Archives of the TeradataForum

Message Posted: Wed, 07 Feb 2002 @ 04:16:01 GMT

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

Hi Tom:

Thanks for your additional thoughts regarding my suggestions on this topic.

Granted, I have not performed extensive research on his system nor seen an EXPLAIN. That is why I suggested that he "experiment" with these options. Short of declaring a different PI on the table to improve the join by eliminating row re-distribution, these might accomplish what he seeks, some speed.

Thank you also for your opinion on time added by the TJ to insure data integrity. Fast, slow or indifferent, overhead is overhead and if it is eliminated, the speed improves. After all, row re-distribution is another form of overhead. Eliminate it and the join normally run faster. On the downside, if the re-distribution skews the data too badly, the join will not execute due to a lack of SPOOL space.

Also, when considering overhead, you must also take into account the potential for things like: block splits, potential for mini-cyl packs, and etc. But these were already mentioned by other people - hence my reference to them. But, I was trying to be a good list citizen and trimmed them from my reply. I did not imply that my comments would provide a perfect solution. Rather, they are things to try.

My impression is that the T-DATA list is a resource and not intended for extensive consultation. Therefore, if the options I suggested do not provide what he needs, then more drastic measures should be attempted.

Thanks again for your follow on comments directed to me,

Mike

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