The typical business intelligence (BI) architecture can be seen as having a stack of layers. The base usually comprises source data systems, from where data is processed by "extract, transform, load" (ETL) software into a data warehouse. Above that are the BI and application layers and then at the top there is the presentation or delivery layer, which can include executive dashboards, scorecards and other decision-making tools.
Software houses used to specialise in making applications for these different layers, which meant that businesses would assemble their own stacks using independent suppliers. So a company might have an SAP enterprise resource planning (ERP) system that feeds into an Oracle data warehouse, for example, and the firm's finance department might use an application from Hyperion for reports and another from SAS for analytics.
As BI has evolved, the greatest challenge has been how to integrate data on different systems accumulated from different vendors over many years. This challenge is being addressed in the following key ways:
* "Service-oriented architecture" (SOA) has been developed as a solution that eliminates the need for point-to-point connections between resources. It provides access to data in legacy systems through linked "services".
* ERP, ETL, data …