JDBC and SQLJ connection pooling support

Connection pooling is part of JDBC DataSource support, and is supported by the IBM® Data Server Driver for JDBC and SQLJ.

The IBM Data Server Driver for JDBC and SQLJ provides a factory of pooled connections that are used by WebSphere® Application Server or other application servers. The application server actually does the pooling. Connection pooling is completely transparent to a JDBC or SQLJ application.

Connection pooling is a framework for caching physical data source connections, which are equivalent to data server threads. When JDBC reuses physical data source connections, the expensive operations that are required for the creation and subsequent closing of java.sql.Connection objects are minimized.

Without connection pooling, each java.sql.Connection object represents a physical connection to the data source. When the application establishes a connection to a data server, the data server creates a new physical connection to the data source. When the application calls the java.sql.Connection.close method, the data server terminates the physical connection to the data source.

In contrast, with connection pooling, a java.sql.Connection object is a temporary, logical representation of a physical data source connection. The physical data source connection can be serially reused by logical java.sql.Connection instances. The application can use the logical java.sql.Connection object in exactly the same manner as it uses a java.sql.Connection object when there is no connection pooling support.

With connection pooling, when a JDBC application invokes the DataSource.getConnection method, the data source determines whether an appropriate physical connection exists. If an appropriate physical connection exists, the data source returns a java.sql.Connection instance to the application. When the JDBC application invokes the java.sql.Connection.close method, JDBC does not close the physical data source connection. Instead, JDBC closes only JDBC resources, such as Statement or ResultSet objects. The data source returns the physical connection to the connection pool for reuse.

Connection pooling can be homogeneous or heterogeneous.

With homogeneous pooling, all Connection objects that come from a connection pool should have the same properties. The first logical Connection that is created with the DataSource has the properties that were defined for the DataSource. However, an application can change those properties. When a Connection is returned to the connection pool, an application server or a pooling module should reset the properties to their original values. However, an application server or pooling module might not reset the changed properties. The JDBC driver does not modify the properties. Therefore, depending on the application server or pool module design, a reused logical Connection might have the same properties as those that are defined for the DataSource or different properties.

With heterogeneous pooling, Connection objects with different properties can share the same connection pool.