Archives of the TeradataForum

Message Posted: Thu, 03 Sep 2009 @ 22:02:06 GMT

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

Ah, here's a method, that I think originally came from Dieter, who suggested it to mi companero Keith Jones on the Master's list, who is implementing it on one of our tables. Pretty cool, but I'll just quote:

"The way I did it before [using HASHBUCKET(HASHROW())] would fail if the hash of a new ... value were zero. It also uses all possible partition IDs, precluding the use of multi-level PPI in V12. This new approach will handle all possible hash values, and will use only 512 partition IDs. ... redo the PPI using the new approach:

     PARTITION BY RANGE_N(HASHBUCKET(HASHROW(char_col))
               BETWEEN 0 AND 65535 EACH 128);"

Yep, I think it's pretty cool.

iv

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