What is a data mart?
A Subset of the Data Warehouse
The data mart is a subject-oriented slice of the data warehouse logical model, serving a narrow group of users. Many only need a subset of data from the full tables in the data warehouse. For example, a mart may only have sales transactions, products, and inventory records. Most only have 5-20 tables instead of 4,000.
Data Mart Fact Tables
The number of tables in a data mart has no relevance to the size of the database. The main tables–called fact tables–can be 100s of terabytes of call detail records for a telecommunications company, for instance. The data mart itself can be huge, but it is a narrow selection of all the data available in the data warehouse.
Data marts solutions are often denormalized, capturing only summaries of data by sorting it and aggregating a result table, usually throwing away detail data. Some are completely reloaded weekly or monthly; it is relatively easy to delete all the data and refresh it so that reports only look at the last 30 days of transactions.