Tuning data source - Connection pool tuning


You can tune the Connection pool from WAS Admin Console


  • Maximum connections Specifies the maximum number of physical connections that can be created in this pool. These are the physical connections to the backend datastore. When this number is reached, no new physical connections are created; requestors must wait until a physical connection that is currently in use is returned to the pool. For optimal performance, set the value for the connection pool lower than the value for the Web container threadpool size. Lower settings, such as 10 to 30 connections, might perform better than higher settings, such as 100

  • Minimum Connections: Specifies the minimum number of physical connections to maintain. Until this number is exceeded, the pool maintenance thread does not discard physical connections. If you set this property for a higher number of connections than your application ultimately uses at run time, you do not waste application resources. WebSphere Application Server does not create additional connections to achieve your minimum setting. Of course, if your application requires more connections than the value you set for this property, application performance diminishes as connection requests wait for fulfillment.

  • Connection Timeout : Specifies the interval, in seconds, after which a connection request times out and a ConnectionWaitTimeoutException is thrown.

    This value indicates the number of seconds a request for a connection waits when there are no connections available in the free pool and no new connections can be created, usually because the maximum value of connections in the particular connection pool has been reached. For example, if Connection Timeout is set to 300, and the maximum number of connections are all in use, the pool manager waits for 300 seconds for a physical connection to become available. If a physical connection is not available within this time, the pool manager initiates a ConnectionWaitTimeout exception. It usually does not make sense to retry the getConnection() method; if a longer wait time is required you should increase the Connection Timeout setting value. If a ConnectionWaitTimeout exception is caught by the application, the administrator should review the expected connection pool usage of the application and tune the connection pool and database accordingly.
    If the Connection Timeout is set to 0, the pool manager waits as long as necessary until a connection becomes available. This happens when the application completes a transaction and returns a connection to the pool, or when the number of connections falls below the value of Maximum Connections, allowing a new physical connection to be created.
    If Maximum Connections is set to 0, which enables an infinite number of physical connections, then the Connection Timeout value is ignored.

  • Reap Time Specifies the interval, in seconds, between runs of the pool maintenance thread.

    For example, if Reap Time is set to 60, the pool maintenance thread runs every 60 seconds. The Reap Time interval affects the accuracy of the Unused Timeout and Aged Timeout settings. The smaller the interval, the greater the accuracy. If the pool maintenance thread is enabled, set the Reap Time value less than the values of Unused Timeout and Aged Timeout. When the pool maintenance thread runs, it discards any connections remaining unused for longer than the time value specified in Unused Timeout, until it reaches the number of connections specified in Minimum Connections. The pool maintenance thread also discards any connections that remain active longer than the time value specified in Aged Timeout.

    The Reap Time interval also affects performance. Smaller intervals mean that the pool maintenance thread runs more often and degrades performance.

    To disable the pool maintenance thread set Reap Time to 0, or set both Unused Timeout and Aged Timeout to 0. The recommended way to disable the pool maintenance thread is to set Reap Time to 0, in which case Unused Timeout and Aged Timeout are ignored. However, if Unused Timeout and Aged Timeout are set to 0, the pool maintenance thread runs, but only physical connections which timeout due to non-zero timeout values are discarded.

  • Unused Timeout: Specifies the interval in seconds after which an unused or idle connection is discarded.

    Set the Unused Timeout value higher than the Reap Timeout value for optimal performance. Unused physical connections are only discarded if the current number of connections exceeds the Minimum Connections setting. For example, if the unused timeout value is set to 120, and the pool maintenance thread is enabled (Reap Time is not 0), any physical connection that remains unused for two minutes is discarded

  • Aged TimeoutSpecifies the interval in seconds before a physical connection is discarded.

    Setting Aged Timeout to 0 supports active physical connections remaining in the pool indefinitely. Set the Aged Timeout value higher than the Reap Timeout value for optimal performance. For example, if the Aged Timeout value is set to 1200, and the Reap Time value is not 0, any physical connection that remains in existence for 1200 seconds (20 minutes) is discarded from the pool. The only exception is if the connection is involved in a transaction when the aged timeout is reached. If it is the connection is closed immediately after the transaction completes.

  • Purge Policy: Specifies how to purge connections when a stale connection or fatal connection error is detected

    • EntirePool:All connections in the pool are marked stale. Any connection not in use is immediately closed. A connection in use is closed and issues a stale connection Exception during the next operation on that connection. Subsequent getConnection() requests from the application result in new connections to the database opening. When using this purge policy, there is a slight possibility that some connections in the pool are closed unnecessarily when they are not stale. However, this is a rare occurrence. In most cases, a purge policy of EntirePool is the best choice.

    • FailingConnectionOnly:Only the connection that caused the stale connection exception is closed. Although this setting eliminates the possibility that valid connections are closed unnecessarily, it makes recovery from an application perspective more complicated. Because only the currently failing connection is closed, there is a good possibility that the next getConnection() request from the application can return a connection from the pool that is also stale, resulting in more stale connection exceptions.
      The connection pretest function attempts to insulate an application from pooled connections that are not valid. When a backend resource, such as a database, goes down, pooled connections that are not valid might exist in the free pool. This is especially true when the purge policy is failingConnectionOnly; in this case, the failing connection is removed from the pool. Depending on the failure, the remaining connections in the pool might not be valid.



1 comment:

Alex moner said...

Lower settings, such as 10 to 30 connections, might perform better than higher settings, such as 100pool plastering nj