Archives of the TeradataForum

Message Posted: Mon, 24 May 2010 @ 18:52:51 GMT

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

Hi -

Is there a real performance difference in Teradata environment whether a null value for a dimension key in a fact table is left as null or is converted to a single dummy dimension key? I see the benefits of outer joins being written as equi-joins, but not certain if the performance difference is material or even if there is any. When the fact table is joined to this dimension table, and if the query plan were to call for the fact table to be spooled and redistributed on the dimension key, whether the dimension key is null or a dummy key, all these rows will be hashed to the same AMP and will cause massive skew, if the dimension key is sparsely populated. How much of a concern is this?

Thanks.

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