I/O priority queueing is used to control deferred I/O requests. If this function is invoked, all deferred I/O requests, except paging and swapping, will be queued according to the I/O priorities associated with the requesting address spaces. Paging and swapping are always handled at the highest priority. An address space's I/O priority is by default the same as its dispatching priority. All address spaces in one mean-time-to-wait group fall into one I/O priority. In addition, address spaces that are time sliced have their I/O queued at their time slice priority. Changes to an address space's dispatching priority when the address space is time sliced up or down, do not affect the I/O priority.

An installation can assign an I/O priority that is higher or lower than the dispatching priority for selected groups of work. For example, if the installation is satisfied with the dispatching priority of an interactive application but would like the application's I/O requests to be processed before those of other address spaces executing at the same priority, the application could be given an I/O priority higher than its dispatching priority.

If I/O priority queueing is not invoked, all I/O requests are handled in a first-in/first-out (FIFO) manner.