Vertica Encoding approach
hi
Trying to look at DBD encoding output using designer_design_projection_encodings function on v7.2. Table had a column int, when looking at data, it has only one value that is repeated for 20MM records, and this column is part of the order by clause in the projection.
However, DBD suggested DELTARANGE_COMP. Documentation says DELTARANGE_COMP is suited better for floating-range or float columns.
For the same table, I had an other column, which is also int, looking at data, it has only one value that is repeated for 20MM records, and this column is also part of the order by clause. But this time DBD suggested COMMONDELTA_COMP.
For the third field, which is numeric(18,0), samples shows only 5 unique values, DBD suggested BLOCKDICT_COMP.
In few of my tests, when I look at the sample data and apply the encodings (mostly with RLE) that I think are the best, query response times are much better. I hope DBD can give us better encodings that we need not have to look back to verify it again or do you think DBD suggestions could be based on something else that I'm missing here? Please advice
thanks