Here is the problem: We have an application that uses about 2 or 3
times physical memory, and hence pages heavily. However, the Digital
Unix pager is paging far more than is necessary, causing the
application to run for days and only get a few hours of cpu time.
What we believe is happening is that when the application accesses a
page of memory, the pager pages in 10 or 20 pages (and hence pages out
10 or 20 pages of memory). But these pages that were just paged out
are soon needed, while the extra 9 to 19 pages that were paged in are
not required. So immediately the pager is called again to pull back
in some of the 10 to 20 that were just paged out. But then the pager
pages in more than are necessary again...
I want to be able to specify how many pages are brought in when there
is a page fault (Dick Seymour is going to snear that I should run VMS
and set the page fault cluster size). Does anyone know of a way to
set this in OSF?
Thanks a lot,
David
(Please reply directly to me to avoid excess list traffic.)
------------------------------------------------------------------------------
David Gordon gordon_at_mbt.washington.edu
Dept. of Molecular Biotechnology (206) 685-4465
Box 352145 FAX: (206) 616-5287
University of Washington
Seattle, WA 98195
------------------------------------------------------------------------------
Received on Tue Nov 21 1995 - 02:28:38 NZDT