4.4 OPENVMS SWAPPING ALGORITHM

The reduction or lack of free physical memory does not limit the number of processes. When the active processes fill all of physical memory-that is, when the number of entries in the free list becomes small-possible response is to inhibit further process creation and image activation. But the OpenVMS operating system does not do this. Instead, it establishes more room in physical memory with three additional strategies: management of the modified list, reduction of all working sets, and removal of inactive processes from memory. Modified list management is called paging, and removing inactive processes is called outswapping. Conversely, inswapping operations restore processes to memory.

While the paging algorithm is responsible for adjusting the working set limit, it is the SWAPPER process that is responsible for these three, more massive memory management activities. The SWAPPER is unusual within the OpenVMS operating system in that it is a process and most of the rest of the OpenVMS elements are procedures. Because it is a process, its time of execution is under the control of the scheduler. In a SHOW SYSTEM or MONITOR PROCESS display, SWAPPER is always found in the HIB state executing at priority 16, as if it were a real-time process. This elevated priority is necessary to ensure that SWAPPER runs before any time-shared process. Periodic (once-per-second) execution of a OpenVMS procedure within the scheduler, called SCH$SWPWAKE (SWAPPER Wakeup), determines whether or not SWAPPER should be scheduled for execution. SCH$SWPWAKE makes a cursory...