Growth in the use of Web services has highlighted their impact on
enterprise application integration, but it would be a mistake to
assume that one technology replaces the other
Recent spin from software suppliers appears to claim that Web
services technology will oust the complex, expensive and
proprietary enterprise application integration (EAI) which
currently dominates the market.
With simple, cheap standards-based Web services integration
platforms, there is certainly major potential in this new
technology. It can be used within companies across all industries
to build new applications which are easier to integrate with
existing ones and to integrate existing applications.
But the idea that Web services technology is set to take over the
software world is complete tosh. And dangerous tosh at that. In the
world of software infrastructure (of which EAI is a part) the crude
association of proprietary technology with "bad" while
standards-based technology is seen as "good" is not a sound idea.
Proprietary technology generally works better, runs faster, and is
more efficient.
Web services technologies prescribe a new way of deploying software
in order to make it available for direct program-to-program access
across the Internet. They also provide an independent mechanism
through which applications can interact. The software that sits
"behind" a Web service could be an enterprise Java application or
component; a Cobol transaction; a Visual Basic program; or a common
object request broker architecture (Corba)-based application - an
attempted technology standard for application server products. Or
it could be something else.
This concept is not new. Microsoft's component object model (Com),
for example, was founded on the same principles - the ability to
build and integrate its own software products, and for third
parties to use it to link their software programs together with
others, on the Windows platform. It is a Windows-only technology,
and continues to play a major part of underpinning .net.
There is also Object Management Group's Corba. However, both Com
and Corba required suppliers to swear allegiance to their
particular camp in the "Microsoft versus everyone else" war. The
difference with Web services - and the exciting bit - is the use of
open standard Internet and Web technologies to expose interfaces.
This makes it easier for different supplier communities to work
together. The war may finally be coming to a close - in this area
at least - and Web services are the peacemakers.
Universal backing for Web services technology from suppliers means
that there are a lot of people interested in using it to ease the
complexity of integrating packaged and custom-built applications.
This problem area has most recently been the domain of EAI products
from suppliers such as webMethods, Vitria, SeeBeyond and IBM.
Web services technology is a potential substitute for proprietary
EAI technologies in some instances, but not in the majority. In
other cases it will play a part in overall integration solutions,
providing just the basic communications infrastructure.
But today, Web service-based middleware technologies are not
capable of driving serious EAI solutions. The uncomfortable truth
for many software infrastructure suppliers is that EAI is
"heavy-lifting" work.
Users need technology stability and maturity when pushing thousands
of messages or function calls through an integration system every
hour. Web services today do not provide for many of the advanced
features required by such scenarios.
Moreover, there is a whole range of EAI scenarios where the focus
of Web services is not really relevant. Ovum calls these scenarios
data migration and reconciliation scenarios, where integration work
is required to homogenise information stored in a number of
applications whose functions and information overlap.
Examples of this would be where two companies merge or where a
company wants to migrate an old system to a new platform.
In these scenarios, integration technology needs to focus on two
issues, which Web services technology is not designed to address.
The first is making it easy for developers to understand the
structure of existing databases and other data sources. The second
is performing highly complex transformations on that data, very
quickly.
The XML-based technologies XSL and XSLT are often pitched as part
of solutions, but they are not powerful enough, and their
implementations are not efficient to support such serious EAI
scenarios.
Web services technology will affect the EAI landscape - but in most
cases it will not remove the need for proprietary EAI technology.
Anyone who tells you otherwise is probably trying to sell you
something that you should not buy.
Neil Ward-Dutton is director of e-infrastructure, Ovum