Paradigms Shift, But Architecture Is Forever

The adoption of Web services may well be the most important idea to impact application architectures for the next three years. However, for many companies that decide to unreservedly embrace it, it won't be much of a big deal. While that probably sounds like a complete contradiction, the answer lies in architecture: paradigms shift, but architecture is forever.

Last month, one of Giga Information Group Inc.'s internal discussions raised the question: Does the use of Web services make central architecture groups and architects redundant? After all, with Web-services interfaces, every development team and workgroup in the organization can choose its own technology path, and Web services will take care of the integration, right? A number of my colleagues across various disciplines--project management, architecture, infrastructure, middleware--shared our opinions, and the consensus was unanimous: Web services highlights the need for architecture--it doesn't eliminate it.

To understand why, it pays to take a historical perspective: The way things change is usually more enlightening than the way things are. Although Web services is the hottest buzz word in IT circles right now, to the architect it is only the latest in a long line of architectural "paradigm shifts" that have redefined the way applications communicate with users, or with each other. This trend can be traced back to the first time somebody hung an intelligent peripheral off of a mainframe. In recent years, we have seen client/server, Web, and now Web services.

Although each generation of technology impacts the implementation, each wave has only strengthened the most basic lessons of architecture. Those fundamental principles include the separation and layering of code (of which the application server is the most recent expression); the abstraction of data and metadata (dominated today by XML schemas); and the ever-increasing reach of information services (both to more people through mobile devices or voice interfaces and to more systems through Web services).

The Next-To-Last Word
Each wave has been touted as the ultimate in application architecture, but IT organizations with a strong architectural core take a different view: none of these waves was the last word, and Web services won't be, either.

The most disciplined organizations knew that when they were building their Web generation of applications, Web interfaces would not be the only users of those applications, and they architected them accordingly. Even without knowing that mobile and wireless user interfaces, for example, were just over the horizon, they were smart enough to know that the Web was not the last word and that something would follow. Organizations that didn't follow architectural principles found they had to discard their client/server applications to build for the Web, and have to discard their Web applications to build for Web services: generation after generation of "instant legacy" systems. It's no wonder that every single paradigm shift is accompanied by a flock of vendors proffering "screen scraper" and "stream scraper" products that promise to revitalize the last wave by presenting it as the new.

The other striking note from a historical view is that in recent years, each technological wave has come along almost twice fast as the previous one--a kind of "Moore's Law" for architecture. Client/server went from early adoption to being supplanted by Web interfaces in under a decade. The Web was the primary interface for less than five years before mobile devices were touted as the next wave, with claims from some vendors that mobile users would rapidly outnumber desktop users. And now, most organizations are barely getting to grips with mobile and along comes Web services.

Brother, Can You Paradigm?
In fact, the waves now come so fast that they crash upon each other, such that many IT organizations will be asked simultaneously to deal with the impact of mobile, Web-services, and maybe voice interfaces in the next 18 months. Without architecture, that's a hopeless cause. My colleague Mike Gilpin has studied the impact of software infrastructure on a number of organizations and, as a rule of thumb, those with strong architecture are able to adapt to a new paradigm, such as mobile devices or Web services, in one-third to two-thirds the time and cost it takes their non-architectural competitors.

So why doesn't every IT organization have a strong architectural function? There are lots of complex reasons, including organizational boundaries and corporate culture, but one of the most important is the difficulty in measuring the value of architecture. A simple cost/benefit approach sees architecture as a large sunk cost with little immediate return. Experience shows, however, that architecture usually pays for itself over and over again in the long run, by insulating you against the constant waves of technology change.

More subtly, however, architectural investment is still valuable even if you never have to use it. When you invest in architecture, one of the things you are buying is flexibility: the option to make a decision in the future at lower cost. Even if you eventually choose never to exercise that option, it has a value today, and a proper valuation of architecture takes it into account. Unfortunately, simple business models such as cost/benefit or return on investment don't value flexibility effectively, which is why at Giga we use a model called Total Economic Impact, which evaluates costs, benefits, risks, and flexibility. Architecture needs to be valued with foresight, not costed with hindsight, and that's what flexibility measures.

I don't want to understate the impact of Web services. It's going to change the development tools you use, the languages you write, the processes you implement for testing and quality assurance, the way you design, implement, and expose business processes, and yes, even the role of architects. But it won't change the fundamental principles of architecture, and whatever comes beyond Web services--and I promise you, there will be something beyond Web services--the companies that will be best placed to profit from it will be those that made architectural, not tactical, investments this time around.