Due to the widening gap between central processing unit speed and electromechanical disk performance, future virtual memory systems based on newer technologies will use devices such as extended semiconductor memory or solid-state disks as an intermediate stage in the memory hierarchy or as a replacement for secondary storage. These newer devices will have smaller access times relative to the main memory access time. The role of the virtual memory management system is to assign to main memory those pages of a program’s virtual address space that are needed for current execution. To avoid excessively expensive page traffic, a program’s memory allocation should he large enough to contain its most economical locality of reference. Too large a memory allocation, however, will needlessly increase space-time execution cost by allowing pages no longer needed to remain in main memory. The choice of a proper memory allocation depends on 0, the ratio of secondary to primary memory access time. The memory allocation should he adjusted so that the space-time cost of retaining pages in memory is balanced against the space-time cost of removing pages and later transferring them in again. This paper analyzes some of the problems encountered in designing algorithms for managing virtual memory systems based on the newer technology secondary storage devices.
Index Terms: Cyclic locality interval, locality of reference, memory management, program behavior, program loop structure, secondary storage device speed, space-time execution cost.
