From: Steve Chapin (sjc@cs.purdue.EDU)
Date: 05/29/92


From: sjc@cs.purdue.EDU (Steve Chapin)
Subject: Re: Linux swapping
Date: 29 May 1992 16:02:07 GMT


}} In article <235kl7h.james@netcom.com> james@netcom.com (James L. Paul) writes:
}}
}} BTW, My understanding is this: Swapping involves management at the
}} process level, and each active process must be fully in physical ram.
}} This limits efficiency, since memory available to be swapped out is
}} limited to that of inactive processes. Paging happens at a lower
}} level, and makes it possible (although slower) to keep only part of
}} an active process in physical ram. Since paging requires memory
}} management hardware to be effective, and Linux needs the 386 for
}} this, I'm guessing that Linux uses paging, not swapping. Do I have
}} this right? (And is demand paging different from paging? How?)

There is a third possibility: paging and swapping. You demand page
each process, and when you notice that your page fault rate is too
high, you reduce the contention by swapping all the in-memory pages
for a process out to the swap device. Later, when the page fault rate
drops, you can swap (at least part) of the process back in.

SunOS does both swapping and paging.

}} James L. Paul

sjc@cs.purdue.edu Steve Chapin Today's Grammar Lesson:
       "If you loose your arrow, you're likely to lose it in the weeds,"
                       was often heard in days of yore.

             If I don't want it in my back yard, I don't want it.