Enterprise Architecture

9 October 2003

Just recently I've picked up a couple of bad reviews on Amazon for P of EAA because there is nothing in the book about enterprise architecture. Of course there's a good reason for that - the book is about enterprise application architecture, that is how to design enterprise applications. Enterprise architecture is a different topic, how to organize multiple applications in an enterprise into a coherent whole.

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.

The problem for central architecture groups is that they are driven by IT management, but the applications they are looking to organize are driven by business needs. If an application team is told to do work that doesn't benefit their application directly, but makes it easier to fit in the architecture, there's a natural reluctance to do it. Furthermore they have the ace card - the business sponsor. If the business sponsor is told the application will ship four months late in order to conform to the enterprise architectural plans, then they are motivated to back up the application team when they say no (spelled "we'll get around to it later"). Since the application is directly connected to providing business value, and the central architectural team isn't, the application team wins. These wins cause the enterprise architecture initiative to bust.

To avoid this the enterprise architecture initiative has to recognize and submit to the political realities.

  • Understand what the business value of any enterprise architectural initiative is.
  • Make sure that any work is supported by incremental short term gains in business value.
  • Minimize costs to the applications

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.

These reasons tend to lead me toward a messaging approach to integration. While it has its faults, it's something that can be applied with minimal impact to existing applications.

By the way, enterprise application architecture can have a big impact upon enterprise integration. Applications that are nicely layered, particularly with a good PresentationDomainSeparation, are much easier to stitch together because you can more easily expose the applications functionality through services. This isn't a cost to the application, because good layering makes the application easier to maintain as well. However too few application developers understand how to do PresentationDomainSeparation. One of the best things an integration group can do is to support education and training to help them to do this (an approach that's best supported if they act like Architectus Oryzus rather than Architectus Reloadus). So in that sense my book has a lot to do with enterprise architecture.