Archives of the TeradataForum

Message Posted: Tue, 09 Jan 2001 @ 20:11:44 GMT

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

Ooooops.....Correction to the second bit below mentioning Unions - I meant to say that a complete Table View would be a large Join and, again, would this get materialised before any query predicates get applied?

The Union thing was something else we're discussing on breaking a large table into separate chronological physical tables (this year, last year etc.) for operational reasons and disguising this behind a single View which Unions these together - seems to present the performance problem of view materialisation before query predicate application.

Thanks,

Michael.

------------------------------------------------------------------------
Thanks all for your incredibly quick responses.
Normalisation wins!!!!

With Very-Thin-Super (eg. Acct-No, Tran-Id, Tran-date, Tran-Type, Tran-Amount) and lots of different Thick-Subs I'm still grappling a little with when (if ever) to promote frequently used data of some dominant Sub up to the Super. If all tables use the same Primary Key data (UPI/NUPI) as John Wight suggests, and queries join via this key for optimal preformance, then why bother?

One other thought, to provide a view of the complete Table I presume this involves Unioning all the Super/Sub combinations. I believe that Teradata will materialise the view (ie. all the Super/Sub Selects) before applying any extra query predicates. Is this so? Is this not a potentially wasteful overhead? Any better techniques?

Cheers,

Michael.

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