What do SOA and enterprise data warehouse have in common? As it turns out, lots. They serve the same purposes, require the same type of governance structure, deliver the same kind of ROI. Most importantly, they need each other.
Consider the similarities: Both enterprise data warehouses and SOA battle the silos. Both are about getting the right information at the right time. Both emphasize "reuse" as a primary value -- SOA reuses application components that get written and validated once and deployed many times; EDW reuses data that has been written and validated once and gets deployed many times. Both require a strong, enterprise-focused governance structure that involves the business and builds support. Both need metadata repositories to succeed on an enterprise scale. Both may have negligible ROI the first time around; but economies of scale grow exponentially as assets are shared or reused.
Teradata has been talking about SOA as an enabler for new applications getting deployed against its warehouse for a couple of years now, and IBM also made the connection earlier in the year when it fused the concepts of enterprise information management and SOA. To recap some observations I made in Webservices.Org at the time: SOA needs enterprise information management. Enterprise information management needs SOA. SOA will bring EIM alive across the enterprise, and EIM will be the killer application for SOA. Enterprises need to cut through the silos and be able to cost-effectively publish data from any application, running on any system, regardless of original data format.
Through Web services, end-users can direct queries and access analytical applications. Such components can provide a service interface that centralizes the computation and sifting of data, versus traditional SQL commands. A Web service may access the right data sources and do the analysis to provide a customer’s profitability history, or calculate the probability of losing the customer, or even seek out abnormalities to detect fraud.
Interesting point of view. I would argue that the best way to bring these together is by building decisioning services that use that enterprise data to deliver decisions that can be used by many services. Check out the blog for more on this.