Much of the feedback we get to the postings on this blogsite contains a healthy amount of skepticism about the whole notion of "SOA." Many feel it's the latest empty container of hype to be unleashed by vendors. Some call it "DOA." (zing!)
For the record, I agree that the concept is being oversold, and I'm sure many vendors will have moved onto something else in a couple of years. But, architecturally, its part of an unstoppable evolution toward -- to borrow from Sun -- the network as computer.
Here are some thoughts from a recent column I'd like to share with you. This originally ran in the Data Center section of Database Trends & Applications. (Available in the print version, not on the Website.)
SOA certainly is overheating the hype meter, and promises to be the overused buzzword of 2005. But a lot of misconceptions swirl around SOA, which is supposed to mean service-oriented architecture, but may more aptly be called State of Anarchy.
SOA advocates which now seem to include every vendor with a piece of code to sell say the new approach can break complex and incompatible applications and processes into granular, standardized components that can be assembled and dissembled as business needs dictate. The Yankee Group conducted a survey and found 75% of organizations have plans to begin building an SOA within the next year.
However, many Web services projects the building blocks of todays SOAs are just getting off the ground, and have yet to touch core mission-critical systems found in enterprise data centers. Rather, they are more likely to be associated with Websites and other edge-of-the-enterprise projects. There seems to be an awful lot of confusion out there about the difference between Web services and SOA. You can have a thousand Web services, but that will not give you an SOA. SOA cant be built in one swoop rather, it will be a phased initiative that will be built out over a period of years.
Another misconception is that SOA is a new approach. In fact, theres nothing new about SOA, other than the fact that its operating at a higher level in the technology stack. CORBA (Component Object Request Broker Architecture), unleashed on the world in the 1990s, was designed as a modular, platform-neutral approach to IT infrastructures. Microsofts DCOM (Distributed Component Object Model) was intended to provide such modularity for the Windows-centric world. So SOA isnt just Web services standards. SOAs will most likely be built using Web services, but many sites will also be incorporating CORBA or DCOM as well. The difference is that CORBA and DCOM addressed application objects, while Web services/SOA address services and processes.
SOA is a grand idea, and vendors should be applauded for any efforts they make to achieve interoperability between platforms and disparate applications, and stovepiped data silos. But many vendors are also pitching SOA solutions in a box, and it just wont happen that way. SOA will take time.