SOA in Action Blog

Service oriented architecture has been doing a stellar job of addressing issues with siloed applications and automating business processes. However, to date, SOA has not been able to address issues with data, because SOA has been about applications.

That's the take of Madan Sheina, principal analyst within Ovum's Software Applications group, who In a Webinar hosted here at ebizQ and sponsored by Informatica, proposed ways to close the gap between application service orientation and data service orientation. In essence, he proposes that organizations look into a "service oriented architecture" for data.

Sheina calls this approach "process-driven data integration," and describes it this way:

"Just as SOA integrates and orchestrates transactional processes and exposes Web services at the application layer, process-driven data integration does the same thing, but at the data layer," he explained. Such architecture would play a role in "orchestrating sophisticated data integration tasks packaged up as data services or data integration tasks. It applies SOA principals to the data integration process itself."

Madan and co-presenter Ash Parikh proposed that organizations adopt process-driven data integration, a process in which data integration functions - such as data movement, transformation, or auditing - are made available as reusable data services across the enterprise.

Parikh calls the current situation in many organizations -- with point-to-point data access schemes -- a "data integration hairball." He observed that over the last three or four years, "people either build their entire SOA stack either through magazine-driven  approaches, or vendor-driven approaches and accumulate every single kind of SOA technology that's out there. But they are really missing the information piece of it."

He adds:

"From a business-user perspective, we want to innovate faster, we want to do more with less. There's another paradigm out there, which is around basically accumulating more information sources, more systems, whether it's through acquisitions or mergers or consolidations, you're just increasing the inventory of all these information sources. And what happens at that point is you want to keep innovating. You still want to keep diffrerentiating. And what you do is request IT to quickly seamlessly bring in that information source. And the way that IT responds is to actually enable you with point to point integrations, which are really hard to maintain over time. What you're really ignoring at this point is the diversity of data access, the diversity of the various information sources that you're dealing with."

A process-driven data integration approach is essential because SOA and BPM alone cannot address these data issues, they said. While SOA - along with its close sibling, business process management (BPM) - have been very effective in integrating application silos and automating business processes at the application layer, they have been a world apart from enterprise data infrastructures.