Skip to Content
0
Former Member
Nov 14, 2012 at 04:47 PM

What is the recommended way of modeling ID, NAME dimensions?

34 Views

Aside from a column-based table (the obvious suggestion), the alternatives also seem to make sense:

1. Get rid of the ID and move the NAME down to the fact table. I've read in the Efficient Transaction Processing in SAP HANA Database – The End of a Column Store Myth paper that Hana uses internal dictionaries for storing CHAR columns, which is equivalent to our initial dimension table. I can see that Hana would have more work to do when loading the data though.

2. Use row-based tabled. In practice, the dimension table is always accessed to translate from NAME to ID or viceversa. You always need both columns, there's no benefit of separating them. But would queries perform poorer when joining a row-based dimension table with a column-based fact table?