Call 2006 the year of consolidation for SOA (service-oriented architecture). Big companies snapped up SOA pure plays one after another like kids scarfing up candy. At the same time, leaders of enterprise SOA initiatives reached an amazing consensus -- that the hard part of SOA isn’t the technology, it’s dealing with the reorganization and governance required to make service orientation work.
The mergers-and-acquisitions mania began with a bang in January, when Mercury Interactive snagged the leading independent registry/repository vendor, Systinet. Less than three weeks later, Progress Software (Sonic’s parent company) announced its intent to purchase Actional, a services management software provider. And, of course, mid-year, Mercury itself received a $4.5 billion offer from HP in a deal that finally closed in November.
Two other vendors of registry/repositories (which combine service directories with metadata management) were snapped up: Flashline by BEA in August and Infravio by webMethods in September. In other summer fun, the largest of the independents, SOA Software, bagged Blue Titan, a highly regarded service network play.
The wild shopping spree, really an acceleration of a two-year trend, is a tad ominous. Big vendors can now say they offer a “complete” SOA solution: service development platform, registry/repository, services manager, ESB (enterprise service bus), service orchestrator, and maybe a little BPM (business process management) on top. Any proprietary linkages among those pieces foretell classic lock-in: Everywhere SOA extends across an organization, so goes a single vendor’s software.
BEA, IBM, Oracle, and Sun can plausibly suggest that notion already, with Microsoft, Progress Software, SAP, SOA Software, Tibco, and webMethods offering various subsets of the complete SOA stack. IBM has even gotten into the game of licensing prebuilt services with this year’s acquisitions of MRO Software and Webify, plus a buildout of services development centers in China and India. HP is whistling a different tune, emphasizing SOA governance (Mercury) and service management (Talking Blocks) while avoiding repeats of its ill-fated past ventures into service development and deployment.
Anecdotally, at least, most customers don’t seem to be buying the single-vendor proposition. Many understand that to identify, say, one company’s ESB as the “backbone” of an SOA flies in the face of the radically simple SOA concept: an array of interoperable, XML-based services strewn across an ordinary IP network, ready to be consumed by applications. The orchestration and Java-based messaging capabilities of an ESB may be beneficial for some subset of related services, but SOAs are ever-expanding, and extending the same ESB everywhere would incur an absurd amount of overhead.
The new focus on governance is a healthy trend. Keeping service end points robust, after all, is what governance is all about. And the fact that many organizations are already thinking hard about service metadata -- how to use it to monitor, manage, and refine an SOA -- is a good indication that things have moved beyond the beginning phases.