Champagne corks pop and the team is noisily patting each other's backs. It's been two weeks since the new data warehouse went live, and feedback from the top floor is positively glowing.
Your mobile phone rings. It's the CFO, who is so happy with the new customer profitability data that she wants profitability by employee! The champagne blush drains from your cheeks. You'll get back to her, you say.
Once again, the axiom "the more successful the solution, the more likely it will change" proves painfully true. Users realize added value from newly available data, but the CFO doesn't realize that HR data is carried in an entirely different data mart.
Most large organizations operate with a collection of automation islands. When information must be shared, the IT department builds bridges between islands. Each change requires more bridges. Wouldn't it be wonderful if stranded systems could put a message in an intelligent bottle and let it sail to exactly the right automation island?
Enterprise application integration (EAI) does exactly that. EAI takes disparate technology automation islands, messages and containers and adds some 21st century capabilities to intelligently move data throughout the enterprise archipelago.
EAI is a recently coined term for enhanced middleware. Middleware is not new; the venerable CICS and Tuxedo transaction processing monitors are examples of middleware. Remote procedure calls are middleware, too. In fact, any software that stands between business logic and the operating system or network protocol can be considered middleware.
EAI allows data sharing between unrelated systems in the organization, provides a single point of interface to which all applications and databases connect, resolves differences between systems, triggers processes and delivers data in the proper format to the proper destination.
Enhanced middleware makes it possible to assemble loosely coupled systems including nonintegrated marts. This means that existing applications do not need to be modified to share data. No bridges or point-to- point interfaces are constructed between systems or platforms. Instead, each system connects to the EAI architecture like the spokes of a wheel connect the rim to its hub. Within this software hub, data is directed to the right spoke and then out to a waiting application.
EAI middleware comes with the ability to transform as well as route data. Data formats and business rules, which govern the exchange of data between systems, are applied to data when it is received by the EAI software. These specifications and rules can be modified, accommodating changes in one application without impacting the other systems or interfaces.
Placing a dynamic middleware layer between applications produces a number of benefits over traditional, point-to-point integration architecture:
Flexibility. Applications produce and read data in their native format. EAI software restructures that data, according to routing rules, so that the data format is meaningful to each application that will receive it. Interface maintenance is performed by reconfiguring these data specifications and business rules within the EAI software. The applications and interfaces themselves are unchanged.
Functionality. System enhancements may be added to the enterprise, as if adding additional spokes to a wheel. Existing spokes remain undisturbed. Data can be redirected from any number of sources to any number of targets, and new data may be combined with existing data. The newly enhanced or enriched data may then be distributed to old and new applications alike.
Real- Time Data. Message-oriented middleware can move data as it becomes available. Processing can be triggered by demand or by system event, rather than strictly by time.
Research firms GartnerGroup and Forrester agree that approximately 35 percent of IT dollars are spent building point-to-point interfaces. Using this percentage to provide a rough estimate, 35 percent of an implementation budget will typically go to integration tasks.
EAI interfaces are not necessarily less expensive to implement than point-to-point, but an EAI hub can result in fewer interfaces overall. If one were to connect every system to every other system, the number of resulting interfaces would follow this basic formula: For "n" interconnected systems, the total number of point-to-point interfaces equals n(n-1)/2.
Following this formula, eight systems would require 28 point-to-point interfaces to fully interconnect them all. An EAI hub-and-spoke architecture would require only eight interfaces, one between each of the systems and the hub.
Rarely do all systems connect with all of the others, but it is not unusual for more than one interface to exist between one pair of systems. So, the hub-and- spoke architecture can result in fewer interfaces; and with fewer interfaces to build, there can be a cost savings.
Unfortunately, development cost is not linear. Complexity is a significant factor. Initial EAI implementations can be highly complex, and EAI software can be expensive and difficult to learn. The difference between costs and savings on the number of interfaces can reasonably cancel one another.
An EAI architecture is not a simple solution. Traditional interfaces can be faster to develop and often deliver faster performance. The suitability of EAI depends upon the size of the enterprise and the relevance of EAI's benefits. Also, the use of EAI does not obviate the need to have an enterprise-wide data warehousing strategy and architecture in place.
Where is the increased return from EAI? This value can be determined by:
- The total cost of interfaces over time.
- The extended life of legacy enterprise systems integrating new functionality rather than completely replacing older applications.
- The total level of effort required for initial implementation, plus long-term maintenance and operations.
- The financial value of operational performance (e.g., system response, load and processing time).
- The need to satisfy both current and future business requirements.
- Anticipated business growth.
Value is primarily realized down the road, where new business requirements, legacy system upgrades, and mergers and acquisitions can alter the enterprise IT landscape. EAI architecture is intended to accommodate such change, whereas rebuilding invalidated point-to- point links will increase the investment in system interfaces over time.
One thing is clear: EAI is a core technology architecture. To realize an increased return on investment, interfaces must be considered at the initial design stage of any project. EAI software won't rescue an inappropriate infrastructure, but it can help accommodate the future.
Michael J. Schroeck is a partner and the global leader of PwC Consulting's iAnalytics solution set. An early pioneer of integrated analytics and data warehousing solutions, Schroeck can be reached at mike.schroeck@us.PwCGlobal.com. PwC Consulting is a business of PricewaterhouseCoopers.