Data cartridges extend the capabilities of the Oracle Database server by taking advantage of Oracle Extensibility Architecture framework. This framework lets you capture business logic and processes associated with specialized or domain-specific data in user-defined data types. Data cartridges that provide new behavior without using additional attributes can do so with packages rather than user-defined data types. With either user-defined types or packages, you determine how the server interprets, stores, retrieves, and indexes the application data. Data cartridges package this functionality, creating software components that plug into a server and extend its capabilities into a new domain, making the database itself extensible.

You can customize the indexing and query optimization mechanisms of an extensible database management system and provide specialized services or more efficient processing for user-defined business objects and rich types. When you register your implementations with the server through extensibility interfaces, you direct the server to implement your customized processing instructions instead of its own default processes.

External procedures are highly extensible because you can enhance their functionality without affecting their invokers. The reason is that the call specification of an external procedure, which has all the information needed to invoke it, is separate from the body of the procedure, which has the implementation details. If you change only the body, and not the specification, then invokers are unaffected.

User-defined PL/SQL functions that can appear in SQL statements or expressions can extend the functionality of SQL.

User-defined aggregate functions are part of the Oracle Extensibility Architecture framework.

The object relational features of Oracle Database are the user-defined Abstract Data Types (ADTs). ADTs are highly extensible because you can enhance their functionality without affecting their invokers. The reason is that the call specification of an external procedure, which has all the information needed to invoke it, is separate from the body of the procedure, which has the implementation details. If you change only the body, and not the specification, then invokers are unaffected.

A data warehouse is a relational database that is designed for query and analysis rather than for transaction processing. It usually contains historical data derived from transaction data, but can include data from other sources. Data warehouses separate analysis workload from transaction workload and enable an organization to consolidate data from several sources. This strategy helps the organization maintain historical records and analyze the data to better understand and improve its business.

In addition to a relational database, a data warehouse environment can include:

• An extraction, transportation, transformation, and loading (ETL) solution

• Statistical analysis

• Reporting

• Data mining capabilities

• Client analysis tools

• Applications that manage the process of gathering data; transforming it into useful, actionable information; and delivering it to business users

Data warehousing systems typically:

• Use many indexes

• Use some (but not many) joins

• Use denormalized or partially denormalized schemas (such as a star schema) to optimize query and analytical performance

• Use derived data and aggregates

• Have workloads designed to accommodate ad hoc queries and data analysis

  Because you might not know the workload of your data warehouse in advance, you must optimize the data warehouse to perform well for a wide variety of possible query and analytical operations.

• Are updated regularly (nightly or weekly) by the ETL process, using bulk data modification techniques

  The end users of a data warehouse do not directly update the data warehouse except when using analytical tools (such as data mining) to make predictions with associated probabilities, assign customers to market segments, and develop customer profiles.

Online transaction processing (OLTP) systems are optimized for fast and reliable transaction handling. Compared to data warehouse systems, most OLTP interactions involve a relatively small number of rows, but a larger group of tables. In OLTP systems, performance requirements require that historical data be frequently moved to an archive.

OLTP systems typically:

• Use few indexes.

• Use many joins.

• Use fully normalized schemas to optimize update, insert, and delete performance, and to guarantee data consistency.

• Rarely use derived data and aggregates.

• Have workloads consisting of predefined operations.

• Have users routinely issuing individual data modification statements to the database, so that the OLTP database always reflects the current state of each transaction.

A SQL analytic function computes an aggregate value based on a group of rows. A SQL analytic function differs from an aggregate function in that it returns multiple rows for each group. For each row, a window of rows is defined. The window determines the range of rows used to perform the calculations for the current row. Window sizes can be based on either a physical number of rows or a logical interval such as time.

SQL analytic functions bring to set-oriented SQL the ability to use array semantics on result sets. They provide coding efficiency, because they enable concise, straightforward coding of logic that is otherwise cumbersome or impossible. They also provide processing efficiency, because they are integral to Oracle Database and use internal optimizations.

A typical use of analytic functions is to retrieve the most current information in a table. For example, a query of the following form returns information from the row with the most recent update time for each customer with records in a table:

SELECT ... FROM my_table t1
  WHERE upd_time = ( SELECT MAX(UPD _TIME)
                     FROM my_table t2
                     WHERE t2.cust_id = t1.cust_id );

The preceding query uses a correlated subquery to find the MAX(UPD _TIME) by cust _id, record by record. Therefore, the correlated subquery could be evaluated once for each row in the table. If the table has very few records, performance may be adequate; if the table has tens of thousands of records, the cumulative cost of repeatedly executing the correlated subquery is high.

The following query makes a single pass on the table and computes the maximum UPD_TIME during that pass. Depending on various factors, such as table size and number of rows returned, the following query may be much more efficient than the preceding query:

SELECT ...
  FROM ( SELECT t1.*,
         MAX(UPD_TIME) OVER (PARTITION BY cust _id) max_time
         FROM my_table t1
       )
  WHERE  upd_time = max_time;

The available analytic functions are:

AVG
CORR
COUNT
COVAR_POP
COVAR_SAMP
CUME_DIST
DENSE_RANK
FIRST
FIRST_VALUE
LAG
LAST
LAST_VALUE
LEAD
LISTAGG
MAX
MIN
NTH_VALUE
NTILE
PERCENT_RANK
PERCENTILE_CONT
PERCENTILE_DISC
RANK
RATIO_TO_REPORT
REGR_ (Linear Regression) Functions
ROW_NUMBER
STDDEV
STDDEV_POP
STDDEV_SAMP
SUM
VAR_POP
VAR_SAMP
VARIANCE

Materialized views are query results that have been stored ("materialized") as schema objects. Like tables and views, materialized views can appear in the FROM clauses of queries.

Materialized views are used to summarize, compute, replicate, and distribute data. They are useful for pre-answering general classes of questions—users can query the materialized views instead of individually aggregating detail records. Some environments where materialized views are useful are data warehousing, replication, and mobile computing.

Materialized views require time to create and update, and disk space for storage, but these costs are offset by dramatically faster queries. In these respects, materialized views are like indexes, and they are called "the indexes of your data warehouse." Unlike indexes, materialized views can be queried directly (with SELECT statements) and sometimes updated with DML statements (depending on the type of update needed).

A major benefit of creating and maintaining materialized views is the ability to take advantage of query rewrite, which transforms a SQL statement expressed in terms of tables or views into a statement accessing one or more materialized views that are defined on the detail tables. The transformation is transparent to the end user or application, requiring no intervention and no reference to the materialized view in the SQL statement. Because query rewrite is transparent, materialized views can be added or dropped like indexes without invalidating the SQL in the application code.

The following statement creates and populates a materialized aggregate view based on three master tables in the SH sample schema:

CREATE MATERIALIZED VIEW sales_mv AS 
  SELECT t.calendar_year, p.prod_id, SUM(s.amount_sold) AS sum_sales
  FROM   times t, products p, sales s
  WHERE  t.time_id = s.time_id 
  AND    p.prod_id = s.prod_id
  GROUP BY t.calendar_year, p.prod_id;

Partitioning is the database ability to physically break a very large table, index, or materialized view into smaller pieces that it can manage independently. Partitioning is similar to parallel processing, which breaks a large process into smaller pieces that can be processed independently.

Each partition is an independent object with its own name and, optionally, its own storage characteristics. Partitioning is useful for many different types of database applications, particularly those that manage large volumes of data. Benefits include increased availability, easier administration of schema objects, reduced contention for shared resources in OLTP systems, and enhanced query performance in data warehouses.

To partition a table, specify the PARTITION BY clause in the CREATE TABLE statement. SELECT and DML statements do not need special syntax to benefit from the partitioning.

A common strategy is to partition records by date ranges. The following statement creates four partitions, one for records from each of four years of sales data (2008 through 2011):

CREATE TABLE time_range_sales
   ( prod_id        NUMBER(6)
   , cust_id        NUMBER
   , time_id        DATE
   , channel_id     CHAR(1)
   , promo_id       NUMBER(6)
   , quantity_sold  NUMBER(3)
   , amount_sold    NUMBER(10,2)
   )
PARTITION BY RANGE (time_id)
 (PARTITION SALES_2008 VALUES LESS THAN (TO_DATE('01-JAN-2009','DD-MON-YYYY')),
  PARTITION SALES_2009 VALUES LESS THAN (TO_DATE('01-JAN-2010','DD-MON-YYYY')),
  PARTITION SALES_2010 VALUES LESS THAN (TO_DATE('01-JAN-2011','DD-MON-YYYY')),
  PARTITION SALES_2011 VALUES LESS THAN (MAXVALUE)
 ); 

Temporal Validity Support lets you associate one or more valid time dimensions with a table and have data be visible depending on its time-based validity, as determined by the start and end dates or time stamps of the period for which a given record is considered valid. Examples of time-based validity are the hire and termination dates of an employee in a Human Resources application, the effective date of coverage for an insurance policy, and the effective date of a change of address for a customer or client.

Temporal Validity Support is typically used with Oracle Flashback Technology, for queries that specify the valid time period in AS OF and VERSIONS BETWEEN clauses. You can also use the DBMS_FLASHBACK_ARCHIVE.ENABLE_AT_VALID_TIME procedure to specify an option for the visibility of table data: all table data (the default), data valid at a specified time, or currently valid data within the valid time period at the session level.

Temporal Validity Support is useful in Information Lifecycle Management (ILM) and any other application where it is important to know when certain data becomes valid (from the application's perspective) and when it becomes invalid (if ever).

CREATE TABLE my_emp(
  empno NUMBER,
  last_name VARCHAR2(30),
  start_time TIMESTAMP,
  end_time TIMESTAMP,
PERIOD FOR user_valid_time (start_time, end_time));
 
INSERT INTO my_emp VALUES (100, 'Ames', '01-Jan-10', '30-Jun-11');
INSERT INTO my_emp VALUES (101, 'Burton', '01-Jan-11', '30-Jun-11');
INSERT INTO my_emp VALUES (102, 'Chen', '01-Jan-12', null);
 
-- Valid Time Queries --
 
-- AS OF PERIOD FOR queries:
 
-- Returns only Ames.
SELECT * from my_emp AS OF PERIOD FOR user_valid_time TO_TIMESTAMP('01-Jun-10');

-- Returns  Ames and Burton, but not Chen.
SELECT * from my_emp AS OF PERIOD FOR user_valid_time TO_TIMESTAMP('01-Jun-11');

-- Returns no one.
SELECT * from my_emp AS OF PERIOD FOR user_valid_time TO_TIMESTAMP( '01-Jul-11');

-- Returns only Chen.
SELECT * from my_emp AS OF PERIOD FOR user_valid_time TO_TIMESTAMP('01-Feb-12');
 
-- VERSIONS PERIOD FOR ... BETWEEN queries:
 
-- Returns only Ames.
SELECT * from my_emp VERSIONS PERIOD FOR user_valid_time BETWEEN 
   TO_TIMESTAMP('01-Jun-10') AND TO_TIMESTAMP('02-Jun-10');

-- Returns Ames and Burton.
SELECT * from my_emp VERSIONS PERIOD FOR user_valid_time BETWEEN 
   TO_TIMESTAMP('01-Jun-10') AND TO_TIMESTAMP('01-Mar-11');

-- Returns only Chen.
SELECT * from my_emp VERSIONS PERIOD FOR user_valid_time BETWEEN 
   TO_TIMESTAMP('01-Nov-11') AND TO_TIMESTAMP('01-Mar-12');

-- Returns no one.
SELECT * from my_emp VERSIONS PERIOD FOR user_valid_time BETWEEN 
   TO_TIMESTAMP('01-Jul-11') AND TO_TIMESTAMP('01-Sep-11');

ALTER TABLE my_emp ADD PERIOD FOR user_valid_time;

-- Add a record for Davis.
INSERT INTO my_emp VALUES (103, 'Davis', '01-Jan-12', null);

-- To change employee 103's last name to Smith,
-- first set an end date for the record with the old name.
UPDATE my_emp SET end_time = '01-Feb-12' WHERE empno = 103;

-- Then insert another record for employee 103, specifying the new last name,
-- the appropriate valid start date, and null for the valid end date.
-- After the INSERT statement, there are two records for #103 (Davis and Smith).
INSERT INTO my_emp VALUES (103, 'Smith', '02-Feb-12', null);
 
-- What's the valid information for employee 103 as of today?
SELECT * from my_emp AS OF PERIOD FOR user_valid_time SYSDATE WHERE empno = 103;

-- What was the valid information for employee 103 as of a specified date?

-- First, as of a date after the change to Smith.
SELECT * from my_emp AS OF PERIOD FOR user_valid_time TO_TIMESTAMP('01-Jul-12')
  WHERE empno = 103;

-- Next, as of a date before the change to Smith.
SELECT * from my_emp AS OF PERIOD FOR user_valid_time TO_TIMESTAMP('20-Jan-12')
  WHERE empno = 103;