Demystifying Data Federation for SOA

Functional Requirements for Data Federation

The basis for just about all service-oriented architectures is reusability. A high level of reuse can be achieved through a layer of data sources and business logic that are exposed as services accessible through an implementation independent service contract. A fundamental design goal of SOA is to provide a logical abstraction to these capabilities and prevent exposure of the physical topology of underlying sources to consuming applications. This is also the design premise of data federation and the basis for its functional requirements (Figure 3). In addition, a data federation solution should contain:

Figure 3

•	Data Source Abstraction Layer for your SOA – Map multiple physical data sources into a set of services defined by logical entities rather than physical location. Reuse data services to define new abstractions as required.
•	Federated, optimized Queries – Queries produce execution plans that use the specific capabilities of underlying databases and apply optimizations of distributed operations and nested data.
•	CRUD-style Data Updates – Data services support creating, reading, updating and deleting of data in place at the original sources. Updates may be distributed and should support some type of reliability or XA compliance on the update cycle.
•	Rich Hierarchical Model – Needs to be able to model relational data implementations as well as interchange with payloads from SOAP or REST services to be processed without format conversion (Figure 4).
•	Security – Security poses new challenges for a data services environment: Services can be reused in unpredictable patterns, requiring flexibility while maintaining control of sensitive information. Data services require rich access control functionality, from policy-based authorization to fine-grained row and column-based security, identity propagation or credential mapping.

Figure 4

A data services federation layer is often seen as a way to take the first step in SOA. This services layer provides data mediation or abstraction between different data consumers and heterogeneous sources. Data services can be virtualized, aggregated views built from sources across the enterprise. They simplify data access and once created, are highly reusable. This approach eliminates the need to build workflows or code Java by hand, making it possible to automate data service creation and maintenance. Other consumers of data services include business processes, business intelligence applications, master data management (MDM), portals, and Web 2.0 applications.