Influencing SOA Behavior
Determining and shaping the behavior patterns that will sustain an SOA effort through time is often the work of change management specialists. However, it is easier to conceptualize the role of behavior in SOA if we examine the intersection of major influences on organizational and individual behavior. Some of these influences are the following:

• Corporate culture
• Major decision-making processes
• Budgeting processes
• Incentives and penalty structures
• Compensation linkages to corporate goals and mantras
• Portfolio management processes
• Architecture process (definition, acquisition, implementation)
• Architecture practice (solutions development)
• Corporate performance metrics, such as return on invested capital (ROIC), revenue and market share growth, cost controls, etc.
• Promotion and advancement criteria

All of these factors influence behavior of organizations and individuals within an enterprise. In the context of an SOA, however, there are a few key factors that determine how effective an SOA can be without fundamental changes to the organization and processes of an organization. To understand these essential factors, we must locate the "center of gravity" of an SOA. Understanding certain key organizational and process relationships will help with the organization of SOA governance, the design of SOA governance processes, as well as understanding the behavior and incentive models that may be required in order to implement SOA governance. For example, the relationship of the budgeting process to project execution and implementation helps determine how effectively budgeting oversight impacts the resulting architecture of IT systems. Without proper budget-to-project alignment, teams will be inclined to build project-centric services that don't necessarily fit the broader needs of the organization's SOA. The relationship of the acquisition/procurement process to enterprise architecture is another key relationship that has a tremendous impact on the resulting architecture of an organization. Does the acquisition process reflect the goals and standards of the EA organization? If not, how can it be changed to better reflect it?

Another critical relationship is the relationship of EA to project or solution architecture - in other words, connecting architecture via governance to downstream activities (more on this in the next section of this article). If there is a disconnection between enterprise architecture and the architecture that is designed at the project level, then there is the possibility of a disjointed architecture.

These important relationships all point to how sociopolitical forces and organizational forces converge to either facilitate or hinder SOA governance. There are no easy answers to these challenges. However, understanding how these organizational tensions either help or hinder SOA governance will point to a path to implementing appropriate organizational institutions and processes to achieve SOA governance objectives.

Production/Distribution/Consumption: Separation of Concerns Within SOA Governance
SOA, in conjunction with other loosely coupled architectural approaches, forces IT organizations to recognize that teams producing services are not likely to be the consumers of those services. Unlike traditional application development, SOA is built upon the premise that a set of services can be employed within a wide range of applications. In other words, SOAs depend upon the separation of the production and consumption concerns within the IT organization, and a distribution vehicle that allows service producers and consumers to communicate and collaborate with each other. Let's define these production, distribution, and consumption concerns with a bit more detail:

• Production: Identification of and governance over the development and maintenance of existing and newly defined candidate reusable services
• Distribution: Publication of those services for widespread dissemination to potential service consumers
• Consumption: Discovery of and governance over the appropriate use of services within application development projects

While it may be technically feasible to execute the responsibilities within these concerns through manually defined and managed procedures, the reality is that for IT organizations of any size to build out a successful SOA, both political/organizational issues such as those discussed in the previous section, and appropriate tooling such as a services repository/registry that automates both service governance and service distribution are essential. The following sections of this article highlight how a services repository/registry can be employed to support key best practices within the production/distribution/consumption life cycle inherent within SOA.

Production Best Practice: Pragmatic Service Definition
Services within an SOA cannot be developed in a "bottom-up," ad hoc manner. Bottom-up development of services is inherently driven by immediate project needs - how do I solve this specific problem with a specific implementation (often driven by the influence of existing applications and their behaviors masquerading as true business requirements). What happens when an organization defines and implements its services with this mindset? The service layer simply becomes YALOT (yet another layer of technology) - more spaghetti code of a different form that didn't improve our business process flexibility, but simply implemented a monolithic application in a different technology.

However services also cannot be defined solely in a "top-down" manner. Top-down business process analysis left to its own devices leads to either "analysis paralysis" - continual refinement of a model hoping to reach perfection (which never comes), or "Big-Bang" projects - trying to define and implement everything at once, usually with disastrous consequences (most typically a combination of "death march" projects and cost and schedule overruns).

Ultimately, what organizations need to make progress in SOA is to develop a coarse-grained business model driven by key business processes (not all of them, but only a representative set of high-priority processes to begin with). Architects and business analysts should collaborate to build this model using those processes to extract and define a normalized set of functions, then grouping those functions together based on behavioral affinity (read components and interfaces for those of you who are UML centric) as a strawman set of initial target service definitions.

At this point, we have a useful framework for the "real work" - detailed analysis, design, and implementation of the services we need for our current set of prioritized projects. Based on business process (and project) prioritization, we identify the needed services from our business reference model and formalize the service definition for these prioritized services. Ideally, each service should be driven by the requirements extracted from at least two separate processes - designing a service based on a single use case is very likely to result in a fragile and narrowly defined service that will not be flexible enough to meet our next set of prioritized projects. Our formalization efforts are likely to result in modifications to our business architecture - which is just fine! We can iteratively enhance our architecture as we make progress towards service implementation.

Production Best Practice:
Recommended Service SDLC Governance/Review Checkpoints
Now that we have our first set of services defined, we need to build and deliver them. Developing services within an SOA (i.e., for purposes of reuse across multiple applications) usually requires more of the production team than a single-use component, module, or object. In order for a service to be considered reusable, it must be maintainable, discoverable, and consumable. Maintainability introduces such concepts as version control (which we will address in more detail later in this article), models and other design documentation, and requirements traceability (why was the asset implemented in this way from a technical and business perspective). Discoverability forces us to consider how we help potential consumers of this asset find the asset in a timely fashion - via keywords, domain taxonomies, and mapping to models, for example. Consumability involves looking at the asset from the point of view of the downstream project planning to use the asset: Is there a user guide, a well-documented API, sample client code, and other artifact available to help the user rapidly understand how to apply this asset to the project at hand? Are dependencies on other assets (and to prior versions of this asset) specified and easily navigated?