Columnar databases are an old idea that becomes actually again. The first introduction of columnar databases was in 1970s. However in the 1980s and 1990s the huge improvement of hardware puts relational databases in a front page. Nowadays, a huge expansion of analytical (business intelligence/data warehouse) solutions working with large volume of data allow columnar databases to back in the game -  as approach that could improve efficiency of analytical systems.

A columnar DBMS stores its content by column rather that by row. However, a database program must show data in a format of two-dimensional tables, but store it as one-dimensional strings. For instance look at following table:

A row-oriented database serializes all of the values in a row together, then the values in the next row, and so on.

Here is the example of row serialization:

1, Smith, John, 10000, 2, Bolton, Nick, 50000, 3, Jones, Leanne, 40000

A column-oriented database serializes all of the values of a column together, then the values of the next column, and so on.

Here is the example of row serialization:

1, 2, 3, Smith, Bolton, Jones, John, Nick, Leanne, 10000, 50000, 40000

The major, simplified advantage of column oriented databases over row oriented databases is in the efficiency of hard-disk access.

References:

Wikipedia

David M. Raab , How To Judge A Columnar Database, Information Management Magazine, 2007

Lou Agosta, Columnar databases, appliances, cloud computing top BI trends,  SearchDataManagement, 2009.