Checkpoints

The database server performs two types of checkpoints: full checkpoints (also known as sync checkpoints) and fuzzy checkpoints. The term checkpoint refers to the point in the database server operation when the pages on disk are synchronized with the pages in the shared-memory buffer pool. The default checkpoint type is fuzzy.

The database server generates at least one checkpoint for each span of the logical-log space to guarantee that it has a checkpoint at which to begin fast recovery.

Although the database server performs checkpoints automatically, you can initiate one manually or control how often the database server checks to see if a checkpoint is needed. You can specify the checkpoint interval in the CKPTINTVL configuration parameter.

To reduce the amount of work required at checkpoint, use the BUFFERPOOL configuration parameter to lower the lru_max_dirty and lru_max_dirty values. For instance, if the database server has several instances with a very large buffer pool, set the lru_max_dirty value to less than 1 to reduce the checkpoint time required.

For more information about CKPTINTVL and BUFFERPOOL, see the chapter on configuration parameters in the IBM Informix Administrator's Reference. For information on monitoring and tuning checkpoint parameters, see your IBM Informix Dynamic Server Performance Guide.