CHALLENGE 3:

Rethink Your Talent Pool

In addition to architecture staff that develop and manage the SOA's big picture, the CIO will need a development staff that is comfortable with both business processes and the services approach to developing applications. And business experts who can partner with IT on process identification and implementation are also critical - the CIO's organization can't do it all.

"Retraining, hiring and redeploying are all needed," says Rogers, so most CIOs will bring in some key hires experienced with SOA and retrain the existing staff in SOA-oriented thinking. "It's an ongoing process, so start out small and do a lot of retraining," says TrueCredit's Metzger.

CHALLENGE 4:

Apply SOA Principles to Your Data Too

Often overlooked in initial SOA efforts is the importance of treating data - not just application functionality - as a service, says ZapThink's Schmelzer. In a SOA, multiple services residing in multiple applications might combine to execute a business process. If each service uses different data sources, or even the same data source in different ways, the results might not be what you expect, he says.

For example, Metzger enforces strict separation between services that access and store data and those that act on the data, such as performing calculations. He also keeps presentation services - those that format the data for the user or for reports - separate from other services. This helps ensure that all services have the same context for a specific piece of data.

THE PAYOFF:

Real Business-IT Alignment

All of these challenges underscore how different SOA is from traditional enterprise technology efforts. The CIO must put the technologist hat to the side and become an advocate for business processes. Better knowledge of those processes leads to better technology design, which reduces the cost of maintaining existing systems - something that now accounts for 70 percent to 80 percent of IT budgets, says Schmelzer.

"We've been talking for years about IT-business alignment," says IDC's Rogers. A SOA lets you actually achieve that goal - if you can develop the architectural vision and execute on it over the long term. The process is not easy, but early adopters like Peck, Metzger and Stubbs would never go back to the traditional IT approach. "A lot of [D.C.] government would not work without SOA," says Peck.

SOA's Technology Underpinnings

Although service-oriented architecture (SOA) is not about technology, technologies can help you deploy an SOA. While there are many good products available, the market is just beginning to mature, notes Ron Schmelzer, a ZapThink senior analyst. That's why so much of the supporting systems for SOA efforts today rely on human effort, such as having a project review team that knows what services already exist and can be used for a proposed new project. Still, a few technologies are emerging now to help ease the SOA effort:

1.Enterprise service bus (ESB): A variation of an application server or EAI platform, an ESB orchestrates the communication among services, as well as with the user and any data sources. "SOA in today's world is very message-centric," says Sandra Rogers, program director for SOA, Web services and integration at IDC (US). Some large enterprises will have an appropriate messaging system in place, she says, but most will need to upgrade their systems to adopt current standards and move beyond simple data transfer and application integration.

2.Service registry or repository: This database system tracks the various service components available for reuse and publishes available services to business analysts and external partners so they know what's available. For example, Sprint Business Services makes its registry available via the Web to encourage use of its services, notes Vijay Musuvathy, manager of solution architecture. Unlike a traditional database, a registry must also store the context of the service, which defines when and how it should be used, in the form of metadata.

3.Master data management systems and metadata repositories: These two categories of products help enterprises manage their data in the same distributed, composite way that they'll manage business logic through SOA services. This ensures that data's meaning is understood by all services that use it, so no hidden assumptions end up corrupting the results that services deliver when using or generating the data.

4.User interface management: In a few years, "the biggest development will be the whole user interface space", says IDC's Rogers. As more services exist in everything from Web portals to office tools, "everything will be a producer or consumer of a service", and managing the user interface across everything will require a service-oriented approach as well, she says. The emergence of asynchronous Java and XML (Ajax) "is a start but it's not enough" for SOA's broad reach, Rogers says.