Information Age: News, analysis & insight for IT & business leaders

Sometimes, technologies – and the ideas around them – are so successful that they ride a huge wave of hype, even reaching into the general public’s dim consciousness. The service-oriented architecture (SOA), deeply technical as it is, may even prove to be one of these.

Information Age itself has been reporting on SOA for perhaps five years, if we think of SOA as “enterprise web services” with some architectural control. But almost from the outset, it was clear that the hype around SOA was running far ahead of the implementations, and, indeed, of the technical capabilities of the software and systems suppliers.

Early evangelists, to be fair, did try to warn of this. In 2002, IBM’s global head of web services told Information Age that SOA was a 10 to 15-year project; and when SAP introduced its Enterprise Service Architecture – its huge project to break its monolithic application in a vast bundle of services – insiders told us that their management were being over-optimistic. It wouldn’t take just three years to unbundle SAP, but seven or 10 or more.

Gartner, the analyst company, of course, has a model for this pattern – the hype lifecycle. A huge peak of hype is followed by a slide down the trough of disillusionment, followed finally by the gradual rise up the slope of enlightenment, as the details are worked through and the benefits finally reaped.

In the case of SOA, it is difficult to know exactly where we are on the lifecycle, not least because of the immensity of the project. One thing is clear, however: we are a long, long way from reaching the enlightened state of agile computing. This is largely because, as with most big projects of this nature, the job starts to look bigger – much bigger – the more you look at it.

The full extent of this came home at a recent seminar organised by Unisys on what it calls the “real time infrastructure” (RTI). In Unisys’ view of the world, every business aspires to be an RTE – a real time enterprise, capable of reacting to change rapidly and painlessly. Applications and services and processes are mapped and modelled, used and reused, coupled and uncoupled at speed by business people benefiting from the architectural fluidity of the SOA.

None of this can work, however, without an underlying, flexible architectural platform – the RTI. That is the part that ensures that if a new service is needed, or another one needs changing, there are processors and disks to support it, that some kind of quality of service, availability, reliability and security can be guaranteed, and that the most cost effective choices are made. And all this must be done immediately – in real time. No lengthy planning cycles, no stalling by the IT department, no long budgetary reviews.

According to Gartner, most businesses today are barely beyond basic RTI – and it will be at least three to five years before an early majority has reached the more intelligent, autonomous, real time stage.

Unisys’ argument is that all this needs a careful methodology that works progressively through the lifecycle, reaping savings along the way through, for example, consolidation and virtualisation, but only delivering full agility sometime towards, or probably after, 2010.

But the RTI layer is only part of it. Everywhere, there are signs that suppliers and integrators are starting to realise the complexity of the mission they have undertaken. In August, for example, Hewlett Packard splashed out $4.5 billion to buy Mercury Interactive (see page 42), a company that has tools for testing services and software, and a register for finding and managing them. It already has an asset management company, Peregrine, in its portfolio. The vendors are starting to realise that a big, distributed, service-oriented management suite will be needed to run the SOA.

There are many other pieces that still need to be put together. Tibco, for example, is arguing the SOA needs to be modelled to carry event information, which will allow rapid reaction and prevent over-complexity. Other companies are trying to address big gaps in the area of orchestration, building composite applications and services, managing trust, security and certification, process modelling, policy management, dynamic resource allocation, and financial control and modelling. Many of these pieces are missing, proprietary or immature.

SOA is often referred to by marketing people as if it is something one or two of their customers did last year, with great results. But no-one should be fooled: there are still some big holes in the middle and few companies, if any, have really got it working together. The slope of enlightenment? It may not always be steep, but it is still long and it is still uphill.