21 Implicit Connection Caching

Connection caching, generally implemented in the middle tier, is a means of keeping and using the cache of physical database connections.

Note:

The previous cache architecture, based on OracleConnectionCache and OracleConnectionCacheImpl, is deprecated. Oracle recommends that you take advantage of the new architecture, which is more powerful and offers better performance.

The implicit connection cache is an improved Java Database Connectivity (JDBC) 3.0-compliant connection cache implementation for DataSource. Java and Java2 Platform, Enterprise Edition (J2EE) applications benefit from transparent access to the cache, support for multiple users, and the ability to request connections based on user-defined profiles.

An application turns the implicit connection cache on by calling setConnectionCachingEnabled(true) on an OracleDataSource. After implicit caching is turned on, the first connection request to the OracleDataSource transparently creates a connection cache. There is no need for application developers to write their own cache implementations.

This chapter is divided into the following sections:

The Implicit Connection Cache

The connection caching architecture has been redesigned so that caching is transparently integrated into the data source architecture.

The connection cache uses the concept of physical connections and logical connections. Physical connections are the actual connections returned by the database and logical connections are containers used by the cache to manipulate physical connections. You can think of logical connections as handles. The caches always return logical connections, which implement the same interfaces as physical connections.

The implicit connection cache offers the following:

  • Driver independence

    Both the JDBC Thin and JDBC Oracle Call Interface (OCI) drivers support the implicit connection cache.

  • Transparent access to the JDBC connection cache

    After an application turns implicit caching on, it uses the standard OracleDataSource application programming interfaces (APIs) to get connections. With caching enabled, all connection requests are serviced from the connection cache.

    When an application calls the OracleConnection.close method to close the logical connection, the physical connection is returned to the cache.

  • Single cache per OracleDataSource instance

    When connection caching is turned on, each cache-enabled OracleDataSource has exactly one cache associated with it. All connections obtained through that data source, no matter what user name and password are used, are returned to the cache. When an application requests a connection from the data source, the cache either returns an existing connection or creates a new connection with matching authentication information.

    Note:

    Caches cannot be shared between DataSource instances. There is a one-to-one mapping between cache-enabled DataSource instances and caches.
  • Heterogeneous user names and passwords per cache

    Unlike the previous cache implementation, all connections obtained through the same data source are stored in a common cache, no matter what user name and password the connection requests.

  • Support for JDBC 3.0 connection caching, including support for multiple users and the required cache properties

  • Property-based configuration

    Cache properties define the behavior of the cache. The supported properties set timeouts, the number of connections to be held in the cache, and so on. Using these properties, applications can reclaim and reuse abandoned connections. The implicit connection cache supports all the JDBC 3.0 connection cache properties.

  • OracleConnectionCacheManager

    The new OracleConnectionCacheManager class provides a rich set of administrative APIs that applications can use to manage the connection cache. Each virtual machine has one distinguished instance of OracleConnectionCacheManager. Applications manage a cache through the single OracleConnectionCacheManager instance.

  • User-defined connection attributes

    The implicit connection cache supports user-defined connection attributes that can be used to determine which connections are retrieved from the cache. Connection attributes can be thought of as labels whose semantics are defined by the application, not by the caching mechanism.

  • Callback mechanism

    The implicit connection cache provides a mechanism for users to define cache behavior when a connection is returned to the cache, when handling abandoned connections, and when a connection is requested but none is available in the cache.

  • Connect-time load balancing

    Implicit connection caching provides connect-time load balancing when a connection is first created by the application. The database listener distributes connection creation across Oracle Real Application Clusters instances that would perform the best at the time of connection creation.

  • Run-time connection load balancing

    Run-time connection load balancing of work requests uses Service Metrics to route work requests to an Oracle Real Application Clusters instance that offers the best performance. Selecting a connection from the cache based on service, to execute a work request, greatly increases the throughput and scalability.

Using the Connection Cache

This section discusses how applications use the implicit connection cache. It covers the following topics:

Turning Caching On

An application turns the implicit connection cache on by calling OracleDataSource.setConnectionCachingEnabled(true). After implicit caching is turned on, the first connection request to the OracleDataSource class transparently creates a connection cache.

Example 21-1 provides a sample code that uses the implicit connection cache.

Example 21-1 Using the Implicit Connection Cache

// Example to show binding of OracleDataSource to JNDI,
// then using implicit connection cache 
 
import oracle.jdbc.pool.*; // import the pool package
 
Context ctx = new InitialContext(ht);
OracleDataSource ods = new OracleDataSource();
 
// Set DataSource properties
ods.setUser("Scott");
ods.setConnectionCachingEnabled(true);    // Turns on caching
ctx.bind("MyDS", ods);
// ...
// Retrieve DataSource from the InitialContext
ods =(OracleDataSource) ctx. lookup("MyDS");  
 
// Transparently create cache and retrieve connection 
conn = ods.getConnection();
// ...
conn.close();  // return connection to the cache
// ...
ods.close() // close datasource and clean up the cache

Opening a Connection

After you have turned connection caching on, whenever you retrieve a connection through the OracleDataSource.getConnection method, the JDBC drivers check to see if a connection is available in the cache.

The getConnection method checks if there are any free physical connections in the cache that match the specified criteria. If a match is found, then a logical connection is returned, wrapping the physical connection. If no physical connection match is found, then a new physical connection is created, wrapped in a logical connection, and returned.

There are four variations on getConnection, two that make no reference to the connection cache and two that specify which sort of connections the cache may return. The non-cache-specific getConnection methods behave in the standard manner.

Note:

When implicit connection cache is enabled, the connection returned by OracleDataSource.getConnection may not have the state reset. You must, therefore, reset all the connection states, such as auto-commit, batch size, prefetch size, transaction status, and transaction isolation, before putting the connection back into the cache.

Setting Connection Cache Name

The ConnectionCacheName property of OracleDataSource is an optional property used by the Connection Cache Manager to manage a connection cache. You can set this property by calling the following method:

public void synchronized setConnectionCacheName(String cacheName) throws SQLException

When this property is set, the name is used to uniquely identify the cache accessed by the cache-enabled OracleDataSource. If the property is not set, then a default cache name is created using the convention DataSourceName#HexRepresentationOfNumberOfCaches.

Note:

The getConnectionCacheName() method will return the name of the connection cache only if the setConnectionCacheName method is called after the setConnectionCachingEnabled method is called.

Setting Connection Cache Properties

You can fine-tune the behavior of the implicit connection cache using the setConnectionCacheProperties method to set various connection properties.

Note:

  • Before setting the cache-specific properties, you must enable caching on the data source; otherwise the setConnectionCacheProperties method will have no effect.

  • Although these properties govern the behavior of the connection cache, they are set on the data source, and not on the connection or on the cache itself.

Closing a Connection

An application returns a connection to the cache by calling the close method. There are two variants of the close method: one with no arguments and one that takes a Connection object as argument.

Note:

  • Applications must close connections to ensure that the connections are returned to the cache.

  • When implicit connection cache is enabled, you must reset all the connection states, such as auto-commit, batch size, prefetch size, transaction status, and transaction isolation, before putting the connection back into the cache. This ensures that any subsequent connection retrieved from the cache will have its state reset.

Implicit Connection Cache Example

Example 21-2 demonstrates creating a data source, setting its caching and data source properties, retrieving a connection, and closing that connection in order to return it to the cache.

Example 21-2 Connection Cache Example

import java.sql.*;
import javax.sql.*;
import java.util.*;
import javax.naming.*;
import javax.naming.spi.*;
import oracle.jdbc.*;
import oracle.jdbc.pool.*;
 
...
  
    // create a DataSource    
    OracleDataSource ods = new OracleDataSource();
    
    // set cache properties 
    java.util.Properties prop = new java.util.Properties();
    prop.setProperty("MinLimit", "2");    
    prop.setProperty("MaxLimit", "10");    
 
    // set DataSource properties    
    String url = "jdbc:oracle:oci8:@";    
    ods.setURL(url);    
    ods.setUser("hr");    
    ods.setPassword("hr");    
    ods.setConnectionCachingEnabled(true); // be sure set to true    
    ods.setConnectionCacheProperties (prop);    
    ods.setConnectionCacheName("ImplicitCache01"); // this cache's name    
 
    // We need to create a connection to create the cache 
    Connection conn = ds.getConnection(user, pass);
    Statement  stmt  = conn.createStatement();
    ResultSet  rset  = stmt.executeQuery("select user from dual");
    conn.close(); 
     
    ods.close();

Connection Attributes

Each connection obtained from a data source can have user-defined attributes. Attributes are specified by the application developer and are java.lang.Properties name and value pairs.

An application can use connection attributes to supply additional semantics to identify connections. For instance, an application may create an attribute named connection_type and then assign it the value payroll or inventory.

Note:

The semantics of connection attributes are entirely application-defined. The connection cache itself enforces no restrictions on the key or value of connection attributes.

The methods that get and set connection attributes are found on the OracleConnection interface. This section covers the following topics:

Getting Connections

The first connection you retrieve has no attributes. You must set them. After you have set attributes on a connection, you can request the connection by specifying its attribute, using the specialized forms of the getConnection method:

  • getConnection(java.util.Properties cachedConnectionAttributes

    Requests a database connection that matches the specified cachedConnectionAttributes.

    See Also:

    For a discussion on connection attributes, see "Other Properties" .
  • getConnection(java.lang.String user, java.lang.String password, java.util.Properties cachedConnectionAttributes)

    Requests a database connection from the implicit connection cache that matches the specified user, password and cachedConnectionAttributes. If null values are passed for user and password, the DataSource defaults are used.

Attribute Matching Rules

The rules for matching connectionAttributes come in two variations:

  • Basic

    In this variation, the cache is searched to retrieve the connection that matches the attributes. The connection search mechanism as follows:

    1. If an exact match is found, the connection is returned to the caller.

    2. If an exact match is not found and the ClosestConnectionMatch data source property is set, then the connection with the closest match is returned. The closest matched connection is one that has the highest number of the original attributes matched. Note that the closest matched connection may match a subset of the original attributes, but does not have any attributes that are not part of the original list. For example, if the original list of attributes is A, B and C, then a closest match may have A and B set, but never a D.

    3. If none of the existing connections matches, a new connection is returned. The new connection is created using the user name and password set on the DataSource. If getConnection(String, String, java.util.Properties) is called, then the user name and password passed as arguments are used to open the new connection.

  • Advanced

    In this variation, the attributes may be associated with weights. The connection search mechanism is similar to the basic connectionAttributes based search, except that the connections are searched not only based on the connectionAttributes, but also using a set of weights that are associated with the keys on the connectionAttributes. These weights are assigned to the keys as a one-time operation and is supported as a connection cache property, AttributeWeights.

Setting Connection Attributes

An application sets connection attributes using the following:

applyConnectionAttributes(java.util.Properties connAttr)

No validation is done on connAttr. Applying connection attributes is cumulative. Each time you call applyConnectionAttributes, the connAttr attribute you supply is added to those previously in force.

Checking Attributes of a Returned Connection

When an application requests a connection with specified attributes, it is possible that no match will be found in the connection cache. When this happens, the connection cache creates a connection with no attributes and returns it. The connection cache cannot create a connection with the requested attributes, because the Connection Cache manager is ignorant of the semantics of the attributes.

Note:

If the closestConnectionMatch property has been set, then the cache manager looks for close attribute matches rather than exact matches.

For this reason, applications should always check the attributes of a returned connection. To do this, use the getUnMatchedConnectionAttributes method, which returns a list of any attributes that were not matched in retrieving the connection. If the return value of this method is null, you know that you must set all the connection attributes.

Connection Attribute Example

Example 21-3 illustrates using connection attributes.

Example 21-3 Using Connection Attributes

    java.util.Properties connAttr = new java.util.Properties();
    connAttr.setProperty("connection_type", "payroll");
 
    // retrieve connection that matches attributes
    Connection conn = ds.getConnection(connAttr); 
    // Check to see which attributes weren't matched
   unmatchedProp = ((OracleConnection)conn).getUnMatchedConnectionAttributes();
   if ( unmatchedProp != null )
    {
    // apply attributes to the connection
    ((OracleConnection)conn).applyConnectionAttributes(connAttr);  
    }
    // verify whether conn contains property after apply attributes
    connProp = ((OracleConnection)conn).getConnectionAttributes(); 
    listProperties (connProp);

Connection Cache Properties

The connection cache properties govern the characteristics of a connection cache. This section lists the supported connection cache properties. It covers the following topics:

Applications set cache properties in one of the following ways:

  • Using the OracleDataSource method setConnectionCacheProperties

  • When creating a cache using OracleConnectionCacheManager

  • When reinitializing a cache using OracleConnectionCacheManager

Limit Properties

These properties control the size of the cache.

InitialLimit

Sets how many connections are created in the cache when it is created or reinitialized. When this property is set to an integer value greater than 0, creating or reinitializing the cache automatically creates the specified number of connections, filling the cache in advance of need.

Default: 0

MaxLimit

Sets the maximum number of connection instances the cache can hold. The default value is Integer.MAX_VALUE, meaning that there is no limit enforced by the connection cache, so that the number of connections is limited only by the number of database sessions configured for the database.

Default: Integer.MAX_VALUE (no limit)

Note:

If the number of concurrent connections exceeds the maximum number of sessions configured at the database server, then you will get ORA-00018 error. To avoid this error, you must set a value for the MaxLimit property. This value should be less than the value of the SESSIONS parameter configured for the database server.

MaxStatementsLimit

Sets the maximum number of statements that a connection keeps open. When a cache has this property set, reinitializing the cache or closing the data source automatically closes all cursors beyond the specified MaxStatementsLimit.

Default: 0

MinLimit

Sets the minimum number of connections the cache maintains.

Default: 0

Note:

  • Setting the MinLimit property does not initialize the cache to contain the minimum number of connections. To do this, use the InitialLimit property.

  • When InitialLimit is greater than MinLimit, it is possible to have any number of connections specified by InitialLimit up to a value specified by MaxLimit. Therefore, InitialLimit does not depend on MinLimit.

  • Connections can fall below the minimum limit set on the connection pool when JDBC Fast Connection Failover DOWN events are processed. The processing removes affected connections from the pool. MinLimit will be honored as requests to the connection pool increase and the number of connections get past the MinLimit value.

TIMEOUT Properties

These properties control the lifetime of an element in the cache.

InactivityTimeout

Sets the maximum time a physical connection can remain idle in a connection cache. An idle connection is one that is not active and does not have a logical handle associated with it. When InactivityTimeout expires, the underlying physical connection is closed. However, the size of the cache is not allowed to shrink below minLimit, if it has been set.

Default: 0 (no timeout in effect)

TimeToLiveTimeout

Sets the maximum time in seconds that a logical connection can remain open. When TimeToLiveTimeout expires, the logical connection is unconditionally closed, the relevant statement handles are canceled, and the underlying physical connection is returned to the cache for reuse.

Default: 0 (no timeout in effect)

AbandonedConnectionTimeout

Sets the maximum time that a connection can remain unused before the connection is closed and returned to the cache. A connection is considered unused if it has not had SQL database activity.

When AbandonedConnectionTimeout is set, JDBC monitors SQL database activity on each logical connection. For example, when stmt.execute is called on the connection, a heartbeat is registered to convey that this connection is active. The heartbeats are set at each database execution. If a connection has been inactive for the specified amount of time, the underlying connection is reclaimed and returned to the cache for reuse.

Default: 0 (no timeout in effect)

PropertyCheckInterval

Sets the time interval at which the Connection Cache Manager inspects and enforces all specified cache properties. PropertyCheckInterval is set in seconds.

Default: 900 seconds

Other Properties

These properties control miscellaneous cache behavior.

AttributeWeights

AttributeWeights sets the weight for each attribute set on the connection.

ClosestConnectionMatch

ClosestConnectionMatch causes the connection cache to retrieve the connection with the closest approximation to the specified connection attributes.

ConnectionWaitTimeout

Specifies cache behavior when a connection is requested and there are already MaxLimit connections active. If ConnectionWaitTimeout is equal to zero, then each connection request waits for zero seconds, that is, null connection is returned immediately. If ConnectionWaitTimeout is greater than zero, then each connection request waits for the specified number of seconds or until a connection is returned to the cache. If no connection is returned to the cache before the timeout elapses, then the connection request returns null.

Default: zero

LowerThresholdLimit

Sets the lower threshold limit on the cache. The default is 20 percent of the MaxLimit on the connection cache. This property is used whenever a releaseConnection() cache callback method is registered.

ValidateConnection

Setting ValidateConnection to true causes the connection cache to test every connection it retrieves against the underlying database. If a valid connection cannot be retrieved, then an exception is thrown.

Default: false

Connection Property Example

Example 21-4 demonstrates how an application uses connection properties.

Example 21-4 Using Connection Properties

import java.sql.*;
import javax.sql.*;
import java.util.*;
import javax.naming.*;
import javax.naming.spi.*;
import oracle.jdbc.*;
import oracle.jdbc.pool.*;
...
  OracleDataSource ds =  (OracleDataSource) ctx.lookup("...");
  java.util.Properties prop = new java.util.Properties ();
  prop.setProperty("MinLimit", "5");     // the cache size is 5 at least 
  prop.setProperty("MaxLimit", "25");
  prop.setProperty("InitialLimit", "3"); // create 3 connections at startup
  prop.setProperty("InactivityTimeout", "1800");    //  seconds
  prop.setProperty("AbandonedConnectionTimeout", "900");  //  seconds
  prop.setProperty("MaxStatementsLimit", "10");
  prop.setProperty("PropertyCheckInterval", "60"); // seconds

  ds.setConnectionCacheProperties (prop);  // set properties
  Connection conn = ds.getConnection();
  conn.dosomework();
  java.util.Properties propList=ds.getConnectionCacheProperties();  // retrieve

Connection Cache Manager API

OracleConnectionCacheManager provides administrative APIs that the middle tier can use to manage available connection caches. This section provides an example of using the Connection Cache Manager.

Example of ConnectionCacheManager Use

Example 21-5 demonstrates the OracleConnectionCacheManager interface.

Example 21-5 Connection Cache Manager Example

import java.sql.*;
import javax.sql.*;
import java.util.*;
import javax.naming.*;
import javax.naming.spi.*;
import oracle.jdbc.*;
import oracle.jdbc.pool.*;
...
// Get singleton ConnectionCacheManager instance
  OracleConnectionCacheManager occm =
  OracleConnectionCacheManager.getConnectionCacheManagerInstance();
  String cacheName = "foo";  // Look for a specific cache
  // Use Cache Manager to check # of available connections 
  // and active connections
  System.out.println(occm.getNumberOfAvailableConnections(cacheName)
      " connections are available in cache " + cacheName);
 
  System.out.println(occm.getNumberOfActiveConnections(cacheName)
      + " connections are active");
  // Refresh all connections in cache 
  occm.refreshCache(cacheName,
    OracleConnectionCacheManager.REFRESH_ALL_CONNECTIONS);
  // Reinitialize cache, closing all connections
  java.util.Properties newProp = new java.util.Properties();
  newProp.setProperty("MaxLimit", "50");
  occm.reinitializeCache(cacheName, newProp);

Advanced Topics

This section discusses cache functionality that is useful for advanced users, but is not essential to understanding or using the implicit connection cache. This section covers the following topics:

Attribute Weights and Connection Matching

There are two connection cache properties that enable the developer to specify which connections in the connection cache are accepted in response to a getConnection request. When you set the ClosestConnectionMatch property to true, you are telling the Connection Cache Manager to return connections that match only some of the attributes you have specified.

If you do not specify attributeWeights, then the Connection Cache Manager returns the connection that matches the highest number of attributes. If you specify attributeWeights, then you can control the priority the manager uses in matching attributes.

ClosestConnectionMatch

Setting ClosestConnectionMatch to true causes the connection cache to retrieve the connection with the closest approximation to the specified connection attributes. This can be used in combination with AttributeWeights to specify what is considered a closest match.

Default: false

AttributeWeights

Sets the weights for each connectionAttribute. This property is used when ClosestConnectionMatch is set to true to determine which attributes are given highest priority when searching for matches. An attribute with a high weight is given more importance in determining a match than an attribute with a low weight.

AttributeWeights contains a set of Strings representing key-value pairs. Each key/value pair sets the weights for each connectionAttribute for which the user intends to request a connection. Each String is in the format written by the java.util.Properties.Store(OutputStream, String) method, and thus can be read by the java.util.Properties.load(InputStream) method. The Key is a connectionAttribute and the Value is the weight. A weight must be an integer value greater than 0. The default weight is 1.

For example, TRANSACTION_ISOLATION could be assigned a weight of 10 and ROLE a weight of 5. If ClosestConnectionMatch is set to true, when a connectionAttribute based connection request is made on the cache, connections with a matching TRANSACTION_ISOLATION will be favored over connections with a matching ROLE.

Default: No AttributeWeights

Connection Cache Callbacks

The implicit connection cache offers a way for the application to specify callbacks to be called by the connection cache. Callback methods are supported with the OracleConnectionCacheCallback interface. This callback mechanism is useful to take advantage of the special knowledge of the application about particular connections, supplementing the default behavior when handling abandoned connections or when the cache is empty.

OracleConnectionCacheCallback is an interface that must be implemented by the user and registered with OracleConnection. The registration API is as follows:

public void
registerConnectionCacheCallback(
OracleConnectionCacheCallback cbk, Object usrObj,  int cbkflag);

In this interface, cbk is the user implementation of the OracleConnectionCacheCallback interface. The usrObj parameter contains any parameters that the user wants supplied. This user object is passed back, unmodified, when the callback method is called. The cbkflag parameter specifies which callback method should be called. It must be one of the following values:

  • OracleConnection.ABANDONED_CONNECTION_CALLBACK

  • OracleConnection.RELEASE_CONNECTION_CALLBACK

  • OracleConnection.ALL_CALLBACKS

When ALL_CALLBACKS is set, all the connection cache callback methods are called. For example,

// register callback, to invoke all callback methods
((OracleConnection)conn).registerConnectionCacheCallback( new UserConnectionCacheCallback(), 
  new SomeUserObject(), 
OracleConnection.ALL_CALLBACKS);

An application can register a ConnectionCacheCallback on an OracleConnection. When a callback is registered, the connection cache calls the handleAbandonedConnection method of the callback before reclaiming the connection. If the callback returns true, then the connection is reclaimed. If the callback returns false, then the connection remains active.

The UserConnectionCacheCallback interface supports two callback methods to be implemented by the user, releaseConnection and handleAbandonedConnection.

Use Cases for TimeToLiveTimeout and AbandonedConnectionTimeout

The following are the use cases for the TimeToLiveTimeout and AbandonedConnectionTimeout timeout mechanisms when used with implicit connection cache. Note that these timeout mechanisms are applicable to the logical connection when it is retrieved from the connection cache.

  • The application considers the connections completely stateless.

    When the connections are stateless, either of the timeout mechanisms can be used. The connections for which the timeout expires are put back into the connection cache for reuse. These connections are valid for reuse because there is no session state associated with them.

  • The application maintains state on each connection, but has a cleanup routine that can render the connections stateless when they are returned to the connection cache.

    In this case, TimeToLiveTimeout cannot be used. There is no way for the connection cache to ensure that a connection returned to the cache is in a reusable condition.However, AbandonedConnectionTimeout can be used in this scenario, only if OracleConnectionCacheCallback is registered on the connection. The handleAbandonedConnection callback method is implemented by the application and ensures that the necessary cleanup is done. The connection is closed when the timeout processing invokes this callback method. The closing of this connection by the callback method causes the connection to be put back into the connection cache in a state where it is reusable.

    Note:

    Do not to close the connection after calling handleAbandonedConnection method because the connection could be in an invalid state. JDBC internally knows how to reclaim a connection even when it is in an invalid state.
  • The application maintains state on each connection, but has no control over the connection and, therefore, cannot ensure cleaning up of state for reuse of connections by other applications or users.

    The use of either of the timeout mechanisms is not recommended.