If you've been following service-oriented architecture for any amount of time, you know that data integration is a problem. How do you apply SOA to data technologies?
Finally -- we have an answer, from Jeff Pollack, a veteran technologist and senior director of Fusion Middleware at Oracle.
Pollack's answer: You don't. Forget about it.
SOA is great for a lot of things, but it's not the best fit for ETL (extract, transform, load); master data management; business intelligence and data integration, Pollack writes in "The Case for Enterprise Data Services in SOA," published recently on ebizQ.
Of course, he's not the only or even the first one to notice that there was a problem. But he's certainly gone over the idea as firmly and completely as anyone to date.
He particularly takes aim at XML and the perception that it's a panacea for data integration. XML and Java-based containers are the current market definitions of data services, but neither are robust enough to handle data transformation on an enterprise scale, argues Pollack.
"The simple and unfortunate reality is that enterprise data requirements are hard, and the dreams of SOA only solution for all enterprise data are likely to remain dreams."
That's the "bad" news.
The good news is that we already have data services and we did even before SOA was a glint in the milkman's eye, according to Pollack. SOA could be used as a control point - not a replacement - for data infrastructures. In this scenario, SOA data services refers to "decoupled end-points that expose highly optimized components for working on all types of data."
Pollack reminds companies that all of this enterprise infrastructure is ultimately about handling the massive amounts of data. Given that fact, why not spend some time evaluating and planning your data services infrastructure? He spends a substantial part of this seven-page article detailing what an architect should include when planning a data services infrastructure.
IT executives and managers will be more interested in how he sees the two coming together for the business:
"A modern Data Service architecture should support synchronizing data grids with master data, publishing high quality canonical data within a messaging infrastructure, and expose control points for commanding BI and data warehouse systems as loosely-coupled services. This vision is not so much a dream, as it is a requirement for modern information-centric businesses that hope to use IT as a competitive edge within their industries. Yet regardless of how grand the IT strategy might be, a good Data Services plan will first solve fundamental tactical issues that simplify the use of data throughout enterprise architectures."