Many people in the planning stages for SOA do not get the proper advice and guidance as to how to proceed, or even what a SOA actually is. Thus, the larger tragedy is that many of these projects will hit the wall, and do so with an impact that will reflect poorly on the notion of SOA, when it's not really a SOA issue at all.
First, be wary if consulting organizations point out their experience in the world of SOA by putting up past projects as proof of their experience. Most, if not all, of these past projects are really JBOWS (just a bunch of Web services) and have no underlying mechanisms to provide agility, which is the core benefit of SOA.
The problem is that many of the people looking to hire SOA consultants don't understand the difference between JBOWS and a true SOA, and accept JBOWS as "experience." In reality, it's an indication that the consultants don't understand the core value of SOA, and thus could send you off in all sorts of dangerous and costly directions. So, make sure to hire consultants who understand that SOA is really about configuration, agility, and changeability, and not just about service enablement. It's very easy to expose services; turning those services into solutions is another level of sophistication.
Second, many consultants are a bit too chummy with vendors. Thus, you'll find that they implement the same vendors and technology each and every time. This should be a huge red flag since SOA problem domains are all very different, and technology solutions that work best for the solution are never, ever, from the same vendor, over and over again. However, when you sell hammers, everything looks like a nail. The path of least resistance is what you know, not what you should do.
Those tasked with managing SOA consultants need to state clearly that you are looking for the right solution, not the one they know, or what may be part of an existing partnership. Indeed, consulting organizations with many preexisting vendor partnerships should be quickly overlooked. You need guys and gals in there who are agnostic when it comes to technology, and are willing to leverage the best-of-breed to address your issues.
Third, there needs to be a predefined process. While many SOA consultants try to use older SDLC and enterprise architecture processes, SOAs need a specific approach that addresses the unique nature of these architectural patterns. For instance, there is a traditional focus on data, but the data needs to be properly bound to services. Moreover, those services must be properly bound to processes. Plus you need to keep track of all the interdependencies, and how all of this stuff links to SOA governance, SOA security, and event management and monitoring.
Many consultants attempt to oversimplify the process, rapidly moving through or even foregoing the planning steps. Their main focus is the selection of the technology, or, in some cases, they attempt to force fit a problem with a predetermined technology solution. This can never be good.
The fact of the matter is that SOA is a complex distributed system, and thus complex to plan, design, build, and test. The time spent in planning will later pay huge dividends. There should be a very rigorous process/methodology defined that, at a minimum, provides you with a semantic-level, service-level, and a process-level understanding of the problem domain, not to mention the governance model and security strategies. Trust me, you can get SOA right the first time, but with more planning and sweat than you expect.
Finally, there is the lack of understanding about ongoing SOA operations and links with traditional enterprise architecture. You just can't build a SOA; you have to live with your SOA, ongoing. That means understanding how the solution exists currently, and how to align it with business as it changes. If your consultant has done his or her job, you should have an architecture where the volatility has been abstracted into a configurable domain. Thus, it's a matter of changing things at the configuration layer to adjust to the changing nature of the business. Therein is the value of SOA.
If your architecture does not seem to be providing that degree of agility, then you're going to have to loop back to the beginning to see where you went wrong. Typically, per my previous point, not enough planning was done and perhaps the wrong technology solution was employed. In any event, you need to bite the bullet, spend the money and fix it, or you'll find that your SOA actually takes you backward.
The issues here are not a case of bad intentions. I think everyone honestly wants to do their best. It's a lack of understanding and perhaps talent in some instances. Those who do best with SOA have a wide range of skills, a good understanding of architecture and the value of SOA, and all of the good work that needs to occur to make it work for the client. However, I don't see many out there who fit that bill, and the amount of bad advice is becoming a huge issue. Unfortunately, many won't figure this out until it's too late. SOA is something you do, not something you buy, but you also need to do it right.
David S. Linthicum (Dave) is a managing partner with Zapthink, LLC a consulting and advisory organization dedicated to excellence in SOA planning, implementation, training, mentoring, and strategy. Dave is an internationally known application integration and SOA expert. In his career Dave has assisted in the formation of many of the ideas behind modern distributed computing including Enterprise Application Integration, B2B Application Integration, and SOA, approaches and technologies in wide use today. Dave keynotes at most major SOA and Enterprise Architecture conferences, maintains one of the most read SOA blogs, is the host of the weekly SOA Report Podcast, and is the author of 10 books, 3 on integration and SOA topics. You can reach Dave at firstname.lastname@example.org.
Timely, incisive articles delivered directly to your inbox.