sstein's picture

HazardSince the middle of last year the SOA hype seems to be gone. Many analysts and opinion leaders have changed their own messages now focusing on broken SOA initiatives or slow adoption rates. So far, I refused to join this SOA bashing, because it is too easy to criticise something. However, now seems to be the time to once again stress an important point: technical SOA is dead, but we have not even tried business-driven SOA... <!--more-->

Let's take a look at how SOA customer projects done by IDS Scheer usually proceed:

  1. Customer asks IDS Scheer about SOA offerings.
  2. IDS Scheer consultant motivates that SOA must be driven by business.
  3. Customer agrees, but sends a request for proposal listing almost all web service technologies known in the universe.
  4. A successful proof of concept is done, transforming a low level (business/workflow) process into a technical web service orchestration (e.g. BPEL).
  5. Customer is disillusioned by the overall complexity and low maturity of web service technology.

A typical outcome of such a SOA project is that the customer stops the SOA implementation and instead focuses on describing his overall IT architecture. Services are defined, capabilities of such services are described, responsibilities for the services are defined, and possible integration scenarios are evaluated. One important task is defining which services to not focus on. Just imagine having a service for every basic IT functionality like sending an email. If services are defined on such a detailed level, much effort is needed to maintain the huge number of service definitions and to compose business processes based on them.

Personally, I was never convinced that encapsulating every IT functionality as a web service will lead to any benefits. I call such an approach a technical SOA. Instead, services must be defined on a much more abstract level so that they fulfil relevant business needs. Here, services do not have to be implemented by technology at all. Instead, it is the aim to define reusable business components, which can be flexibly combined to support business processes. At the moment, some early adopters and consultants are practising this business-driven SOA approach, but it is too early to say if this approach will work out as expected. Therefore, I say that technical SOA is dead, but that we have not even tried business-driven SOA.

By the way, in a recent paper we have published many details about how the ARIS modelling method was extended to support business-driven SOA. The paper is published in a scientific journal and therefore a tough read. If you are not interested in all the heavy details but more in a practical example, see the belonging ARIS expert paper.

Tags: soa