Subject: Re: memory tester shows up swap/page tuning bug
To: Jonathan Stone <jonathan@DSG.Stanford.EDU>
From: Jukka Marin <jmarin@pyy.jmp.fi>
List: current-users
Date: 09/15/1996 10:22:14
> The following shar'ed program allocates and touches large amounts of
> physical memory.  (I wrote it to check for a bad SIMM).  Run it with
> the size, in megabytes, of physical memory on the machine. Do this
> where you can hear swapping activity to the disk. During the
> ``interesting'' behaviour regime, normal tools don't seem able to run,
> at least not for me.

I have run a similar program in my i386 box with 32 MB RAM.  I told the
memory tester to use 24 megs - and the machine locked up for 20 seconds
or so.  The only thing it was doing was responding to ping (couldn't
see the HD LED from here).  Giving 32 megs to the tester would be even
worse (my SS1+ _paniced_ while I was running crack and the swap problem
showed up).

> IMNSO, a VM system should keep a pool of free pageframes at all times,
> to satisfy pagefaults without having to first page something else out.

I don't know how the VM system works, but the way it behaves now is not
really good. ;-)

  -jm