In this article I'll cover how a model-driven design and development process can pave the way to a Service Oriented Architecture that will in turn produce accurate applications that meet business needs and increase development team productivity.

SOA is hot topic in our industry right now. One of the important aspects of SOA is that it's a business approach and methodology as much as it's a technological approach and methodology for developing distributed computing systems.

SOA is a software architecture for building applications that implement business processes or services by using a set of loosely coupled components orchestrated to deliver a well-defined package. With SOA, existing software assets already in use are packaged in such a way that they're self-contained, loosely coupled, and can be called and executed by other new applications as a service. Any new computing modules would be created as services the first time around, and so would enter into the services library to be used over and over again. Well, to call a service, we have to know how to call it, where to call it, and what to expect back as a response; that is, a service must have a callable interface that's well defined, well published, and well secured. In short, not only should a service be good quality code, it must be well understood and well documented.

How Many of You Have Done This or Seen It Happen in Your Organization?
A business analyst sketches out a system feature-by-feature, either on paper or using a flowcharting tool and hands it over to a team for development.

When the system comes back from development for testing, the analyst returns to his sketch and system tests the functionality of each feature, one feature at a time.

Whenever a feature doesn't perform as the analyst thinks he defined it, he makes a note that goes back to the developers, who make changes to the code, and the whole thing starts again...all manual, all by hand. And remember the testing team isn't always made up of analysts who wrote the "sketch"...it's a perfect world assumption that the analyst and some business users are testers along with the testing professionals. How many understandings of the "system" are in that group and how many "notes" go back to development teams? How Neanderthal!

We are no longer Neanderthals. We now have modeling tools. We now have iterative and prototyping development processes. Programming language technology has advanced such that all of our well-meaning attempts at good modular programming over the years can be expanded into bigger modules with more of a business meaning. We recognize the need to communicate and collaborate. It's taken us 40 years...but we're getting there. The idea of architecture implies a thoughtful plan using a standard set of guidelines and rules to construct a blueprint. A good software architecture will have the blueprints for how data is stored, how users interact, how blocks of functionality (services) are linked together to form programs, how programs communicate, the capacity needed of the hardware, and so forth.

The idea behind SOA is that these blocks of functionality (services) are available to be used by many programs, and that, over time, when you have more and more services those programs can be compiled and put together more quickly. And since the services are already defined, their design can be reused too. Also, if a service implements a specific business function or rule, it's simpler, more cost-effective, and less time-consuming just to change the related service code rather than search through all the independent programs for code that may have implemented that business rule and then change all the programs using it like we used to.

Of course, you still have to test it, but the actual design and development time to implement a change from the business is much less of an impact when you have models to work from, impact analysis reports to guide you, and independent services to change.

A set of loosely coupled components does the same work that used to be done inside tightly structured programs. But now the components can be combined and recombined in many ways. This makes the overall infrastructure more flexible and agile. This way SOA can make it easier and faster to build and deploy IT systems that directly service the goals of a business and we can prove it because we have our models of the SOA architectures linked directly back to the model of our business goals, business requirements, and business processes. Not only do we reuse the code, we reuse the models.

Model-Driven SOA
The whole point of SOA is to make business applications more manageable, more flexible, and more responsive to change. Businesses are constantly changing how they do things - not necessarily changing what they do. Models allow the business people to focus and change the business processes and workflows without having to focus on the technological plumbing. The IT folks can focus on designing, changing, and improving those business processes by codifying the business services. The models are their common point of communications and it documents the agreements of what needs to be done. Models can automate and drive impact analysis for better communication between IT and business and, when enough textual information has been captured, most modeling tools can generate quality code that provides a good start toward overall development.

Sybase's view of modeling (Figure 1) is that in an enterprise world your modeling environment has to be able to address the broader range of application classes and has to integrate them. You do not want a silo'd environment; a SOA environment is all about sharing and reuse. Your models have to be able to import, export, capture, forward generate, reverse engineer, and report between the common implementation techniques of service-driven, composite, event-driven, mobile, and federated systems using today's integration protocols.

To Tie It All Together
Business Process Modeling, Data Modeling (with conceptual and physical data models, data warehouse models, and data movement through the information liquidity model), Application Modeling with UML (all diagrams for UML 2.0), XML modeling and Free Models (a semantic-free diagramming and drawing facility that can be adapted to document ANY notation desired) are all used to built applications today.

Requirements should be attached or linked to any elements of any models, with multiple requirements documents tied to multiple models simultaneously for complete requirements traceability and visible as a document or a matrix. Models should be used to round-trip engineer to process engines (BPEL engines), databases, application servers as well as object-oriented development languages including VB, .NET, C#, and Java. By using patterns and templates, you can take a model-driven approach to development on J2EE, .NET and Web Services. By using techniques like object/relational mapping and data source mapping as well as integrated metadata management, you can link and synchronize all the models together. Changes that occur in one model can be communicated effectively and easily as change requests to any other model type to ensure a complete, cohesive, and consistent response to change throughout the project team, and throughout the enterprise. With your common metadata repository, you have a secure environment to manage, share, and collaborate on modeling projects.

SDLC
Modern software development systems are usually developed using model-driven methodologies and processes. The nice thing about comprehensive modeling tools is that all the stakeholders throughout the SDLC (Software Development Lifecycle) have their own view into the metadata or information about the system to be developed. This common detailed definition of what must be built to meet the customer's needs is how development teams can be productive and not waste time because of misunderstandings or because one expert didn't pass some information on to another (Figure 2).

• End users and business analysts have a business-centric view into the system that captures information about requirements, goal, ands conceptual business processes.
• IT management and enterprise architects have detailed functional models, more technical but not a code-level view of system components, hardware, and software deployment details, and shared application components.
• Data architects and DBAs have conceptual data models, physical data models, and the DBMS-specific artifacts necessary to implement the data structure and the services necessary to support the processes being automated or implemented.
• Application/system architects and application developers have their view into the use cases, sequences of events, and objects that have to be implemented to develop the application code and the services necessary to codify the detailed business process.

Where the real challenge comes in is when there's change: change in requirements, change in business model, change in technological capabilities (new hardware or software standards) - or change in understanding the existing requirements. If we can become more effective and more efficient in coordinating the communication of that change between business and IT professionals, we can increase the agility of our IT development efforts, and so increase the agility of our business. As the team moves through the software development lifecycle, the business system model becomes more and more detailed as each phase of design adds more to the information about what's to be built through adding more diagrams and more descriptive text to the "model"...and in some cases, the code to be generated.