Hbase column family design importance

Question

I am studying HBase but can't find for myself answer for one question.

Let's consider the following situation. We have five physical (hardware) servers (0-4). Hmaster is installed on server 0 and four hregion servers are installed on server 1-4. And we have one very big table which we need to work with these five servers.

As I understand every region server is responsible for certain region (some set of rows(!)). It means that always one row (including ALL its column family,columns and cells) is located only in ONE region server (in our example in ONE physical server).

If what I wrote is right I can't understand what is the use and importance of column family. Please correct me if I am wrong and/or exmplain what column families are used for.

Answer 1

It's more for IO performance when you scan/fetch. If you find yourself only using columns X, Y, and Z but not A, B, and C during a scan/fetch, you can partition your data into two column families to improve IO performance.

There is probably also a small benefit for compressed tables too since like data could be physically grouped together and thus more easily compressed.