OK, start SOA small if you want - but don't skimp

Suddenly everyone is saying that SOA should start small - work incrementally, get a few small wins under the belt, benefits grow over time. Joe McKendrick summarized it well, with vendors such as BEA, Intel and HP jumping in. In a sense there is nothing new here. Incremental approaches to integration have been the way forward for a number of years now. Admittedly, with all the hype around SOA and its ability to 'transform the business', some companies have got seduced into trying for a full-scale top-down strategy, but generally companies start small because the complexity stops them from doing anything else.

However, there is an important warning here. Be careful the SOA initiative is not strangled at birth. I have written before about the need for an SOA ecosystem or infrastructure, and indeed there is a (free) paper on the subject available from Lustratus. The problem is that if you start a small SOA project in the belief that it is OK to skimp on the ecosystem until more projects are implemented, you are heading for trouble. For example, suppose you decide to start with just web services for the first few projects. Then, the services created will be built without a registry in which to record them - so others can't find them and reuse them, because they don't know they are there. And once you DO decide to get a registry, will all the existing services be recorded in there, together with all the detials and information that will prove valuable in the future?

Start small if you want - that's fine. In fact, it's a good idea. Just make sure the basic SOA ecosystem is in place before you start. To misquote an old proverb, start in haste, repent at leisure.

Steve   

Comments

[Jack van Hoof]

Ofcource it would be beautiful if you could get all the right things in place to succeed with SOA. But in many (if not most) cases you just can't get all the right things in place. That's not our world and will never be our world. We will have to deal with chaos. Also when in comes to SOA approaches:
http://soa-eda.blogspot.com/2007/06/there-is-no-one-size-fits-all-approach.html

[Steve Craggs]

Good point Jack,

I agree that all situations will be different. However, my point is that if you don't at least make some attempt to get some of the enabling tools in place, you are doomed to failure in the longer term IF you are expecting SOA to deliver its much-vaunted benefits at some point.

Steve