The reason consultants are brought in to do integration work
is because it’s extremely messy, extremely difficult, and extremely thankless. “Integration” usually means preserving
legacy applications that are cumbersome and often inexplicably sacred to your client,
squeezing miracles out of the none-too-clean data they yield, and trying to
bring up faster, cleaner apps from an enterprise resource planning (ERP) source
you’re learning on the fly. You didn’t design it, but yours is the head that
will roll if the whole thing doesn’t absolutely sing.

The word middleware
covers a great deal of territory. In general, it defines a class of integration
software that governs communications between components in an application
system. Specifically, it’s used to monitor and route intercomponent
requests and manage the app system’s runtime environment. It can be found in
any number of places, and for each form of intrasystem
request, there’s a specific class of middleware. Here are the most common
varieties you’ll encounter:

  • Distributed
    This middleware handles interfaces in the
    application system, coordinating requests between components. It basically
    enables app components to talk to each other, imparting fine control over both
    event handling in the application and the dynamic disposition of data.
  • Database
    —This middleware sits
    between applications and data, and sorts out what data goes where and under
    what conditions; it’s not just about databases, but can also include
    conventional file I/O. It handles request management and can oversee database
  • Remote procedure calls—This old friend is
    one you’ve used many times before, though perhaps not in this context. RPCs are the mainstay for initiating application events
    across a network. The concept is to run programs in an application system
    transparently, giving the system a level of process control that requires no
    user intervention. While they’ve been around a lot longer than Enterprise Application
    Integration has, their usefulness here is obvious.

Middleware is intended to save you bargeloads
of time you’d otherwise waste in rewriting legacy apps and reconfiguring
databases. But more than a few IT professionals shout curses rather than
sing praises when middleware needs to be put to work. Why is this?

The hard part

The point of application integration is to preserve your
investment in existing applications. The reason integration is required is that
you no longer operate in a world where users are people sitting in front of
terminals. Users, from the legacy app’s standpoint, can now include those
original users plus downline enterprise apps, lateral
apps producing similar but enhanced data, Internet users, Internet apps, and
many other destinations.

On top of this, a legacy app’s input can now include data
that is similar to what it was designed for (multiple databases, real-time
queries, remote real-time queries), and worst, these inputs can become

In short, when you wrap middleware around legacy apps, you
must deal with much more than we typically have to handle in writing an app in
the first place. You might easily ask yourself: Why not just rewrite the legacy

What lies beneath

The power of application integration seems almost
serendipitous, and we tend to get tangled in the technical side and overlook
the incredible benefit that accommodation of legacy apps has bestowed upon us.
Because industry efforts to accommodate legacy system investment influenced
enterprise development early on, a typical application integration framework
operates on the principle of stateless messaging.

Applications no longer necessarily operate in linear series.
Instead, apps generate data, which becomes generally available to other lateral
or downline apps. Via messaging, those apps may then
make use of it as needed. Bodies of data become free agents, in a sense, and
new mechanisms for the initiation and regulation of app execution are now in
place, giving us tremendous discretionary powers over processes as a whole and
even redefining the role of the individual application in a business process.

All of this now rests in your hands. Most books and articles
will coach you on the techniques by which you can implement data queuing,
messaging, and application execution security, but there are broader
implications to these abilities. You can do more than streamline processes with
this level of control. Not only can you greatly enhance effectiveness, but you
can have unprecedented control over departmental functionality and
interdepartmental cooperation. You can also extend to interactions with partner
companies, as well.

Data portability increases dramatically via middleware

In most legacy application systems—and in many contemporary
distributed systems—transactional data exists as an interim step in a linear
process. While data-queuing and interapplication
messaging were originally created to bring legacy app data together with
enterprise app data, the truth is that middleware “liberates” transactional
data regardless of source and destination. Transactional data now becomes
available to all.

Opening databases that were originally department-specific
is obviously a key step in ERP in that it more deeply integrates the awareness
of your client’s various users. It’s a mistake to overlook the possible
advantages of making transactional data available across functional boundaries
because live, momentary, real-time data can hold critical process cues that may
never have been evident before. It’s important to understand that these new
free-standing pools of fresh information, made available beyond the legacy
system’s original destination, can often be useful in ways you never thought

Passively monitor transactions and implement a “super app”

Component transaction monitoring (CTM) has been with us for
some time. Microsoft would love to claim it, but IBM was really the first
player at the table. Transaction monitoring allows you (via real-time
middleware) to keep an eye on a program, a file, a database table, or any other
system component. Then, depending on the sophistication of the middleware
you’re using, you can set up conditional responses to changes in that component—the
sending of e-mails, the setting of flags, the storage of messages in your app
integration messaging system, the opening of permissions for program
executions, etc. Basically, you can now write a “super app” to surround
individual apps and even impart a new layer of control and sophistication on
an existing system, increasing its efficiency, availability, robustness, or all

Programs communicate

By combining transaction monitoring of a legacy app with the
messaging utility of middleware, you effectively give the app the power to
communicate with other apps. Sensing an instance of a program’s use,
identifying the user, and noting the data going in and out basically give you
the power to fully report on the app’s activity to the middleware message pool.
In this way, programs can “talk” to each other, and no distinction is needed
between legacy apps and full-blown enterprise apps—they can now be on equal

What does this mean? Programs that contribute to a business
process previously relied upon their position in a linear process chain to
function, but they can now operate dynamically (read “over the Web”) without
having to be rewritten. And here’s the real punch: Since the value of the very pricey
application software suites that your ERP provider wants to sell you operate in
this manner, very often legacy apps
can function with the same dynamic utility of expensive ERP counterparts, given
the right middleware configuration.

Restructure processes without revising the applications

Because you can impose conditional responses to any program
activity based on the content of the reporting, you can reconfigure the individual
steps of an application system into new, more powerful processes, affecting any
or all of the following:

  • Precision
    of event timing in the process
  • Dynamic
    levels of user access, enhancing not only security but situational
    appropriateness of user interaction with the system
  • Fine
    resolution of reporting, down to real time
  • Enforcement
    of critically timed decision making
  • Timeliness
    of decision support data
  • Efficiency
    of planning through detection of errors (in real time) at a finer level
  • Quality
    of data
  • Higher
    degrees of communication and awareness between users and departments
  • Much
    greater contingency handling in business processes
  • Isolation
    of human decision points throughout a business process
  • Empowerment
    of users at the points in a process (and only
    at those points, if desired) when empowerment is appropriate
  • Precise
    enforcement of process requirements from users
  • Greatly
    enhanced process management overall

And you don’t have to touch a single line of legacy code or
enterprise app code to accomplish these things. It’s no longer about using
middleware to reach predefined goals. You can use middleware to leapfrog your
thinking to higher, more effective goals.