Page 1 of 4   next page »

Services-oriented development of applications (SODA) is an important development model for enabling organizations to reorient business processes in the transition to a service-oriented architecture (SOA). This article describes one such approach.

Services-Oriented Development of Applications (SODA)
Gartner refers to SOA and SODA as foundational elements of future computing. SODA is a new style of developing software, designed to work specifically within the SOA paradigm. SOA represents a collection of loosely coupled, coarse-grained, heterogeneous components that can be easily snapped together using Web services. The result is enhanced developer productivity, code reuse, and business agility. The SOA Blueprints (www.middlewareresearch.com/soa-blueprints/), an industry-derived set of best practices and associated reference implementation, is an excellent resource for those looking to move to an SOA.

SODA is centered on the creation and assembly of services and service contracts first, deferring the design and implementation of the objects and components that realize the services until after the coarse-grained service contracts have been ironed out. SODA developers focus more on the process flow within and between applications, and less on the code that creates the underlying system. For a brief and to the point description of SODA, see www.serviceoriented.org/soda.html.

Pain Points
The issues described below are certainly not all-encompassing, but are meant to exemplify the issues that often impact a development process (particularly when undertaken by larger enterprises with multiple development teams), and which might be alleviated by transitioning to a SODA-based development approach.

Dependencies / Bottlenecks
Bottlenecks related to the ordering of dependencies are usually the biggest issue. Developers are often waiting for a dependent item to be designed and/or implemented before they can continue with their tasking. One of the issues is the division of labor in a multi-team environment. A standard division of labor often employed in this environment consists of having different teams working on the same vertical slice of functionality in different tiers, causing these dependency-related bottlenecks to spring up.

Management and mitigation of dependency-related issues is crucial to ensuring that developers can continue to be productive throughout development iterations; however, it isn't possible to "manage" a dependency issue away if many of the pieces are being developed in parallel (which is often the case with a fixed deadline).

Creation of the data model for each of an organization's releases (iterations) oftentimes begins roughly in parallel with the development of the software that interacts with it and not much earlier, which is not ideal but not always possible to avoid. Similarly, the domain model for a release may not be completed prior to a development iteration that will be using it. Proper planning in building out the data and domain models first is critical to avoiding these issues but if the project is already well underway, it is what it is ("next time, we'll do it right"). These issues result in a bottleneck for the business services, object/relational persistence mapping, and enterprise application integration (EAI)/business-to-business (B2B) developers, who require both data and domain models to be completed before much of their work can commence. The dependency problems can quickly cascade.

Division of Labor
Development teams are often divided horizontally by tier (especially those that are geographically dispersed). One team may be working on the presentation tier and the other on the business tier. This division often exists out of necessity due to the fact that off-site teams may not have access to the back-end systems with which the business tier needs to interact. Development of the presentation-tier functionality does not require access to these systems (at least during the design and initial implementation phases of the iteration). This division of labor has its problems, though, in that the development of discrete pieces of functionality (use cases) is being performed by separate (again, often geographically separated) teams. These teams often communicate through interfaces to business services and the data transfer objects (DTOs) (http://java.sun.com/blueprints/corej2eepatterns/ Patterns/TransferObject.html) exposed through them - a standard Java 2 Enterprise Edition (J2EE) best practice. The DTO structure and content often changes during a development iteration, as both the presentation and business services developers dig deeper into their designs and subsequent implementations. In addition, the structure/content of the DTOs is (mostly) driven by the structure of the domain model, and as it evolves, so do they. These issues could be lessened somewhat if the same team were working on an end-to-end slice of functionality (one team would be in control of the structure/content of the domain model, DTOs, business service interfaces, and the associated presentation). The key roadblock to going down a road with this approach is the capability to consistently simulate interfaces with the back-end systems, so that an off-site team can be given a full vertical slice. When teams are collocated, such back-end systems often have usage restrictions, especially when the back-end system is connected to and/or owned by another project, department, or business.

How Can a SODA Approach Help?
All resources are considered to be as services with SODA (from UI components on the front end to interaction with an external business partner on the back end). The primary development activity for the vast majority of developers here is the orchestration of those services. A small number of J2EE and integration gurus are responsible for the development of the implementation and/or extension of services in the case where the chosen SODA tools do not provide them off-the-shelf.

Page 1 of 4   next page »