Les DSI rencontrent bien des soucis avec la SOA. Mais le principal problème est que, avant de se consacrer au "comment", il faudrait déjà savoir ce qu'est véritablement la SOA au delà des discours marketing qui déclinent le concept jusqu'à l'overdose.

The benefits of a SOA are widely touted by advocates, and easily understood by anyone in the C-suite cost savings, risk reduction, and business agility through high reuse, interoperability by design, and loosely-coupled integration.

But while practitioners agree on the benefits, the methods for attaining them are evolving and conflicted. Spend any time perusing the SOAsphere, and it's clear that few areas of IT are as politically charged. Pervasive misinformation, coupled with religious drama, are the real reasons why many SOA initiatives stall or fail (Google SOA failures for some interesting reading).

At the root is a vendor feeding frenzy the likes of which we haven't seen since Sun was the dot in dot-com. IT investment in SOA will hit $52 billion over the next four years, with 77 percent of IT shops doing the SOA dance, according to AMR Research, Boston. That's up from 53 percent of companies with over 500 employees investing in SOA now, AMR says.

Present-tense surveys and future projections from other analyst orgs paint similarly dramatic pictures. Assuming such numbers are accurate, that's one big honkin' slab of red meat dangling over a pool of hungry piranhas. It's no wonder there are, literally, countless vendors touting their brand of SOA over others.

The good news choice for CIOs, including all the usual EAI, BPM, ERP, AppDev, modeling, and repository suspects (IBM, HP, Oracle, Software AG, Sun, BEA, Microsoft, IDS Scheer, SAP, more) present and accounted for. Add a bumper crop of lesser known but no less savvy software and service providers, appliance makers, consultants, and open sourcers, and you have a veritable SOA supermarket.

The bad news choice for CIOs. Vendors are settling into different camps with different agendas, different technological approaches, and most important, differing views of what constitutes a SOA, and why SOAs fail. Several ecosystems have emerged (Oslo rising next), which might or might not play well with each other (my money's on "not"). Is SOA doomed to suffer the same fragmented fate as other like-minded approaches of the past?

The industry can't even agree on the basics. It is a "soah" or an "s-o-a?" Consider the term "service," which is horribly overloaded. There are Web services, Software as a Service, Service Oriented Architecture... and these are just the hot buttons. How many times has someone confused SaaS with SOA? How many times have you confused SaaS with SOA? The two can be and often are mutually exclusive.

Many on-demand and Web 2.0 apps are, in truth, classic information silos with nary a SOA as far as the eye can see. Yet the two service-oriented approaches are often associated conceptually in conversation, but rarely in implementation, much to the detriment of Web 2.0. Likewise Web services, which, contrary to popular sentiment, are not a prerequisite for a SOA (donning asbestos suit, powering up shields, writing last will and testament).

Make no mistake about it. CIOs are confused and concerned.

So what is a SOA? Understand "what" a SOA is, and getting to "why" and "how" gets easier. Next week, I'll take a crack at answering that enigmatic question, and define "what" is wrong with SOA in the process.

Rich Levin

. Web-profils.com, la place de marché du conseil et de l'ingénierie