working on applications and technology

There 're many tools in the market that we could use to architect a business intelligence/data warehousing project, The key components for such project usually includes

• Business Intelligence tools
• Data storage and data modeling layer
• Data integration or ETL/ELT tools

This is an simple overview

The business intelligence architecture 1 - use only OBIEE

This architecture is perhaps the most simple architecture for business intelligence , we could build business intelligence applications using only Oracle BI tools - OBIEE.

The reason we could neglect the data storage layer and data ETL tools in this architecture is that OBIEE comes with the BI Server component which could build virtualized multi-dimensional data model over many existing herterougenous data source. All the analysis request for the 'virtual' data model will be translated into SQL request  and sent to backend datasource for executions. BI server will then collect and consolidate results from different data source.

Of course this is more like an demonstration purpose architecture, for most of the time we usually draw data from different data source and place it into the data storage layer and this is the 2th architecture.

The business intelligence architecture 2 - use OBIEE plus Oracle database and ETLtools

The difference between this architecture and the above architecture is to add Oracle database and data ETL tools , We will use ETL tools to draw data into this database and organize the relational tables into star schema data model , and such approach is also called the ROLAP approach , which use relational tables for multi-dimensional modelling.

This is an common and popular architecture for many business intelligence project.

It is simple and good for a starter business intelligence project.

The business intelligence architecture 3- add Essbase

The more complex architecture for business intelligence is to add MOLAP component (e.g. Essbase) into the architecture.

Why we need another database (essbase) anyway if we could build multi-dimensional data model using RDBMS database?

This is an tricky and puzzled question for many people. You will not understand only until you learn the detail of Essbase.

Essbase does comes with very powerful data analysis capability that is difficult for the RDBMS OLAP model and here 's a few such features

• Write-back capability - very useful for What-If analysis
• physically and precomputed storage cell value - very fast query speed
• ...