Tuesday, October 4, 2011

Mission Impossible? Or how to achieve the SOA vision.

When I am asked about the state of SOA, I sometimes comment that anything involving architectural change is bound to take a little time. But my more considered response would be that whilst the impression of SOA is now widespread, true implementation of the SOA vision, for most enterprises remains a distant vision, if indeed they still remember what that was.

For me the vision was encapsulated in the report by one of our customers on their SOA progress in 2009. They reported their systems were exploding in size and complexity. They had scant standardization, and there was no single truth. If a core process broke they would change it to fit the application, rather than the other way round. This was crazily expensive to maintain. After four years of transformation they report a 20% reduction in IT staff, 1500 systems closed down, the ability to turn services on automatically for customers virtually as they place their orders and a massive reduction in complexity demonstrated by a rental price change that previously required changes to 42 systems – followed by three months of testing, now requires just one platform adjustment that automates the change process. THAT’S STRATGIC!

In contrast I read a Forrester survey[1] from last year that reported while 47.4% of respondents work in organizations where SOA projects are underway, the original reasons for SOA, reuse and cost reduction, have morphed into data integration, legacy integration, flexibility of application development, and department-level application integration. Perhaps this is why we at Everware-CBDI are observing numerous inquiries about “SOA Reboot”, which is variously explained as interest in doing SOA properly, realizing the vision and or delivering real business benefit.

For many enterprises the root cause of this lack of achievement is very straightforward - SOA requires a strategic initiative that looks longer term than most enterprises are able to do. But for most enterprises this is mission impossible, they are bound by short term goals and budgets.

The solution is not rocket science. What’s needed is a governance system that manages a progression from tactical to strategic. Many SOA efforts today are business process project focused, because simply put that’s where the business priority is today. What’s needed is a governance system that ensures project service solutions can be evolved to become enterprise services, where it makes sense. The overhead in making this leap is that a few new policies are needed that spell out better working practices. Consider some candidate policies.

  • All new components and services MUST comply with a defined minimum level of reference architecture.
  • Implications and strategy for future service reuse is a REQUIRED element of all Plan or Feasibility phase end reports.
  • All projects MUST reuse and evolve existing (loose coupled) services and components before acquiring or building new components

There’s more; to make this work needs good governance plus a product (sic) management system in place, because it will get complex. But it works.

I am writing this practice up for the Quarter 4 CBDI Journal. Make sure you are a registered subscriber so you get a copy on publication.


[1] TechTarget/Forrester Research State of SOA Survey for 2010

http://media.techtarget.com/searchSOA/downloads/TTAG-State-of-SOA-2010-execSummary-working-523%5B1%5D.pdf

1 comment:

  1. Mea culpa - the practice research I was writing (see above) didn't make the cut line for the October Journal. If you would like to get an early view, contact me. I expect this will publish in the next CBDI Journal. David

    ReplyDelete