82 Automated Physical Database Design and Tuning
To obtain this cost, we calculate cost(q, {I}) − cost(q, ∅) (either by using
optimization calls or by leveraging the cost model of the optimizer). Therefore,
we derive the cost of a nonatomic configuration C for an UPDATE query q as
cost(q, C) = min
C
⊂C∧atomic(C
)
cost(q, C
) +
I ∈ inner SELECT
cost(q, {I})
−cost(q, ∅)
So far we have discussed the definition of atomic configurations and how
they can be used to avoid some optimization calls but have given no proce-
dure to identify such configurations for a given query. We next introduce two
approaches to address this issue.
5.2.1.1 Static Strategy Based on Query
Processor Restrictions
The characteristics of the query processor can determine what constitutes ...