ESB: The lighter alternative for integration

Traditionally, organisations looking to integrate applications and software have tended to look to Enterprise Application Integration (EAI) platforms from players like Tibco or Vitria.

But while still the gold standard for heavy-duty integration, the platform approach -- often also called 'hub and spoke' -- is a non-trivial task in terms of both initial cost and consultancy. A new alternative gaining traction among analysts is a Java-based alternative, Enterprise Service Bus (ESB).

"ESB is good because the system is inherently networked and distributed," notes IDC's director of European software research, Rob Hailstone. He likes ESB because of it being standards-based, but also because it's extensible.

ESB has come out of two sources: messaging and Java. The idea is that applications are connected to a more lightweight backbone than the traditional EAI platform, which then takes care of functions such as intelligent routing and transformation, connecting and coordinating the interaction of services across highly distributed contexts.

Messaging may be an incomplete way of appreciating what ESB is all about, however. "ESB is a really new sort of message-broker based architecture, taking on a further stage architectures like the Common Object Request Broker Architecture of the 1990s," thinks Christophe Toulemonde, integration programme director for Meta Group in Paris.

No matter what their intellectual provenance, there's no denying ESB is a hot topic for the integration market. In a recent research note (December 2003) Gartner went so far as to predict that ESBs will replace traditional communication middleware in new applications by 2007.

Integration observers think this will be because people can't really afford to wait around to do everything EAI-fashion, and also can use this as an entry point into the coming SOA (service-oriented architecture) future. "Traditional EAI hubs make sense if you're taking on a big project which can cost justify linking a large number of systems," says IDC's Hailstone. "But in reality most integration is about linking more like two. ESB can help in making that move more cost effective. And in terms of product maturity we're up to round two now, with some reasonably mature products. So the good news is that in the end, service-oriented architecture will make it much more possible to do 'mix and match' integration projects."

ESB fans claim this is a both a lower-cost and less complex alternative, which though at the moment probably best suited for tactical integration, usually around linking existing systems with the Web, is set for explosive growth on the back of rising deployment of Web services.

Independent Enterprise Service Bus (ESB) vendors include Sonic Software, Spiritsoft, Blue Titan and Cape Clear. But they're not going to be the only games in town for much longer. IBM is expected to launch some kind of an ESB product under the Websphere brand later this year, while established EAI players like SeeBeyond have started to extend their offerings in this direction too.

Spiritsoft, a UK-US ESB start-up, numbers Etrade, Barclaycard and Dutch financial institution Rabbo Bank among its customers. Its UK-based chief technology officer Rob Davies explains more. "There's growing interest in the integration stack coming out of J2EE, but in many ways ESB is a true successor to EAI, but this time we're putting much more of an open standards-based face on integration."

Spiritsoft, which grew out of a big middleware project at the City offices of Japanese investment bank Nomura, is set to launch a new product in April tied to the new Apache Geronimo J2EE-based Web server and connector system. Meanwhile its main rival, Sonic Software, a subsidiary of Progress Software, claims to be doing around $15m worth of straight ESB business already per year, up from $2.5m in its first year of 2001, according to its VP and chief technology evangelist Dave Chappell.

"We spotted back at the time that JMS was first launched that the best way to do the kind of highly distributed integration customers wanted to do was through this new form of message-oriented middleware," he says. "What's great about ESB as compared to the old hub and spoke is it's less monolithic and easier to span boundaries in both companies and beyond, potentially. The ESB combination of Web Services, enterprise standard messaging and intelligent routing means this is something all new projects that involve linking together applications should look at. ESB in essence is changing the economies of scale of integration itself. ESB really provides the architecture Web services can live in."

All this talk of Java Messaging Service and J2EE shouldn't obscure the fact that Java itself is set to be more integration friendly. It's very early stages, but (yet another) new Java standard being worked over right now is called JSR 208, or Java Business Integration, which seems promising (see www.jcp.org to learn more).

Neither should the fact we've mentioned Java so often let us leave Microsoft out of the equation: look for announcements later this year of its so-called Indigo ESB-like technology, slated to be an interesting factor in its upcoming Longhorn operating system technology.

ESB seems definitely worth investigating then, especially if you are already a J2EE-oriented environment, with the proviso that Microsoft will soon have something to say here. On that same note, bear in mind that until the big boys produce their own ESB versions of their integration offerings you may be forced to bake your own. Gartner advises: "Leading-edge enterprises implementing a SOA-capable enterprise nervous system in 2004 should consider ESB vendors that are smaller than IBM and Microsoft, or create ESB-like infrastructure by using separate MOM and Web services products, integrating them through custom code."

So why not ESB uber alles? The short answer is that at the moment at least it seems to be being positioned as primarily a way to link new applications, more usually Web-facing ones. That's fine for that class of applications of course, but it may not be of too much immediate help to those of us struggling to integrate big back office systems, let alone legacy or bespoke software.