Architecting HBase Applications

Chapter 15. Too Many Column Families

Many HBase users start to design HBase tables before reading about it and before knowing about all the HBase features and behaviors. People coming from the RDBMS world, with no knowledge of the differences between a column family and a column qualifier, will be tempted to create a column family for each column they have of a table they want to migrate to HBase. As a result, it is common to see tables designed with too many column families.

For years, it has been recommended to keep the number of column families under three. But there is no magic number like this. Why not two? Why not four? Technically, HBase can manage more than three of four column families. However, you need to understand how column families work to make the best use of them. The consequences explained here will give you a very good idea of what kind of pressure column families are putting on HBase. Keep in mind that column families are built to regroup data with a similar format or a similar access pattern. Let’s look at these two factors and how they affect the number of column families:

Regarding the format

If you have to store large text data, you will most probably want to have this column family compressed. But if for the same row you also want to store a picture, then you most probably do not want this to be compressed because it will use CPU cycles to not save any space so will have negative impact on the performances. Using separate column families make sense here. ...