TABLE OF CONTENTS Virtual memory is one technique for managing the resource of physical memory, including the caches, main memory, and the disk subsystem. It was originally invented to provide to software the illusion of a very large amount of main memory [Kilburn et al. 1962]. Though administrators today typically purchase enough main memory that nearly all code and data needed by software is actually in main memory (i.e., virtual memory has outlived its original purpose), the mechanism is nonetheless the most popular front-end for the memory system today, in general-purpose systems and, to an increasing degree, embedded systems as well.
The basic functions of virtual memory are well known [Denning 1970]. One is to create a virtual-machine environment for every process, which among other things allows text, data, and stack regions to begin at statically known locations in all processes without fear of Conflict. Another is demand-paging setting a finer granularity for process residence than an entire address space, thereby allowing a process to execute as long as a single page is memory resident. Today's expectations of virtual memory extend its original semantics and now include additional features such as virtual-address aliasing (enabling different processes to map shared objects at different virtual addresses), protection aliasing (enabling different processes to map shared objects using different protections), and support for virtual caches.
In general-purpose systems, the virtual memory system is the primary client of the memory system, in that a user process is not aware of physical memory and does not access it directly.
