Organic SOA: Developing a sustainable lifecycle

Creating a scalable service-based enterprise ecosystem requires a broad architectural vision

Nothing in enterprise IT should exist in isolation. That’s the big idea underlying SOA -- leverage existing data and application assets by provisioning them as shareable services, and when you build new functionality, create it as a service so that current and future apps can draw on its power.

That architectural vision can result in a harmonious, scalable ecosystem or an unholy mess of point-to-point spaghetti, depending on how you plan, build, deploy, and manage. One great advantage of SOA compared with other models is that you can start small and grow organically. But the idealized endgame of SOA, which few organizations will ever reach but all should bear in mind, is an entire enterprise infrastructure that functions as one giant service-based meta-application that can meet any business requirement.

Just as applications have lifecycles, so should an SOA. Yet whereas application lifecycle management organizes requirements, assets, testing, and change control for finite IT projects, an SOA is a broad initiative that lacks a completion date and requires a unique master plan.

To develop this lifecycle, we tapped InfoWorld’s extended family of practitioners to examine each node in the SOA lifecycle. David S. Linthicum, an enterprise integration veteran and InfoWorld’s Real World SOA blogger, has provided the basic framework -- plan, discover, design, deploy, and manage -- as well as specific advice on taking inventory of current data and apps, an examination of SOA deployment issues, and succinct advice on services design. James Borck draws on his experience with business process modeling to illuminate the planning process. And Phillip J. Windley wraps it all up with a discourse on SOA management.

The advice here applies to the broadest array of SOA implementations -- from small-scale deployments that serve a limited set of requirements to projects that encompass an entire organization. Because the focus is squarely on the lifecycle, we have not addressed infrastructure issues, such as the messaging technologies that enable services to communicate. SOA security also falls outside the lifecycle, but Linthicum provides his security viewpoint at the end of this article nonetheless.

As enterprises get in the SOA game, they’ll likely traverse the lifecycle more than once as initiatives grow more and more ambitious. Initial Web services efforts driven by IT may give way -- and become modified beyond recognition -- to bigger plans driven by top-down business process analysis. You may not reach, nor may there exist, an SOA end state. But at least the path to SOA success is getting clearer.