As it turns out, I can get pretty cynical about enterprise architecture. This cynicism comes from what seems to be the common life-cycle of enterprise architecture initiatives. Usually they begin in a blaze of glory and attention as the IT group launches a major initiative that will be bring synergy, reuse, and all the other benefits that can come by breaking down the stovepipes of application islands (and other suitable analogies). Two or three years later, not much has been done and the enterprise architecture group isn’t getting their phone calls returned. A year or two after that and the initiative quietly dies, but soon enough another one starts and the boom and bust cycle begins again.
So why does this cycle happen with such regularity? I think that most people involved in these initiatives would say the reason they fail is primarily due to politics – but what they often miss is that those political forces are inevitable. To succeed in these things means first recognizing the strength of those political forces.
. . .
[Duct Tape Architecture?] A good way to think about this is that these initiatives should less about building an overarching plan for applications, and more about coming up with techniques to integrate applications in whatever way they are put together. (After all ApplicationBoundaries are primarily social constructs and they aren’t likely to conform to anyone’s forward plans.) This integration architecture should work with the minimum impact to application teams, so that teams can provide small pieces of functionality as the business value justifies it. I think you also need to focus on approaches that minimize coupling between applications, even if such approaches are less efficient than a more tightly coupled approach might be.
Be sure to check out the “Patterns and Best Practices for Enterprise Integration” link.