Paging occurs if the total amount of memory of all the running applications exceeds the amount of physical memory. In this situation, the operating system temporarily moves parts of the running applications onto the disk to keep the applications running when memory limits are exceeded. When a paged-out memory location is accessed, the operating system restores that area of memory from disk and then, to make room, moves some other part of memory to disk. Consequently, performance can suffer. Some amount of paging is fine on the client side. But on the server, where many queries must be performed simultaneously, paging must be minimized.

Procedures that are known to require large amounts of memory and exceed the allotted managed memory should be cached and scheduled for refresh at non-peak processing hours (provided the business requirement allows for that). That query should be forced to process on disk to allow other queries to run and execute efficiently for more optimized performance. See the TDV Reference Guide for the use of the FORCE_DISK query engine option.