Transaction Processing in a Distributed System

A transaction is a logical unit of work constituted by one or more SQL statements executed by a single user. A transaction begins with the user's first executable SQL statement and ends when it is committed or rolled back by that user.

A remote transaction contains only statements that access a single remote node. A distributed transaction contains statements that access more than one node.

The following sections define important concepts in transaction processing and explain how transactions access data in a distributed database:

Remote SQL Statements
Distributed SQL Statements
Shared SQL for Remote and Distributed Statements
Remote Transactions
Distributed Transactions
Two-Phase Commit Mechanism
Database Link Name Resolution
Schema Object Name Resolution

Remote SQL Statements

A remote query statement is a query that selects information from one or more remote tables, all of which reside at the same remote node. For example, the following query accesses data from the dept table in the scott schema of the remote sales database:

SELECT * FROM scott.dept@sales.us.americas.acme_auto.com;

A remote update statement is an update that modifies data in one or more tables, all of which are located at the same remote node. For example, the following query updates the dept table in the scott schema of the remote sales database:

UPDATE scott.dept@mktng.us.americas.acme_auto.com
  SET loc = 'NEW YORK'
  WHERE deptno = 10;

Note:

A remote update can include a subquery that retrieves data from one or more remote nodes, but because the update happens at only a single remote node, the statement is classified as a remote update.

Distributed SQL Statements

A distributed query statement retrieves information from two or more nodes. For example, the following query accesses data from the local database as well as the remote sales database:

SELECT ename, dname
  FROM scott.emp e, scott.dept@sales.us.americas.acme_auto.com d
  WHERE e.deptno = d.deptno;

A distributed update statement modifies data on two or more nodes. A distributed update is possible using a PL/SQL subprogram unit such as a procedure or trigger that includes two or more remote updates that access data on different nodes. For example, the following PL/SQL program unit updates tables on the local database and the remote sales database:

BEGIN
  UPDATE scott.dept@sales.us.americas.acme_auto.com
    SET loc = 'NEW YORK'
    WHERE deptno = 10;
  UPDATE scott.emp
    SET deptno = 11
    WHERE deptno = 10;
END;
COMMIT;

The database sends statements in the program to the remote nodes, and their execution succeeds or fails as a unit.

Shared SQL for Remote and Distributed Statements

The mechanics of a remote or distributed statement using shared SQL are essentially the same as those of a local statement. The SQL text must match, and the referenced objects must match. If available, shared SQL areas can be used for the local and remote handling of any statement or decomposed query.

Remote Transactions

A remote transaction contains one or more remote statements, all of which reference a single remote node. For example, the following transaction contains two statements, each of which accesses the remote sales database:

UPDATE scott.dept@sales.us.americas.acme_auto.com
  SET loc = 'NEW YORK'
  WHERE deptno = 10;
UPDATE scott.emp@sales.us.americas.acme_auto.com
  SET deptno = 11
  WHERE deptno = 10;
COMMIT;

Distributed Transactions

A distributed transaction is a transaction that includes one or more statements that, individually or as a group, update data on two or more distinct nodes of a distributed database. For example, this transaction updates the local database and the remote sales database:

UPDATE scott.dept@sales.us.americas.acme_auto.com
  SET loc = 'NEW YORK'
  WHERE deptno = 10;
UPDATE scott.emp
  SET deptno = 11
  WHERE deptno = 10;
COMMIT;

Note:

If all statements of a transaction reference only a single remote node, the transaction is remote, not distributed.

Two-Phase Commit Mechanism

A database must guarantee that all statements in a transaction, distributed or non-distributed, either commit or roll back as a unit. The effects of an ongoing transaction should be invisible to all other transactions at all nodes; this transparency should be true for transactions that include any type of operation, including queries, updates, or remote procedure calls.

The general mechanisms of transaction control in a non-distributed database are discussed in the Oracle Database Concepts. In a distributed database, the database must coordinate transaction control with the same characteristics over a network and maintain data consistency, even if a network or system failure occurs.

The database two-phase commit mechanism guarantees that all database servers participating in a distributed transaction either all commit or all roll back the statements in the transaction. A two-phase commit mechanism also protects implicit DML operations performed by integrity constraints, remote procedure calls, and triggers.

Database Link Name Resolution

A global object name is an object specified using a database link. The essential components of a global object name are:

Object name
Database name
Domain

The following table shows the components of an explicitly specified global database object name:

Statement	Object	Database	Domain
`SELECT * FROM joan.dept@sales.acme.com`	`dept`	`sales`	`acme.com`
`SELECT * FROM emp@mktg.us.acme.com`	`emp`	`mktg`	`us.acme.com`

Whenever a SQL statement includes a reference to a global object name, the database searches for a database link with a name that matches the database name specified in the global object name. For example, if you issue the following statement:

SELECT * FROM scott.emp@orders.us.acme.com;

The database searches for a database link called orders.us.acme.com. The database performs this operation to determine the path to the specified remote database.

The database always searches for matching database links in the following order:

Private database links in the schema of the user who issued the SQL statement.
Public database links in the local database.
Global database links (only if a directory server is available).

Name Resolution When the Global Database Name Is Complete

Assume that you issue the following SQL statement, which specifies a complete global database name:

SELECT * FROM emp@prod1.us.oracle.com;

In this case, both the database name (prod1) and domain components (us.oracle.com) are specified, so the database searches for private, public, and global database links. The database searches only for links that match the specified global database name.

Name Resolution When the Global Database Name Is Partial

If any part of the domain is specified, the database assumes that a complete global database name is specified. If a SQL statement specifies a partial global database name (that is, only the database component is specified), the database appends the value in the DB_DOMAIN initialization parameter to the value in the DB_NAME initialization parameter to construct a complete name. For example, assume you issue the following statements:

CONNECT scott@locdb
SELECT * FROM scott.emp@orders;

If the network domain for locdb is us.acme.com, then the database appends this domain to orders to construct the complete global database name of orders.us.acme.com. The database searches for database links that match only the constructed global name. If a matching link is not found, the database returns an error and the SQL statement cannot execute.

Name Resolution When No Global Database Name Is Specified

If a global object name references an object in the local database and a database link name is not specified using the @ symbol, then the database automatically detects that the object is local and does not search for or use database links to resolve the object reference. For example, assume that you issue the following statements:

CONNECT scott@locdb
SELECT * from scott.emp;

Because the second statement does not specify a global database name using a database link connect string, the database does not search for database links.

Terminating the Search for Name Resolution

The database does not necessarily stop searching for matching database links when it finds the first match. The database must search for matching private, public, and network database links until it determines a complete path to the remote database (both a remote account and service name).

The first match determines the remote schema as illustrated in the following table:

User Operation	Database Response	Example
Do not specify the `CONNECT` clause	Uses a connected user database link	`CREATE DATABASE LINK k1 USING 'prod'`
Do specify the `CONNECT TO ... IDENTIFIED BY` clause	Uses a fixed user database link	`CREATE DATABASE LINK k2 CONNECT TO scott IDENTIFIED BY tiger USING 'prod'`
Specify the `CONNECT TO CURRENT_USER` clause	Uses a current user database link	`CREATE DATABASE LINK k3 CONNECT TO CURRENT_USER USING 'prod'`
Do not specify the `USING` clause	Searches until it finds a link specifying a database string. If matching database links are found and a string is never identified, the database returns an error.	`CREATE DATABASE LINK k4 CONNECT TO CURRENT_USER`

After the database determines a complete path, it creates a remote session, assuming that an identical connection is not already open on behalf of the same local session. If a session already exists, the database reuses it.

Schema Object Name Resolution

After the local Oracle Database connects to the specified remote database on behalf of the local user that issued the SQL statement, object resolution continues as if the remote user had issued the associated SQL statement. The first match determines the remote schema according to the following rules:

Type of Link Specified	Location of Object Resolution
A fixed user database link	Schema specified in the link creation statement
A connected user database link	Connected user's remote schema
A current user database link	Current user's schema

If the database cannot find the object, then it checks public objects of the remote database. If it cannot resolve the object, then the established remote session remains but the SQL statement cannot execute and returns an error.

The following are examples of global object name resolution in a distributed database system. For all the following examples, assume that:

Example of Global Object Name Resolution: Complete Object Name

This example illustrates how the database resolves a complete global object name and determines the appropriate path to the remote database using both a private and public database link. For this example, assume the following:

The remote database is named sales.division3.acme.com.
The local database is named hq.division3.acme.com.
A directory server (and therefore, global database links) is not available.
A remote table emp is contained in the schema tsmith.

Consider the following statements issued by scott at the local database:

CONNECT scott@hq

CREATE PUBLIC DATABASE LINK sales.division3.acme.com 
CONNECT TO guest IDENTIFIED BY network 
  USING 'dbstring';

Later, JWARD connects and issues the following statements:

CONNECT jward@hq

CREATE DATABASE LINK sales.division3.acme.com 
  CONNECT TO tsmith IDENTIFIED BY radio; 

UPDATE tsmith.emp@sales.division3.acme.com 
  SET deptno = 40 
  WHERE deptno = 10;

The database processes the final statement as follows:

The database determines that a complete global object name is referenced in jward's UPDATE statement. Therefore, the system begins searching in the local database for a database link with a matching name.
The database finds a matching private database link in the schema jward. Nevertheless, the private database link jward.sales.division3.acme.com does not indicate a complete path to the remote sales database, only a remote account. Therefore, the database now searches for a matching public database link.
The database finds the public database link in scott's schema. From this public database link, the database takes the service name dbstring.
Combined with the remote account taken from the matching private fixed user database link, the database determines a complete path and proceeds to establish a connection to the remote sales database as user tsmith/radio.
The remote database can now resolve the object reference to the emp table. The database searches in the tsmith schema and finds the referenced emp table.
The remote database completes the execution of the statement and returns the results to the local database.

Example of Global Object Name Resolution: Partial Object Name

This example illustrates how the database resolves a partial global object name and determines the appropriate path to the remote database using both a private and public database link.

For this example, assume that:

The remote database is named sales.division3.acme.com.
The local database is named hq.division3.acme.com.
A directory server (and therefore, global database links) is not available.
A table emp on the remote database sales is contained in the schema tsmith, but not in schema scott.
A public synonym named emp resides at remote database sales and points to tsmith.emp in the remote database sales.
The public database link in "Example of Global Object Name Resolution: Complete Object Name" is already created on local database hq:
```
CREATE PUBLIC DATABASE LINK sales.division3.acme.com 
  CONNECT TO guest IDENTIFIED BY network 
  USING 'dbstring'; 
```

Consider the following statements issued at local database hq:

CONNECT scott@hq

CREATE DATABASE LINK sales.division3.acme.com;

DELETE FROM emp@sales 
  WHERE empno = 4299;

The database processes the final DELETE statement as follows:

The database notices that a partial global object name is referenced in scott's DELETE statement. It expands it to a complete global object name using the domain of the local database as follows:
```
DELETE FROM emp@sales.division3.acme.com 
  WHERE empno = 4299; 
```
The database searches the local database for a database link with a matching name.
The database finds a matching private connected user link in the schema scott, but the private database link indicates no path at all. The database uses the connected username/password as the remote account portion of the path and then searches for and finds a matching public database link:
```
CREATE PUBLIC DATABASE LINK sales.division3.acme.com 
  CONNECT TO guest IDENTIFIED BY network 
  USING 'dbstring'; 
```
The database takes the database net service name dbstring from the public database link. At this point, the database has determined a complete path.
The database connects to the remote database as scott/tiger and searches for and does not find an object named emp in the schema scott.
The remote database searches for a public synonym named emp and finds it.
The remote database executes the statement and returns the results to the local database.

Global Name Resolution in Views, Synonyms, and Procedures

A view, synonym, or PL/SQL program unit (for example, a procedure, function, or trigger) can reference a remote schema object by its global object name. If the global object name is complete, then the database stores the definition of the object without expanding the global object name. If the name is partial, however, the database expands the name using the domain of the local database name.

The following table explains when the database completes the expansion of a partial global object name for views, synonyms, and program units:

User Operation	Database Response
Create a view	Does not expand partial global names. The data dictionary stores the exact text of the defining query. Instead, the database expands a partial global object name each time a statement that uses the view is parsed.
Create a synonym	Expands partial global names. The definition of the synonym stored in the data dictionary includes the expanded global object name.
Compile a program unit	Expands partial global names.

What Happens When Global Names Change

Global name changes can affect views, synonyms, and procedures that reference remote data using partial global object names. If the global name of the referenced database changes, views and procedures may try to reference a nonexistent or incorrect database. On the other hand, synonyms do not expand database link names at runtime, so they do not change.

Scenarios for Global Name Changes

For example, consider two databases named sales.uk.acme.com and hq.uk.acme.com. Also, assume that the sales database contains the following view and synonym:

CREATE VIEW employee_names AS 
        SELECT ename FROM scott.emp@hr; 

CREATE SYNONYM employee FOR scott.emp@hr;

The database expands the employee synonym definition and stores it as:

scott.emp@hr.uk.acme.com

Scenario 1: Both Databases Change Names

First, consider the situation where both the Sales and Human Resources departments are relocated to the United States. Consequently, the corresponding global database names are both changed as follows:

sales.uk.acme.com becomes sales.us.acme.com
hq.uk.acme.com becomes hq.us.acme.com

The following table describes query expansion before and after the change in global names:

Query on `sales`	Expansion Before Change	Expansion After Change
`SELECT * FROM employee_names`	`SELECT * FROM scott.emp@hr.uk.acme.com`	`SELECT * FROM scott.emp@hr.us.acme.com`
`SELECT * FROM employee`	`SELECT * FROM scott.emp@hr.uk.acme.com`	`SELECT * FROM scott.emp@hr.uk.acme.com`

Scenario 2: One Database Changes Names

Now consider that only the Sales department is moved to the United States; Human Resources remains in the UK. Consequently, the corresponding global database names are both changed as follows:

sales.uk.acme.com becomes sales.us.acme.com
hq.uk.acme.com is not changed

The following table describes query expansion before and after the change in global names:

Query on `sales`	Expansion Before Change	Expansion After Change
`SELECT * FROM employee_names`	`SELECT * FROM scott.emp@hr.uk.acme.com`	`SELECT * FROM scott.emp@hr.us.acme.com`
`SELECT * FROM employee`	`SELECT * FROM scott.emp@hr.uk.acme.com`	`SELECT * FROM scott.emp@hr.uk.acme.com`

In this case, the defining query of the employee_names view expands to a nonexistent global database name. On the other hand, the employee synonym continues to reference the correct database, hq.uk.acme.com.