Service-oriented architecture, or SOA, is the modern notion of connecting systems together at both the information and service levels. Indeed, enterprises are racing to enable their existing applications to externalize services, as well as build the appropriate integration infrastructure around it.

However, extending your SOA to automate your business means you must work and play well with other organizations. Managing services, orchestration layers, and connections between companies is not as easy as one might imagine.

Truth be told, while we've understood the value of SOA for some time now, the concept is still new to most enterprises. Not until the advent of Web services did we have a widely accepted standard and enabling technology that allows us to access all types of systems through a common services interface. In fact, we may be at a point in time where more is understood about the technology than the ways in which it fits into the enterprise or value chain. Organizations seem to adopt Web services without thought of strategic fit and function. Adoption is only half the battle.

To address strategic concerns, many enterprises are attempting to figure out how to best leverage SOA within their firewalls, as well as between organizations that are, or should be, part of their business processes. This notion of extending SOA to externalize and share processes and services is really the ultimate destination for SOA; certainly it was the vision for Web services.

Value of Services
Today most interorganization electronic business is conducted using traditional information-oriented mechanisms such as EDI or simple FTP exchanges. These exchanges deal with simple information, and they tend to occur within nightly and weekly batch transfers, meaning that latency is a consideration.

While traditional information-oriented exchanges are the way business gets done today, there are two basic needs for electronic business: the need to have access to information approaching zero latency, and the need to view external customer or supplier systems as sets of remote services as well as clusters of information sinks and sources.

The ability to leverage services will provide organizations with a clear advantage. In addition to the ability to see information in real time, they can abstract application behavior and leverage many remote services inside their enterprise systems as if they were local. This is the basic notion of Web services, so we won't get too deeply into it here.

Access to services implies that business processes existing in and between companies can be coupled at the services layer, meaning that services are shareable (if allowed) among the partner organizations. For instance, the service that defines how inventory is allocated supply chain-wide is shareable, and thus the service is not only consistent, but does not require reinvention within each organization. Moreover, since these services are always visible, information bound to these services is produced and consumed in real time. In essence, you're creating a virtual set of applications that exists between trading partners and allows those trading partners to function like a single entity, and thus service common business processes as if they existed in a single company.

By leveraging this type of architecture, businesses have the opportunity to reduce inventory costs a hundredfold. For example, all manufacturer systems would have service- and information-level visibility into all retail systems, and all of the parts suppliers have the same access, and perhaps their raw materials providers as well. With everyone sharing both information and services, business processes are fully automated and inefficiencies, such as overstocking, understocking, or manufacturing delays, drop dramatically. What's more, and perhaps more important, customer satisfaction goes up since items they demand are available, and at the best possible price.

Functional Components
So, how do you begin sharing your SOA? You must first break the SOAs down into several basic components before attempting integration. These include:

• Private and public services
• Public and private processes
• Data and abstract data
• Monitoring and event management
• Points of integration
• Directory services
• Identity management and security services
• Semantic management

Public services are those that are redundant within your trading community, such as logistics, inventory, or billing. By exposing these services to outside organizations, you allow them to share the service and thus avoid their own development cost, and also allow them to leverage a shared service as a point of integration and a binding point for common processes.

There are a few key criteria for selecting services that are public, or, exposed to trading partners. First, the service should be redundant to two or more entities. In other words, you solve the same problem for several partners. Second, the service should be unique to the trading community; otherwise it makes sense to look for other public services to solve the problem. Finally, the service should offer ease of integration, including the ability to discover semantics as well as interfaces.

In order to make services public, you must create or leverage an existing shared directory service that allows those outside of the organization to locate, discover, and leverage the service you deem to be public. Directories may be proprietary, LDAP-based, or UDDI-based (or mixed and matched). Typically these directories are public, but support the notion of public and private services, processes, and semantics.

Public and private processes provide orchestration of services, binding them together into a business process to drive information movement and invocation of services. You may consider processes or orchestrations as a group of services gathered together to solve a particular business problem, an overriding control mechanism, if you will (see Figure 1).

There are three types of processes to visualize enterprise and cross-enterprise processes: private, public, and specialized processes.

• Private processes exist at the intra-company level, allowing the business user to define common processes that span only systems that are within the enterprise and not visible to the trading partners or to community-wide processes. For example, the process of hiring an employee may span several systems within the enterprise, but should not be visible to processes that span an enterprise or trading community or other organizations.
• Public processes exist between companies and consist of a set of agreed-upon procedures for exchanging information and automating business processes within a community. This is the core notion of intercompany SOA, since it's really the concept where we create intercompany orchestrations.
• Specialized processes are created for a special requirement, such as collaboration on a common product development effort that only exists between two companies and has a limited life span.

In the world of BPEL, process is one of two things:

• Executable business processes that model actual behavior of a participant in a business interaction.
• Business protocols that use process descriptions specifying the mutually visible message exchange's behavior for each of the parties leveraging the protocol (does not reveal internal behavior).

To this end, BPEL leverages a well-defined language to define and execute business processes and business interaction protocols, thus extending the Web services interaction model by providing a mechanism to create meta-applications - process models, really - above the services that exist inside or outside the company. What's both different and compelling about BPEL is the use of a common syntax that is designed to be transferable from process engine to process engine. This is in contrast to other process integration standards, such as BPMI or WFMC, which are more about approaches than a common language. There is more momentum beyond BPEL, and all technology vendors are declaring support for BPEL.

The notion of data and data abstraction, in terms of intercompany SOA, lets us think about collections of data or services as abstract entities, thus represented in a form that is most useful to the integration server or the application integration architect. It's this notion that provides for the grouping of related pieces of information, independent of their physical location and structure, as well as defining and understanding what meaningful operations can be performed on the data or services. Thus, we can create any representation needed for data that exists anywhere, and bound to any service.

What's more, we need to separate the implementation from the abstraction itself. This allows us to change the internal representation and/or implementation without changing the abstract behavior, and allows people to use the abstraction in terms of intercompany SOA without needing to understand the internal implementation.

Monitoring and event management encompass the ability to analyze all aspects of the business and enterprise or trading community to determine the current state of the process in real time, and adjust those processes as needed and in real time. Optimization, or the ability to redefine the process at any given time in support of the business and thus make the process more efficient, is an aspect of event management (see Figure 2).

Points of integration allow services to interact with other services, or perhaps an orchestration layer. Services, especially those build for intercompany SOA, need to be designed to interact with other systems. For instance, they should provide more robust discovery of metadata and management of connections. Thus, the service developer needs to architect a service as a point of integration, not simply a point of abstracted functional behavior.

Identity management and security services seem to go without saying when you think of intercompany SOA, due to the naturally occurring exposures, and a detailed discussion is out of the scope of this article. However, it's worth a mention that there are three A's of security which you need to consider: authentication, authorization and audit.

In identity management, especially in the work of inter-company SOA, you must pay a lot of attention to authentication. However, rights and permissions are identity-based attributes and should play a very important role in the identity management.

Finally, we need to manage semantics between any number of systems that have very different application semantics and ontologies. We typically do this through a semantic repository, where the semantics are understood and persisted, and a semantic mapping layer that understands the semantics of the source or target systems and can account for the differences during runtime.

Share and Share Alike
It's not a matter of when intercompany SOA will become a reality. The evolution is under way today, as organizations attempt to extend their SOAs to other partner organizations, and want service-level visibility in return. The key words here are real-time automation and visibility. You can either get good at it now and create a more competitive and responsive business, or play catch-up later.

However, like any other new way of doing old things, you have to consider the architectures and how they mesh together over time. The new dynamic here is that you're dealing with many IT departments and many approaches to building applications The best approach is to understand your own value, and work directly with your partners to ensure that everyone is adjusting their way of thinking. The goal is to build sharable services and processes that automate your business, as well as processes between businesses.