Archives of the TeradataForum

Message Posted: Wed, 07 Jul 2004 @ 16:47:47 GMT

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

Q1 : Question on confidence levels in TDQM.

Manual indicates that for TDQM release 04.02 and release 5, the minimum confidence levels can be set by selecting one from this list (in this order) * High, * Medium (index), * low, * no. This seems to imply that 'index confidence' is of lower quality than 'high confidence', but I thought 'index confidence' was the highest level of confidence. What is the order of confidences? did TDQM get it wrong?

Q2 : If I create a TDQM rule with limits of row count N and time T, and set the minimum confidence as medium, then for the query with explain below:

a) is the row count applied this way? Reject if any {n2, n4} >= N.

b) Is n5 never used in TDQM evaluations?

c) Is the time limit applied this way?

Reject if t1+t2+t3+t4 >= T

or do we take only the estimates with the minimum confidence, reject if t2+t4 >=T

or is the time limit applied on every individual step reject if any {t2, t4} >= T

-----------------------------

Lets assume the explain for a query is

     Step 1.
        The size of Spool x is estimated with no confidence to be n1 rows.
        The estimated time for this step is t1 seconds.

     Step 2.
        The size of Spool y is estimated with index confidence to be n2 rows.
        The estimated time for this step is t2 seconds.

     Step3.
        The size of Spool z is estimated with low confidence to be n3 rows.
        The estimated time for this step is t3 seconds.

     Step 4.
        The size of Spool p is estimated with high confidence to be n4 to n5 rows.
        The estimated time for this step is t4 seconds.

-----------------------------

Q3 : In general, how have people come up with TDQM threshold numbers? Did you run explains and pull the numbers from it? Or was it based on your knowledge of tables and row counts?

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