Archives of the TeradataForum

Message Posted: Wed, 09 Sep 2009 @ 16:16:39 GMT

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

Might be worth investigating in each particular case. You need to separate deterministic and random factors affecting the events and then statistically evaluate the hypotheses on the probabilities involved. There are many software packages such as R, SPSS or SAS which facilitate the calculations.

This is not an endorsement of any of them, just an indication of the genre. These or similar packages are often used by DWH business users but they are less known in the technical circles, which is a shame because statistical methods can successfully be applied to the technical aspects of Teradata.

The working of the Teradata optimizer is based on the same statistical principles so why not extend the methods to the daily DBA work? Statistical methods can be very efficient on such large systems as Teradata leading to the same or better results and insights than traditional number crunching.

Instead of expensive logging of performance data for each and every user and query DBA could consider less intrusive sampling based on statistical methods.

On the less serious note, there seem to be people who feel much more confident with predictive analytics:

uk.tv.yahoo.com/blog/article/245760

Could be a great way for the UK to get out of recession in style.

Victor

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