Glossary - Relational Database Index Design and the Optimizers - page 300

Merge scan A join method whereby one or more tables are sorted into a consistent
order if necessary (after local predicates have been applied), and the qualifying rows
in the tables or work files are merged (Oracle: sort-merge join ).
Mirroring Writing all pages on two drives.
Multiblock I/O Oracle sequential read.
Multiple index access The collection and comparison of pointers from several
indexes or indeed from several slices of a single index, followed by the access of the
required table rows. Also called index ANDing (index intersection )and index ORing
(index union ).
Multiple serial read-ahead SQL server sequential read.
Multirow FETCH An SQL call used in cursor operations to request multiple rows at
a time.
Nested loop A join method whereby the DBMS first finds a row in the outer table
that satisfies the local predicates referring to that table. Then it looks for the related
rows in the next table, the inner table, and checks which of these satisfy their local
predicates and so on.
Nonleaf page Index pages other than leaf pages that contain a (possibly truncated)
key value, the highest key together with a pointer, to a page at the next lower level.
Null An empty or unknown value; when storing the table row for which no value has
been provided for a column, the DBMS stores a special indicator for null.
Optimizer A component of a relational database management system, which chooses
the access path for each SQL statement. It estimates the cost of feasible access paths,
usually based on a weighted sum of I/O time and CPU time.
Page Index and table rows are grouped together in pages (Oracle uses the term
block ); these are often 4 K in size, but other page sizes may be used. The page size
will determine the number of index and table rows in each page. An entire page will
be read from disk into a buffer pool and so several rows are read with a single I/O.
Predicate A search argument in the WHERE clause of an SQL statement.
Primary key A column or columns that uniquely identify a table row.
Query A request expressed in SQL that provides the rows satisfying the
search arguments.
RAID 5 A redundant array of inexpensive disks level 5—a commonly used way to
store data—logical volumes are striped over several disk drives that form a RAID
array; the first 32 K stripe, for instance, is written to disk drive 1, the second to drive
2, and so on.
RAID 10 Appropriate for databases with frequent random inserts, updates, or deletes;
actually RAID 0 mirroring + RAID 1 striping. Instead of redundant parity data, an
updated page is written to two disk drives; a page can be read from either drive. Disk
load (drive busy) caused by random writes is lower than with RAID 5 but more disk
drives are needed.
Read cache An area in the semiconductor memory (RAM) of the disk server used to
store the most recently read pages from the disk drive. The objective is to reduce the
number of reads from the disk drives. Often much larger than the database buffer