Glossary - Relational Database Index Design and the Optimizers - page 299

Hash join A join method where one table is first stored in a temporary table (local
predicates having been applied), hashed by the join column(s); for each row of the
other table that satisfies its local predicates, the temporary table is checked for
matching rows, using the hash value.
Hint A specification given in an SQL call or a bind option to influence the optimizer;
the syntax is product specific. It should only be used when the optimizer is not able
to choose the best access path because of inappropriate cost estimates.
Host variable A program variable that is used in a WHERE clause, such as
:SALARY in WHERE SALARY > :SALARY.
Index matching The use of predicates to restrict the size of an index slice to be
scanned. One or more columns (sometimes called matching columns) identify the
beginning of the slice and use the B-tree index structure to find the first index entry
required. The columns will also determine the end of the scan. Matching predicates
are sometimes called range-delimiting predicates.
Index only An access path that is able to provide all the data requested without
requiring access to the table.
Index read-ahead An SQL Server term used to represent reading-ahead the next leaf
pages following leaf page splits.
Index screening Index columns that cannot be used in the matching process may still
be compared to the values in the predicates; table access then need only take place
when it is necessary to do so.
Index skip scan An Oracle term used to represent reading several index slices instead
of doing a full index scan.
Integrity A state of a database in which the defined constraints and rules for the data
are valid.
I/O A request from the processor to read a page from disk or to write a page to disk
following an update.
Join method The optimizer's decision about how to join tables together; the normal
choice is the nested-loop join, although others are available.
Leaf page The lowest level of an index; the pages contain the key and pointer
combinations arranged in key sequence.
Least recently used The algorithm normally used by buffer and disk server cache
managers to identify which pages should be overwritten to satisfy new requests.
List prefetch A facility used by DB2 for z/OS to sort the index pointers of the
required rows into page number sequence, so that the table rows may be accessed
using skip-sequential.
Local response time Of a transaction excluding transfer and wait times between the
work station and the server; the server response time.
Lock A construction for serialization processing to ensure logical integrity; normally
relates to a table, page, or row.
Materializing result rows Performing the database accesses required to build the
result set. In the best case, this simply requires a row to be moved from the database
buffer pool to the program. In the worst case, the DBMS will request a large number
of disk reads.