Connection pooling and connection concentrator


While connection pooling and connection concentrator seem to have similarities, they differ in their implementation and address different issues. Connection pooling helps reduce the overhead of database connections and handle connection volume. Connection concentrator helps increase the scalability of your DB2® for OS/390® and z/OS® and DB2 Connect™ solution by optimizing the use of your host database servers.

When using connection pooling, the connection is only available for reuse after the application owning the connection issues a disconnect request. In many 2-tier client-server applications users do not disconnect for the duration of the workday. Likewise, most application servers in multi-tier applications establish database connections at server start up time and do not release these connections until the application server is shut down.

In these environments, connection pooling will have little, if any, benefit. However, in web and client-server environments where the frequency of connections and disconnections is higher than connection pooling will produce significant performance benefits. The connection concentrator allocates host database resources only for the duration of an SQL transaction while keeping user applications active. This allows for configurations where the number of DB2 threads and the resources they consume can be much smaller than if every application connection had its own thread.

When it comes to fail-safe operation and load balancing of workload connection concentrator is clearly the right choice as it allows reallocation of work with every new transaction. Alternatively, connection pooling can only offer very limited balancing and only at connect time.

Connection pooling and connection concentrator should be used together although they address different issues.