Archives of the TeradataForum

Message Posted: Fri, 12 Sep 2003 @ 18:07:52 GMT

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

The original implementation of DT left the result wherever it was created - local spool. It then included a separate redistribution to a new spool (if necessary) to prepare it for the next join. The separate redistribution is no different than any other spool redistribution, it is hashed on the column(s) for the join it is getting ready for. Closer to R5, more intelligence was added for the simpler cases so many cases will not have the extra redistribution. And with R5 (assuming no aggregates...) the DT is just pretty syntax - the DT itself will not be materialized, but will just be folded into the join plan.

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