Archives of the TeradataForum

Message Posted: Sat, 10 Dec 2011 @ 09:07:46 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

I have a large table with a 2-column NUPI and 2 levels of MLPPI.? The data types and cardinalities are as follows:

NUPI_col_1 = integer, ~100 distinct values
NUPI_col_2 = varchar(20), ~1 billion distinct values PPI_L1_col = byteint, 4 distinct
values PPI_L2_col = date, several thousand of distinct values

Normally, these are the stats I would collect:

     COLLECT STATS tablename COLUMN ( PARTITION ) COLLECT STATS tablename COLUMN ( NUPI_col_1
     ) COLLECT STATS tablename COLUMN ( NUPI_col_1 , NUPI_col_2 ) COLLECT STATS tablename
     COLUMN ( PPI_L1_col ) COLLECT STATS tablename COLUMN ( PPI_L2_col ) COLLECT STATS
     tablename COLUMN ( PPI_L1_col, PPI_L2_col )

Now, DIAGNOSTIC HELPSTATS also recommends:

     COLLECT STATISTICS tablename COLUMN ( PARTITION , NUPI_col_1 , NUPI_col_2 ) COLLECT
     STATISTICS tablename COLUMN ( PARTITION , NUPI_col_1 , NUPI_col_2 , PPI_L1_col,
     PPI_L2_col );

Question: Are these additional stats necessary or are they redundant??

Attila Finta

	<Prev	Next>		<<First	<Prev	Next>	Last>>