Integration involves more than data. It encompasses a myriad of tools, systems, models, standards, and functions. But you can boil down this complex process into three distinct integration models: data consistency, application integration, and process integration.
 Tom Kevan |
Whether your data come from sensors, control devices, or production management systems, one of the first steps toward integration is to create data consistency across various independent applications. You need an environment where transactional information—information that is updated, modified, or created—automatically and seamlessly flows between systems so that when information is updated in one system the change is automatically effected in the other linked systems without manual processing. To achieve the necessary data consistency, you must convert the data so that they use common formatting, data models, and logic.
This type of integration starts with message-oriented middleware, which ties independent systems together to achieve data consistency. Data integration is also achieved through the use of adapters (similar to network gateways) that understand the business logic and the data models of each of the systems and abstract them into services that can be reused in support of solutions that will integrate the systems.
The next model—application integration—represents the convergence of application development and application integration. In the past, applications were developed to address stand-alone functions or to solve a particular operational problem, and they were developed from the ground up. Today, new applications have to layer new functionality on top of existing fundamental applications (e.g., manufacturing execution systems and enterprise resource planning systems), while at the same time reusing the functionality and data that exist in the legacy systems.
In a recent conversation, Trevor Matz, Managing Director of Ensemble for InterSystems Corp., which provides integration platforms, observed, "You can leave the application infrastructure that you have in place and layer on top of it an integration platform upon which you can develop additional functionality while reusing and leveraging the existing application functionality that you have. That is what is called composite applications." This level of integration is also supported by service-oriented architectures that provide a way of creating services from existing applications by decoupling functionality from existing applications to allow it to be used as a service or a component in support of a new application.
In the third model—process integration—you model and orchestrate workflow between different applications to automate, optimize, and streamline previously independent processes. The goal is to integrate long-running business processes that span multiple applications, as opposed to just trying to coordinate transactional data between systems.
"Instead of manually moving that process between different applications, you're trying to model the process from an analytic business management perspective," says Matz. "And once you've modeled the process, you integrate and orchestrate the underlying applications to automatically support the execution of the process according to the model that you have created."
InterSystems and Motoman, Inc., are putting these theories into practice in an integration implementation for a medical laboratory. Here, Motoman, the second largest robotics company in the Americas, is automating the laboratory's pre-analytic processing of specimen samples.
"In the process of implementing the automation to support the handling and sorting of specimens, we are interfacing with the customer's business system to obtain information about each sample, including the sorting criteria and the test target associated with each of the incoming specimens," says Tony DaSilva, Director of Operations for Motoman. "The business system, via the middleware, will provide information on when specimens arrive and are sorted, as well as notification of exceptions that may need to be resolved either manually or through another process."
In addition, the new system will provide operational information about the laboratory to the lab managers so that they can make decisions based on real-time process data. If managers need to dynamically reconfigure the automation platforms to meet changing conditions throughout the day, the system will provide them with the necessary information.
In the end, integration's reward is a better grasp of the big operational picture and enhanced control for managers at all levels through access to comprehensive real-time data.
Tom Kevan is a freelance writer/editor specializing in information technology and communications.