Glossary - Relational Database Index Design and the Optimizers - page 295

Glossary

INDEX DESIGN APPROACH

The following terms and abbreviations are those we have used for the index design
approach described throughout this topic. They are not the official terms used by any of
the database management systems in use today.

Algorithm for deriving the best index A fat index (third star) is designed, where the
scanned index slice is as thin as possible (first star). If this index does not imply a
sort (second star), it is a three-star index. Otherwise, it will only be a two-star index,
having sacrificed the second star. Another candidate should then be derived that
eliminates the sort, thereby having the second star but having sacrificed the first star.
One of the resulting two star indexes will then be the best possible index for the
given SELECT statement.
Assisted random reads A single term used within this topic to represent automatic
skip-sequential, list prefetch, and data block prefetching.
Assisted sequential reads Atermused within this topic to represent splitting a cursor
into several range-predicate cursors, each of which will scan one slice; when several
processors and disk drives are available, the elapsed time will be reduced accordingly.

BJQ, Basic Join Question Is there an existing or planned index that contains all the
local predicate columns (this includes the local predicates for all the tables involved)?
BQ, Basic Question Is there an existing or planned index that contains all the
columns referenced by the WHERE clause (a semifat index)?
Best index The best index that can be designed for a SELECT statement. It may have
all three stars, in which case it is also an ideal index; it may, however, have only two
stars, having sacrificed the first or the second star because of the presence of an
ORDER BY together with range predicates.

Call-level exception monitoring Producing exception reports that show the slowest
SQL calls in a monitoring period; they can be of great assistance in finding the slow
SQL calls in a slow program that has a large number of different SQL statements.
Candidate A and B The two possible indexes used in the algorithm for deriving the

best index.
CPU coefficients The values used to calculate the CPU time for random and
sequential touches, FETCH, and sort.
CQUBE, CPU Quick Upper-Bound Estimate A quick, rough estimate of the SQL
CPU time using four variables: TR, TS, F, and RS.