Archives of the TeradataForum

Message Posted: Tue, 19 Oct 2004 @ 09:05:49 GMT

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

Hi Pete,

My understanding is...

The two forms of redistribution are used by FastExport (and nothing else as far as I know) to arrange the data rows into blocks ready for the data transfer stage.

Once Teradata has built the answer set, using normal duplication, redeistribution, sorting etc as required by the SQL and table definition, the "fastexport" processing takes over.

Without an "ORDER BY" clause the answer set is chopped into 32 (64 ?) KB blocks and then sent across the lan/mainframe channel to the host. Thinking this through I'm not certain if this requires a redistribution of the data or if that's just used to describe a process.

With an ORDER BY clause rows that will be consecutive in the final answer set are placed on the same AMP (I guess within ranges) so that when they are divided into the final data blocks and transferred to the host you get the output rows in the correct order.

Does that help ?

Cheers,

Dave

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