The effect of look-ahead paging in a virtual memory system as determined by simulation

  • Authors:
  • Leon R. Wechsler

  • Affiliations:
  • -

  • Venue:
  • ANSS '73 Proceedings of the 1st symposium on Simulation of computer systems
  • Year:
  • 1973

Quantified Score

Hi-index 0.00

Visualization

Abstract

The virtual memory concept, introduced in 1961 with the Atlas computer, has gained widespread acceptance in recent years, and has been designed into a number of 3rd and 4th generation large processors. Basically, the concept distinguishes between name space (program-generated addresses) and physical space (memory location addresses) with the computer performing the function of mapping name space into physical space. Program pages are moved from auxiliary memory to main memory as required for execution, and physical space assignments are made in accordance with a set of allocation algorithms. In particular, the ability of the virtual memory to realize automatic storage allocation has made it well suited for time sharing computer applications. With limited computer resources, the user is given the illusion that a very large main memory is at his disposal by the dynamic allocation of memory space. For each user's time slice, control of the associated working set of pages is accomplished with demand paging. Immediately preceding the start of a time slice, the technique is to load only that page containing the next instruction, and thereafter load additional pages of the working set those necessary to start processing as demanded during the time slice. This approach is reasonably effective providing that the program working sets change membership slowly and that the computer systems incorporate a high speed access auxiliary memory like Large Capacity Storage (LCS) to store latest program working sets.