BI Applications and Embedded BI, Part 1
Posted by Dylan Wan on July 7, 2007
Oracle BI Applications is build on top of the platform from the Oracle BI Enterprise Edition. The dashboard and report components can be seamlessly embedded into Oracle Applications, like how it is integrated with Siebel application. However, it is also a very typical data warehouse architecture.
Sometime, people feel that only those BI application built directly on top of the OLTP schema can be embedded into the OLTP applications and provide the real time BI. Many also believe that only the people build the OLTP application can understand how the BI is used in the business flow. In fact several years ago, I have both of these views. After I worked for Oracle BI Applications for a couple of months and get deeper into it, now I think that both views are wrong.
Embedding the BI into operational flow and empower everyone in the enterprise is absolutely needed. However, the data stored in the typical OLTP schema is still need to be summarized, computed, aggregated to become meaningful business metrics. All these processes require computer resources including the processing power and disk space. The space growth pattern is very different from the space management for OLTP. A data warehouse requires you to hold the historical data which can grow very fast while the data is not frequently updated. Unlike the typical OLTP process, which typically include individual transactions and each of them add or update records in several tables, the data warehouse process involves much more data in each run and the batch process is required. You do not want the transactional system’s performance to get impacted.
A key driver is that a typically enterprise system is running a hetergeneous environment. The data can actually sitting in many different places. The embedded BI solution directly based on top of the OLTP tables assumes only a single OLTP application to be used. It is typically not the case. The consumer apps of the “intelligence” data is very likely a different apps from the provider apps. Even worse case is that the intelligence data can only be meaningful and valuable when you combine data from multiple systems!