Intelligent Enterprise

Better Insight for Business Decisions

RSS
Webcasts
Whitepapers
Subscribe
Home

December 10, 2003

Smart Synergy

With real-time business becoming a key objective, the worlds of enterprise application integration (EAI) and business intelligence (BI) are converging. What will it take to forge the most perfect union?

by David S. Linthicum

Continued from Page 1

The New Stack

To understand how integration fits with BI, we need to create and understand a new architecture, or stack. It has several components that are adapters (both information- and service-oriented), translators (both real time and batch), persistence, abstractors, and analysis (real time, historical, and both). (See Figure 2.)

The adapters portion of the stack, similar to the adapters that come with almost every EAI software available, provide the critical link between a source system (such as a database or ERP application) and the application integration/BI infrastructure. We need two types of adapters, however: information-oriented and service-oriented.

Information-oriented adapters let us externalize information, including content and schema, out of a source database or application. Service-oriented adapters do similar things, but they interact with application behavior, letting us either abstract that behavior so it's usable within other applications (in other words, composite applications) or abstract the information through the invocation of a remote service.

The goal is to use these adapters for information and service visibility rather than information replication, which is a more traditional use for adapters. Also, it's important to note that it isn't practical to use more than one adapter on any given system, because a single adapter already serves dual purposes. The adapters need to service all aspects of replication in support of a business process, as well as give visibility for real-time analytic systems.

As the name implies, translators let the stack change information schema and content as needed to create a single abstracted schema and content for use in the higher portions of the stack. Translators need to transform information both in real time to support the real-time business processes and in batch to support more traditional ETL-type activities. For BI, the translator needs to provide visibility into an abstracted schema, making real-time information — perhaps from many systems — appear as a single schema, not unlike traditional federation middleware or the newer IIS technology. Again, only a single translator is needed rather than several types of translators from several vendors operating on the same information. This translator needs to handle very complex data.

Persistence is required to capture real-time data, which will shortly become near-time and eventually be historical. In the world of traditional EAI, we did this through a mechanism known as message warehousing, where real-time data was captured and held offline for any period of time and for any use (typically rudimentary analysis). In the context of the type of integrated system we propose here, the persistence level would act as a staging area of sorts, a source for information that's neither real time nor historical.

Although translators perform some abstraction to prepare the real-time information for viewing, we still need to abstract this abstracted real-time information with existing information contained in operational data stores, data warehouses, or both, at the same time. We call these components abstractors. Abstractors combine the old with the new, providing a single view into meaningful schemas and content for use by the analytic tools.

Finally, the analysis portion of the stack gives us a mechanism for making sense of the real-time, near-time, and historical information. We use this technology to slice and dice the information as needed, extracting the right information that lets us make great decisions that take into account all the relevant information from the current state of the business, as well as past states. For the most part, this type of analysis draws heavily from the data warehousing world, with newer applications within the emerging world of BAM. It's important to note that we are considering all the current and past information, where traditional BI tools typically consider only older aggregated data, and traditional BAM considers only real-time data. This valuable synergy lets us understand the present through an understanding of the past, and it helps us compare past and present, side by side, using any analytic view required.