I somewhat agree with Anne Thomas Manes SOA is Dead, but I think it is more appropriate to claim: "Big SOA is Dead; Little SOA is Thriving".
What do I mean by this?
I think BPM, BAM, and ESB initiatives (i.e. Big SOA) are suffering from being large, complex, and somewhat bloated efforts to gaining SOA momentum. Customers want rapid results. They want quick services, quick ROI, they want [sic] SOA Today. That translates to lightweight SOA that is simple and digestible without having to buy or learn an entire software suite or vendor tool platform just to get your first 50 services out the door. This is Little SOA. Tools that can rapidly create services in minutes, rather than months.
Now before I get Savvion, Lombardi, and other BPM player lining up to slice me to virtual pieces, I will precursor by saying BPM, BAM, and ESB are still necessary in an SOA Architecture...just not necessary to start and progress your first chapter of your SOA Journey. Take any of the 1,000 SOA maturity models out there, and the first step is to create services, right? So, why invest in Big SOA to start? The capabilities of these Big SOA tools like BPM, BAM, and ESB can be delayed for 6-9 months, possibly even a good year until your business requires these efforts. How does that sound to your 2009 budget! Finally, some good news! Application-to-Application and Composite Application consumption of business services can thrive without these Big SOA tools; however, there is a caveat that Business Process Automation (BPM, Workflow, BAM) will have to wait-- sorry, Little SOA isn't perfect!
As you can see, my proclomation favors a bottoms-up or middle-out strategy in developing services, as opposed to Top-Down and modeling all your business processes before defining your servcies. If you had asked me 6 months ago, I would have told you Bottoms up is bad and that all your business processes need to be defined before you start mapping activities to services. And, Top-Down is certainly a viable academic approach. It's just not practical. Especially in today's economic climate. Think about how time consuming BPM is, and most CIO's will not buy-into this 1-2 year Top-Down BPM investment. Most CIO's only last a couple of years at each job stop , so they don't want to do BPM on their dime. How can you blame them when it takes an large time and monetary investment to model As-Is and To-Be processes, not to mention standaridzation, education, and maintenance? Although Big SOA maybe the more academically powerful approach, my customers just don't have the patience for Big SOA. They want [sic] SOA Today!
To summarize, buying an SOA software stack for $500kk-$1M+ is not starting small! Organizations can not afford these Big SOA products anyhow in 2009. Gartner and IDC will certainly agree that you don't start with an ESB in your SOA Journey, so you can buy your SOA stack when you are more knee deep into your SOA project and the capabilities are required. (Hey, I'm just echoing analyst sentiment!) However, you do need something to get your services out the door and for the beginners to get off the SOA sidelines. Think long and hard by making investments into Little SOA products to start, and then you can realize the value of being simple, rapid, agile, and most importantly AFFORDABLE.