Better compression due to column based storage grouped by native datatype ?

Hi HANA devs,

lets assume 2 column based table definitions: (c40 = 40 chars width, n3 = 3 digits width, uVal = value (unique by unique index))

Table T1:

Table T2:

Does HANA store these two tables internally like ... ?

Table T1_Col_1

Table T1_Col_2

Table T2_Col_1

Table T2_Col_2

--> This assumes a "per table" columnar design, good compression if original tables containing enough duplicate values to squeeze out.

Several years before HANA, we emulated a "columnar" design in a row based DB by normalizing all DB tables to a small amount of technical key/value tables grouping the values just by technical datatypes. With this approach ("store together, what's technically the same type"), the 2 tables could be stored like

Table T_c40

Table T_n3

--> This assumes a "per database" columnar design, allows much more potentional duplicates to be squeezed out

We also tried a design with just one data storage table (besides the index tables to link the T_uVal entries to its original tables) like ...

Table T_uVal

--> we got all the "really unique" values in one big table now.

How is this done internally right now ?

Thanks for clarification,

Matthias