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.
