MacLochlainns Weblog

Oracle Partitioned Tables

Learning Outcomes

• Learn about List Partitioning.
• Learn about Range Partitioning.
• Learn about Hash Partitioning.
• Learn about Composite Partitioning.

Lesson Material

Partitioning is the process of breaking up a data source into a series of data sources. Partitioned tables are faster to access and transact against. Partitioning data becomes necessary as the amount of data grows in any table. It speeds the search to find rows and insert, update, or delete rows.

Oracle Database 21c supports four types of table partitioning: list, range, hash, and composite partitioning.

List Partitioning

A list partition works by identifying a column that contains a value, such as a STATE column in an ADDRESS table. Partitioning clauses follow the list of columns and constraints.

A list partition could use a STATE column, like the following (the complete example is avoided to conserve space, and the three dots represent the balance of partitions not shown):

CREATE TABLE franchise
( franchise_id    NUMBER CONSTRAINT pk_franchise PRIMARY KEY
, franchise_name  VARCHAR(20)
, city            VARCHAR(20)
, state           VARCHAR(20))
PARTITION BY LIST(state)
( PARTITION offshore VALUES('Alaska', 'Hawaii')
, PARTITION west VALUES('California', 'Oregon', 'Washington')
, PARTITION desert VALUES ('Arizona','New Mexico')
, PARTITION rockies VALUES ('Colorado', 'Idaho', 'Montana', 'Wyoming')
, ... );

This can be used with other values such as ZIP codes with great effect, but the maintenance of list partitioning can be considered costly. Cost occurs when the list of values changes over time. Infrequent change means low cost, while frequent change means high costs. In the latter case, you should consider other partitioning strategies. Although an Oracle database supports partitioning on a variable-length string, MySQL performs list partitioning only on integer columns.

Range Partitioning

Range partitioning is very helpful on any column that contains a continuous metric, such as dates or time. It works by stating a minimum set that is less than a certain value, and then a group of sets of higher values until you reach the top most set of values. This type of partition helps you improve performance by letting you search ranges rather than complete data sets. Range partitioning is also available in MySQL.

A range example based on dates could look like this:

PARTITION BY RANGE(rental_date)
( PARTITION rental_jan2011
  VALUES LESS THAN TO_DATE('31-JAN-11','DD-MON-YY')
, PARTITION rental_feb2011
  VALUES LESS THAN TO_DATE('28-FEB-11','DD-MON-YY')
, PARTITION rental_mar2011
  VALUES LESS THAN TO_DATE('31-MAR-11','DD-MON-YY')
, ... );

The problem with this type of partitioning, however, is that the new months require constant management. Many North American businesses simply add partitions for all months in the year as an annual maintenance task during the holidays in November or December. Companies that opt for bigger range increments reap search and access benefits from range partitioning, while minimizing ongoing maintenance expenses.

Hash Partitioning

Hash partitioning is much easier to implement than list or range partitioning. Many DBAs favor it because it avoids the manual maintenance of list and range partitioning. Oracle Database 21c documentation recommends that you implement a hash for the following reasons:

• There is no concrete knowledge about how much data maps to a partitioning range.
• The sizes of partitions are unknown at the outset and difficult to balance as data is added to the database.
• A range partition might cluster data in an ineffective way.

This next statement creates eight partitions and stores them respectively in one of the eight tablespaces. The hash partition manages nodes and attempts to balance the distribution of rows across the nodes.

PARTITION BY HASH(store)
PARTITIONS 8
STORE IN (tablespace1, tablespace2, tablespace3, tablespace4
         ,tablespace5, tablespace6, tablespace7, tablespace8);

As you can imagine the maintenance for this type of partitioning is low. Some DBAs choose this method to get an initial sizing before adopting a list or range partitioning plan. Maximizing the physical resources of the machine ultimately rests with the DBAs who manage the system. Developers need to stand ready to assist DBAs with analysis and syntax support.

Composite Partitioning

Composite partitioning requires a partition and subpartition. The composites are combinations of two types of partitioning—typically, list and range partitioning, or range and hash composite partitioning. Which of these you should choose depends on a few considerations. List and range composite partitioning is done for historical information and is well suited for data warehouses. This method lets you partition on unordered or unrelated column values.

A composite partition like this uses the range as the partition and the list as the subpartition, like the following:

PARTITION BY RANGE (rental_date)
 SUBPARTITION BY LIST (state)
 (PARTITION FQ1_1999 VALUES LESS THAN (TO_DATE('1-APR-2011','DD-MON-YYYY'))
  (SUBPARTITION offshore VALUES('Alaska', 'Hawaii')
  , SUBPARTITION west VALUES('California', 'Oregon', 'Washington')
  , SUBPARTITION desert VALUES ('Arizona','New Mexico')
  , SUBPARTITION rockies VALUES ('Colorado', 'Idaho', 'Montana', 'Wyoming')
  , ... )
,(PARTITION FQ2_1999 VALUES LESS THAN (TO_DATE('1-APR-2011','DD-MON-YYYY'))
  (SUBPARTITION offshore VALUES('Alaska', 'Hawaii')
  , SUBPARTITION west VALUES('California', 'Oregon', 'Washington')
  , SUBPARTITION desert VALUES ('Arizona','New Mexico')
  , SUBPARTITION rockies VALUES ('Colorado', 'Idaho', 'Montana', 'Wyoming')
  , ... )
, ... )

Range and hash composite partitioning is done for historical information when you also need to stripe data. Striping is the process of creating an attribute in a table that acts as a natural subtype or separator of data. Users typically view data sets of one subtype, which means organizing the data by stripes (subtypes) can speed access based on user access patterns.

Range is typically the partition and the hash is the subpartition in this composite partitioning schema. The syntax for this type of partition is shown next:

PARTITION BY RANGE (rental_date)
 SUBPARTITION BY HASH(store)
  SUBPARTITIONS 8 STORE IN (tablespace1, tablespace2, tablespace3
                           ,tablespace4, tablespace5, tablespace6
                           ,tablespace7, tablespace8)
   ( PARTITION rental_jan2011
     VALUES LESS THAN TO_DATE('31-JAN-11','DD-MON-YY')
   , PARTITION rental_feb2011
     VALUES LESS THAN TO_DATE('28-FEB-11','DD-MON-YY')
   , PARTITION rental_mar2011
     VALUES LESS THAN TO_DATE('31-MAR-11','DD-MON-YY')
   , ... )