-->
  • October 7, 2005
  • By Marshall Lager, founder and managing principal, Third Idea Consulting; contributor, CRM magazine

Where's the SOA Train Headed?

Services-oriented architecture is the application design trend of the present and the future, but its very nature indicates that it will not be with us forever, according to an essay published today. In "Dateline 2010: A Look Back At SOAs," Bruce Richardson, chief research officer at AMR Research, considers several angles of the SOA issue, including value proposition, performance, necessity of skills, and possible vulnerabilities. He reminds us: "SOA is a journey, not a destination." The enterprise applications market is going through a period of consolidation, which appears to be limiting the options of smaller companies. "A number of companies received funding when CRM and ERP became important issues in business, but now they're stranded with no hope of an IPO, hoping that SAP or Oracle will buy them out," Richardson says. "We got excited about client/server architecture back in 1991, and now I'm wondering about what questions we should have asked ourselves back then, and how to apply those lessons to the emerging SOA." The questions are wide-ranging, from the increased computing infrastructure requirements and the possibility of slow performance, to governance of the spread of new services and the shortage of personnel skilled in creating, integrating, and maintaining SOA components. Intel estimates that the reuse of services and components will improve deployment efficiency by 25 to 56 percent over traditional deployments, with the attendant IT budget savings, but Richardson ponders the other side of the issue. "When we shifted from dumb terminals to PC clients, most of us did so without thinking about how much more it was going to cost to support the new desktops. Instead, we viewed it as a productivity boost. Now, many large companies have engaged in a very expensive, instance consolidation exercise designed to move us back to a more cost-effective centralized model." This, he believes, will lead to hidden costs in the form of hardware upgrades, real-time BPM and business activity monitoring tools, and consulting services fees for integrators. Richardson also notes a lack of security due to multiple versions of existing services. "In the initial [ESA] pilot, there is no way to keep track of the various versions and changes to specific Web services." He suggests creating a metadata repository to define the services catalog and prevent loopholes from forming. He's not entirely pessimistic, however: "I see a lot of hope for analytics and handling unstructured information. The space for knowledge management and content management is getting larger," Richardson says, and that space will be filled by specialist organizations that can offer their expertise as a component in a services platform. "The ERP platform of the future consists of a new hub-and-spoke architecture. The center consists of a much thinner, stripped-down applications hub. The spokes are a set of much richer Web services that radiates out through the enterprise and into the extended trading partner community," Richardson says. "Intel sees ESA as offering a new plug-and-play, ERP-as-a-service platform that will accelerate business changes and software upgrades. It also provides an opportunity for business process reengineering, especially for processes that extend to trading partners." The current issue is how to cut through the clutter of SOA offerings, but "in a me-too environment, the tie goes to the ERP vendor," Richardson says, noting those vendors have control of the aggregated services in the short term. SOA requires interoperability, however, which will open the door for a new crop of best-of-breed vendors, as well as the ability for users to pick and choose from among the best platform features. But the new best-of-breed will have much shorter time to innovation. SOA will change and evolve for the foreseeable future. "When will the future arrive? That's too hard to predict," he says. "Client/server took five to eight years to reach the promises made in 1990 and 1991. With SOAs, the trip will be longer. I think we are embarking on a 10-year journey--all aboard." Related articles: Oracle's Fused Future: Support and Interoperability
SOA Is Consulting's Bread and Butter Integration Still Leaves Some Users Confused
CRM Covers
Free
for qualified subscribers
Subscribe Now Current Issue Past Issues