[Info-vax] Raxco VMS Tuning Seminar Notes

[jls]

On Mon, 8 Nov 2010 17:29:13 -0800 (PST), AEF <spamsink2001 at yahoo.com>
wrote:


>>
>> Swapper trimming can be a significantly lower overhead for many
>> circumstances.  
>
>Can you comment on the
>
>3.7 Active Memory Reclamation from Idle Processes
>
>and how it compares to your methods?

"My method", which is really just VMS tuning information gleaned from
the docs, and experience, *is* "Active Memory Reclamation from Idle
Processes." (AMRIP)

The swapper is part of AMRIP.  But, as I've explained numerous times
now, taking pages away from a process requires you to fault them all
back in again when they become active.

Swapping larger processes entirely reduces this paging activity, given
that longwait is adequate enough and that the number of processes
maintains a low probability of inswapping (i.e., low inswap rate).