Service Oriented Architecture (SOA) has emerged as a promising harbinger of agility, but is certainly not without its critics. Although SOA is not a new concept, it's not a mainstream approach and has not yet demonstrated widespread, repeatable success. It's no surprise then that it's often met with passionate resistance. Given the lack of SOA case studies and the fact that it usually demands a dramatic departure from the norm, SOA is often sidelined to make way for the more traditional, generally accepted point-to-point (P2P) integration approach.

In the absence of objective SOA performance data, the industry needs a simple model that compares the distinguishing characteristics of SOA and the P2P approach on a fundamental level so that we can make more objective judgments about the fit of SOA in our enterprises. This article presents such a model and demystifies the short- and long-term financial effects of each, illustrating that although each is aimed squarely at the goal of enterprise integration, their effects on the bottom line are vastly different - in fact exactly opposite.

This model demonstrates that a service-based IT infrastructure is significantly more financially sound than a point-based IT infrastructure when integrating a heterogeneous IT environment. Applying traditional enterprise architecture practices to the challenge of building agility is not only a recipe for disaster, but companies that currently rely on point-to-point solutions are likely to already be caught in financial quicksand.

The Anatomy of Agility
To effectively compare P2P and SOA they must first be put on a level playing field. This model assumes that the primary goal of IT is to leverage its technology and resources in a way that maximizes its value back to the business. Since today's businesses demand quick action in the face of accelerating change, speed of delivery is an integral component of the IT value proposition. Therefore, each of the two approaches will be evaluated on its ability to provide and sustain IT agility while maximizing return on investment (ROI). Before we dive into the math we need to define what agile IT looks like.

Fundamentally, companies need their IT organizations to turn strategic business ideas into valuable technology solutions as quickly as possible. However, what is considered quick for one company may be lethargic for another, which makes it difficult to establish a universal benchmark of agility. But let's assume that the market will select for the quickest companies, who are likely to have achieved as close to zero latency as practicable for their industries. For our purposes, agility is characterized by IT's ability to turn solutions around almost instantly.

For IT as an organization to operate in real-time, its infrastructure must also operate in real-time. This means that the technology that goes into producing solutions must be ready to be integrated and delivered at a moment's notice. Building such a product invariably involves blending the capabilities of many software functions from multiple sources into a unified package. So each component function must be ready to be called into the game at any time.

Furthermore, functions are chained together in different sequences to serve different business needs. And with potentially hundreds, or even thousands, of useful functions teeming in the typical enterprise, the number of possible combinations is enormous. To be agile, IT must assume that any of these integration "value chains" may need to be shrinkwrapped and delivered to the business at any time. The only way to meet this extreme, real-time integration challenge is to preemptively link every component in the enterprise to every other component. Otherwise, the integration work must occur after the order comes down from the business, thereby increasing time to delivery and diminishing agility. Through total connectedness IT can assemble applications by simply activating pre-integrated chains of components in a sequence that suits the business. This takes the integration step out of the delivery cycle and moves IT toward real-time execution. But what does a fully connected, agile IT infrastructure cost?

Measuring ROI
The primary unit of investment in the agile IT environment is the dedicated connection - the physical bit of infrastructure that links one component directly to another. Physical connections provide a transport mechanism through which information can flow between otherwise isolated functions. Component integration, and thereby total enterprise integration, can't happen without them.

Each connection, of course, comes with a price tag, which represents the sum of all short- and long-term expenses associated with keeping the connection in working order. Short-term costs encompass the work required to build the connection, such as planning, designing, implementing, and testing. Long-term costs stem from the work required to maintain the connection over time such as bug fixing, upgrading, and enhancing. Since these costs are directly related to work performed by humans, we will consider them all labor expenses and base them on an average blended labor rate. The investment portion of the ROI equation, therefore, will be the total ownership cost (i.e., the aggregated lifetime labor cost of owning each dedicated connection) of operating a fully connected, change-ready IT infrastructure.

Each time a component is linked to another, value-added information emerges from the IT environment. The connectedness of that environment lets IT combine information in new ways that are of value to the business. The more integrated nodes there are in the network, the more ways information can be creatively assembled the more value can be extracted. Of course, not all relationships between connected nodes will be active all of the time, which suggests a distinction between active links and passive links.

Active links are integration relationships that are "online." They are in use and serve a concrete business need. Passive links are relationships between components that exist but are inactive because the value provided by the relationships isn't currently in demand. When a business need calls for them, though, they are quickly switched on and become active. Whereas active links provide actual value to the business and can be used to measure the present value of the network, passive links provide potential value and can be used to measure the future value of the network. Since agility is a measure of both the present and future performance of IT, we are interested in both active and passive links. Therefore, the total return delivered by an agile IT infrastructure is the combined value of all active and passive connections. The ROI of the network is simply an expression of the difference between total value and total cost.

Now we are ready to begin the analysis. The remainder of the article compares two approaches to achieving total enterprise integration and evaluates them based on the ROI the resulting integration network offers back to the business. The first is the "classic" approach to enterprise architecture, characterized by simple point-to-point connections between isolated components and often favored for its simplicity and historical success at meeting business needs. The second is the SOA approach, characterized by loosely coupled reusable links between components and gaining popularity for its overall flexibility.

The ROI of P2P
IT traditionally tackles integration challenges reactively, responding to the arrival of concrete business requirements with the commencement of a software development lifecycle. Since the process unfolds within the scope of a funded project with defined schedule and resource constraints, the traditional model rewards IT for quick solutions that don't inject complexity beyond that required to satisfy the immediate business need. Any deviation outside the defined boundaries of the project is generally construed as unnecessary risk. Consequently, components are integrated using methods that have minimal impact on the project plan. The method that fits this profile best is the P2P approach, in which all isolated components are joined together through direct physical connections.

The benefits of this approach are its simplicity, low upfront costs, and long-standing reputation for getting the job done. Development teams aren't burdened by technical requirements that don't relate directly to the business solution and managers aren't accountable for expenses that don't directly support their project objectives. The consequence, though, is that each integration solution provides little or no value to any business case other than the one for which it was specifically designed. Because reuse is not a consideration, each IT component must have a dedicated, physical connection to all the others to achieve total connectedness. Let's examine the economic implications of this approach.

Suppose we have an IT infrastructure, or segment thereof, consisting of 10 components that each perform a business function (e.g., "create application," "calculate credit score," "authorize user") that has value outside its native system. We wish to link them all together into an agile, change-ready network. As illustrated in Figure 1, we begin by linking two nodes together, yielding one connection. Adding a third node, we must link it to each of the first two. Adding this node has caused the total number of connections to increase to three. Adding the fourth node requires linking it to each of the existing three, producing a grand total of six dedicated connections.