Nov 23, 2009

Is SOA a Bad Idea?

In consulting at various companies, I'm seeing a lot of "SOA failure". Oh, the technology works and the tools are capable. Further, their developers are using a variety of SOA(ish) techniques. But their formal SOA programs, their Integration Centers, are failures. (I define failure as an unhappy department, senior management that is unhappy with the results, a general feeling of failure, unable to demonstrate and most likely without ROI, and an impression that Integration is in-the-way and provides unsufficient benefit versus the effort to engage the process.)

These projecs or departments have almost always started bottom up, they are an IT department solution to pinpoint (narrow) IT problems. They've never really mapped out how they're going to get from here to "there", and have defined "there" as only an IT solution without connecting to concrete business goals.

First a basic question, do you want to be “there”? The easy answer is yes, integration is getting more complex and must be managed, while at the same time demand for systems agility is increasing. Like the change from procedural oriented programming to object oriented programming, the benefits are real and the future of IT will be “there”.

But a major mistake being made by many IT organizations in moving to Service Oriented Architecture is moving for purely technology reasons rather than business reasons…

Example: A regional utility company called me in as a SOA and Integration consultant. We spent a day reviewing the SOA model and benefits, issues such as reliability and security of the new model, and discussing how an organization begins making such a transition. At the end of the meeting they said, “great, that was very helpful and enlightening. Now we know we don’t want to do that.”

I was a bit taken aback. I tried to explain that like the transition from procedural to object oriented programming, the SOA model was going to be part of the future. Just like procedural programming was left in the past and exclusively in legacy systems… At which point they stopped me and said “but we still do procedural programming.” Huh? “We’re still doing new development in procedural methods, languages and environments.”

“See,” they explained, “we need absolute systems reliability. We know procedural technologies. They are thoroughly understood, all operational aspects well known. We know exactly what we’re getting and how it’s going to work. Even if other technologies offer a variety of benefits, our primary concern is reliability, supportability, stability.”

In another ten years they’ll be doing SOA. But today their business concerns require they NOT do SOA. Business and IT management concerns of environmental complexity, systems and business process agility, and gaining futuristic high speed full systems development advantages through applications assembly and BPM (Business Process Management) capabilities are great reasons to begin moving towards SOA. “It’s a bunch of really neat technologies” is not.