Breaking SOA bottlenecks

legacy technology for your services,” says GM’s Zhang. SOA per se does not require new tools, though you may discover that having new tools can make the effort easier.

For example, an ESB can consolidate messaging, logging, metrics, application management, and alerting functions into a common engine, simplifying the infrastructure, says Pete Conner, CTO of Primitive Logic. But you may have these functions already in the forms of messaging middleware, XML managers, and EAI systems. And you can’t assume that any specific ESB will do the job you need, as there is no standard definition of what an ESB should do, Conner says.

“Often, an ESB is very much an extension of EAI; it’s about integration, not about reuse or contracts,” he warns.

 

Addressing the data problem

One of the most unloved parts of IT is ensuring the integrity of data. Data comes from multiple sources — disparate applications, outside partners — typically with different assumptions about meaning and usage. That has led to difficult, ongoing efforts to rationalize data through transformation, scrubbing, and master-record systems. In the past, the effects of inconsistent data models and metadata could be confined to the interfaces between applications, usually through transformation efforts.

In SOA, however, individual services that make up a composite application may use data from a variety of sources, so the data integrity problem can no longer be neatly contained. SOA requires an underlying data architecture, so no matter where the data originates, the metadata describing it is consistent enough to be understood the same way by all services using it. “At runtime, decisions are made by a set of rules that early on master data to be correct, so core data management and master data management is very important,” says Tata’s Mohanty.

“SOA magnifies your data issues,” says Ed Vazquez, a vice president at the MomentumSI consultancy. “Before, you could paper those over, but not with SOA. A reusable service means reusable data.”

This means different business units must finally agree on standards for things such as customer information, even if no one group uses all of that information, and that IT must

focus its data-cleansing efforts on the sources of data rather than just on what makes it into the data warehouse. “That’s why at the same time you are defining the enterprise process model and architecture, you need to identify the entire data and semantic model,” says Tata’s Iyers.

If you have a data mess on your hands, however, remember that cleaning it up will be time consuming, and will probably require long, boring meetings that involve the business side as the details of consistent data representation are ironed out.

 

Governance, governance, governance

By its very nature, an SOA encourages services to be shared across organizational boundaries. Among departments or business units, disputes arise over how to build, consume, modify, and retire services, as well as over levels of service availability. SOA breaches the usual departmental silos, so conventional management approaches tend to fall short.

Many early adopters are feeling their way through these issues. “Governance is a hard sell and requires organizational change,” says Bobbie Young, CTO of Unisys. She recommends that organizations evaluate their capabilities to manage such change before going down an SOA path and discovering that endemic resistance blocks any real SOA benefit from being realized.

In some cases, resistance may be unexpected. For example, although the SOA promise of agility through the composition of reusable services sounds appealing to business execs, it also can threaten some business leaders who place higher value on consistency and predictability. “Often, the CFO and COO like the fact that the system is inflexible because it makes it hard for the little guys to change the processes,” notes AMR analyst Finley. And the planning stage of an SOA can resemble Congress trying to pass a bill, he says. “There’s a huge amount of politics about how to standardize processes. It’s easier to say, ‘ This is it.’”

Inertia can also be a barrier, notes Keane’s Daly: “With SOA you’re talking about enterprise-wide standards and approaches. But most companies have so much invested in their systems that there’s no incentive to change or share.”

Other governance issues center around control of the