During a visit to a data centre recently, the company's CIO mentioned to me in passing that the computer giant IBM was "ending support for SNA next year". For me, this was almost a Proustian moment, with the now obscure acronym reviving, in my mind's eye, countless headlines, industry disputes and politicking spinning back through two decades.
By any standards, and even by the dramatic standards of the IT industry, the story of SNA – and the rise and fall of IBM's Network Systems business – is an extraordinary one. But unlike, say, the rapid rise of Microsoft and the destruction of IBM's domination in small computers, the history of SNA has largely been left untold. But as with all these stories, there are important lessons, for today – for all businesses, but especially for telecoms companies, and maybe even for Microsoft itself.
From the early 1980s to the mid 1990s, Systems Networking Architecture, a broad set of mainframe-centric networking technologies and de facto standards, was enormously important to IBM's hold on the computing industry. For a period, there were few commercial networks of significance anywhere in the world that did not heavily rely on IBM's technology.
By about 1990, IBM's networking business, if services are included, had annual revenues of nearly $9 billion – somewhere around the size of an SAP today. IBM's domination of corporate networking was greater than Cisco's is today. There was even an SNA ecosystem, with several computer companies making equipment compatible with the SNA standards, which all computer makers also followed.
But the standards were not open – SNA was almost entirely proprietary. So while other companies implemented them as best they could, IBM maintained jealous control over the pace and direction of mainframe networking. As a result, SNA was consistently controversial, as governments and rivals vainly sought to undermine IBM's dominance through both legal and technical initiatives.
This was a prolonged, bitter and colourful battle. Open Systems Interconnection (OSI), a rival set of standards, failed to win wide acceptance in spite of huge investment and the backing of the European community and many public-sector purchasers. After years of wrangling, the European Community also imposed on IBM the same kind of settlement it is trying to force on Microsoft today, insisting the critical interface information for SNA was made available within 90 days of any product announcement.
But suddenly, in the early 1990s, it all began to go wrong – spectacularly wrong. A young competitor named Cisco systems was having success with an alternative approach, sending traffic over a wide area network by replacing the one-to-one SNA ‘session' links with connectionless links. It worked perfectly on the US Department of Defense's Arpanet, progenitor of the Internet, and then on the Internet itself.
Soon, IBM was falling victim to "disruptive innovation": a cheaper, but less functional, product was undermining sales of a product that had become over-engineered for its market.
By 1993, Cisco already had sales of $649 million – a tenth of IBM's – and revenues were doubling each year. Cisco opened up a recruiting and engineering centre directly opposite IBM's own development labs in North Carolina, and soon a large number of IBM engineers were driving into a different car park each morning. "[IBM] Network Systems is just being slaughtered. It is punch drunk," said analyst Anura Guruge at the time. Canute-like, IBM refused to admit the depth of the problem, dismissing the Cisco router as "a big mistake" that simply "doesn't work".
Belatedly, IBM attempted to fix its product line, cut costs, and establish ways for users to build simpler, more flexible networks. But it was too late: Cisco had the momentum. At a press conference in Cannes in the mid-1990s, IBM tried to convince a sceptical press that it could recover. But, unfortunately for the company, the banner that hung over the hotel-front, celebrating the launch for The Mummy, had not been changed from the film festival launch party the day before. It read: "Abandon hope, all ye who enter here" – a phrase the journalists present used mercilessly.
Eventually, IBM admitted total defeat. It formed an alliance with Cisco, and allowed it to supply equipment for most IBM customers' networking needs. The bulk of IBM's networking staff de-camped to Cisco or were made redundant. In just over a decade, one of the world's biggest technology operations – and one that was strategically far more important that IBM's PC business ever was – had crumbled away.
The lesson is for the reader to decide. But here are some clues: voice-over-IP; WiMax; open source.