I'll start by saying you should go get a book about data warehousing. There is no way I can explain this in a few short posts.
A very common data warehouse strategy is to build a Data Mart. This is a database full of information from one or more OLTP systems that will be used to build OLAP cubes (Analysis Services). One commonly used structure for this Data Mart is called the Star Schema. This is basically a single table (Fact) that contains your rows of information - let's call them units and dollars for this example. In addition, you would find a number of foreign key fields in this Fact table referring to the Dimension tables. These dimension tables would hold the information that describes the attributes about these facts.
Since this schema is not very relational, it is very easy to traverse the joins (they are all a single level and can only return a single dimension record), this structure makes it very easy to calculate aggregations on the fact information (units and dollars) for any attribute in a dimension table. So, if you had a dimension for Customer and one for Product, it would be very easy to calculate the units and dollars for each Customer, for each Product, of for a comination of Custmers and Products.
This calculation of aggregations is basically what you get from an OLAP cube. It not onl can hold the information, but it can hold some or all of these aggregations making it easier to display them to a user very quickly.
Again, there are lots of books about this. If you plan to do any large-scale reporting projects, pick one up and at least read the first chapter.