So is it SOA or DIY?

If I insisted that you should dump your current automobile and buy a build-it-yourself kit car, what would you say? I suspect you’d say that, given the fact you hardly have enough time to sleep these days, you are unlikely to find the time to build a car. You’d probably follow that up with “and you must be nuts.”

What if I told you that for your efforts you could get double the gas mileage and reduce your maintenance costs?

You might counter with “but I’d have to go to the trouble of finding and buying all the parts or purchase a complete kit and then I’d have to learn how to build it, which would require time that I don’t have, and the only way my maintenance costs would be reduced would be if I did the maintenance myself, which would require that I find still more time that I still don’t have, and I still think you’re nuts.”

What you’re really saying is that you’re willing to trade off the cost of owning a completed car for the time and effort it would cost you to build and maintain a kit car. And that I may not be playing with a full deck.

Given that you are so rational about cars, why are you not more suspicious about service-oriented architectures (SOA)? According to a recent Aberdeen Group study, “Enterprise Service Bus and SOA Middleware,” 90 per cent of companies “are adopting or have adopted service-oriented architectures and will exit 2006 with SOA planning, design and programming experience.”

OK, so all of these organizations are experimenting, dipping a toe into the murky waters of the SOA ocean, right? Well, apparently not. The report found that, “In comparison with previous information technology adoption cycles, SOA’s rush into early adoption is remarkable.”

So why are these companies diving so eagerly into SOA? Aberdeen Research Director for Enterprise Integration Peter Kastner, who authored the report, is quoted as saying, “Redesigning business processes, high IT integration costs, and customization challenges are eating up 40 per cent of the IT budget in integration expenditures. . . . SOA is broadly seen as a real technology step forward, with the largest companies, who have the biggest integration problems, leading the way.”

Interesting. So we have the companies with the biggest integration problems adopting SOA incredibly fast even though the concept is only a couple of years old?

Have you guys forgotten about the whole object-oriented architecture hoopla? In an IBM paper titled “The Tao of e-business services” published in 2000, author Steve Burbeck notes, “The lesson we should have learned from the failure of object-based systems is that the way services are described, organized, specified by potential users and discovered amidst the clutter of the Internet will determine the success of B2B services.”

The point of Burbeck’s paper is to argue for focusing on services as the architectural basis for enterprise systems, but unless I’m missing something, why should SOA solutions be better than OOA? The idea that one architecture is going to fix all of your enterprise application problems is at best optimistic. At worst, if you put all your eggs in the SOA basket, it will be a great risk.

I’m not saying that there’s no value in SOA. I’m simply pointing out that there’s a faddish quality to current SOA adoption trends that is out of all proportion. SOA has no real history, isn’t cheap to implement and often results in a collection of spare parts flying in loose formation.

What it comes down to is this: If you like the idea of building something from a lot of spare parts then you should perhaps try building a kit car. Should the car turn out to be a disaster, it will cost you a lot less than building an SOA solution that doesn’t do what you hope it will.

QuickLink: 065161