Archives of the TeradataForum
Message Posted: Fri, 15 May 2009 @ 09:50:05 GMT
Note that the PPI does not overrule the PI. The PI distributes the records across the amps, the PPI then creates a virtual partion on each amp so that they do not all have to be scanned for most Select queries and joins.
PPIs should be designed in the light of table usage. If the table is large and users consistently use one or two columns as a constraint e.g. Branch_Number(PI) and Month, then PPI will effectively accellerate those queries by constraining the number of records scanned. If most queries are only constrained by another column that is not in the PI, e.g. Month, then simply use that for PPI.
|Copyright 2016 - All Rights Reserved|
|Last Modified: 27 Dec 2016|