DBD recommendations based on amount of data?

DreaBrandford

Is the amount of data in the tables important for the design recommendations made by DBD? In other words, does a large amount of data required for the DBD to make a good recommendation ?

Asked during BDC session Vertica Database Designer – Today and Tomorrow
@ybei

JoeGonzalezVertDBA

I think it makes a difference when it comes to recommending a projection that is unsegmented (which is recommended for smaller tables) vs a projection that is segmented. My personal rule of thumb is to default to an unsegmented projection on a table with less than 1 million rows. I am not sure exactly what amount the DBD uses when recommending one vs the other but I do believe itt takes itt into account for this purpose only.

Jim_Knicely

Check out:

dbadmin=> SELECT parameter_name, default_value, description FROM configuration_parameters WHERE parameter_name IN ('DBDRepLargeRowCountPct', 'DBDRepSmallRowCountPct', 'DBDLargestTableRowCountBoundary');
         parameter_name          | default_value |                                                                                                      description                                         
---------------------------------+---------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
 DBDRepLargeRowCountPct          | 1.000000      | If the largest fact table has more than <DBDLargestTableRowCountBoundary> rows, then use this percentage off the largest fact table to define the number of rows below which a table should be replicated Default: 1%
 DBDRepSmallRowCountPct          | 10.000000     | If the largest fact table has less than <DBDLargestTableRowCountBoundary> rows, then use this percentage off the largest fact table to define the number of rows below which a table should be replicated Default: 10%
 DBDLargestTableRowCountBoundary | 1000000       | Number of minimum rows expected in Fact Table, default 1M rows
(3 rows)