In deployments with exceptionally heavy utilization of the poll() mechanism, you can add the ability to skip calls to poll() with the startup parameter -prefetchpriority. This parameter takes an integer value that represents the number of records to add to the network message of an in-process prefetch query without additional polling. The poll skip is off by default, and only engaged when the previous poll() call returns with nothing waiting.

In some circumstances, the poll() call actually takes longer than the addition of a record to the message in the ready queue. Adding more than one record to the message at a time can greatly reduce the number of poll() calls being made on large deployments. The expected outcome is that skipping calls to poll() greatly improves overall system performance due to a decrease in "System CPU" cycles, allowing for increased "User CPU" cycles.