advanced Research Projects Agency Network (ARPANET) was a nationwide computer communications network, the first to use packet switching technology, now a standard for the way data and voice traffic is sent. Using packet switching, a system can use one connection to communicate with more than one machine by assembling data into packets – discrete units of information. The connection can be shared and every packet can be routed independently of other packets. ARPAnet was not, as was widely believed, developed to withstand nuclear attack.

1971 Laser printer
Has its origins appropriately enough in Xerox’s Palo Alto Research Centre and actually uses dry printing technology known as electrophotography – very similar to the photocopier. The original laser printer called EARS was developed by adapting Xerox copier technology and adding a laser beam to it. Incidentally, research at PARC to speed up the connection between the computer and printer resulted in the development of Ethernet, which is still the bedrock of local area networks.

1981 PC
The term ‘personal computer’ existed for years before the IBM PC came to market and even then, many rival manufacturers sold machines that were incompatible with one another. Under an extremely tight deadline, the team leading this project opted to build the machine with off-the-shelf components, founded on the twin pillars of Intel chips and Microsoft’s operating system (that’s why the PC bears no resemblance to IBM computers that preceded it). Its design effectively laid down the blueprint still followed today, as it was based on an architecture specifically designed to be copied by others.

By Gordon Smith