SOA Lite
In some of my interactions with clients, discussions around SOA are met with diverse outcomes. They vary from
- We have decided to have an enterprise-wide SOA. Can you help us build a strategy?
- Help us get started by putting foundational components and work on a portfolio
- We are not ready for SOA yet
- Why do we need an ESB, UDDI, repository, registry? Suite vendors are asking us to buy stacks!
While 1, 2 and 3 are acceptable outcomes, the aggressive move by suite vendors to sell all encompassing solutions has bewildered and completely confused users who were almost ready to buy their products in an incremental way. The absolute necessity of an ESB in an SOA is almost as necessary as a canon on commercial airlines. It is the cart before the horse syndrome. In one of Jason Bloomberg’s articles on Who’s killing SOA, he points out that the platform vendors have a major role to play.
I can easily identify myself with clients who need to show real business outcomes in 2 – 3 quarters, if not less, whereas in the buy my whole suite scenario, the environment setup alone could take a few quarters and a full service SOA can take a couple of years (btw standards are still happening).
The progression from asset discovery to service modeling, which is the core of an SOA implementation is hardly mentioned. Neither is the need for change management with respect to organizational roles alignment and funding models which are foundational problems for SOA.
Many organizations which have been reasonably successful in their SOA path already are those that have mentored their teams into building applications from loosely coupled services.
