Mailinglist Archive: opensuse (3644 mails)

< Previous Next >
Re: [SLE] vitual ram
  • From: Derek Fountain <fountai@xxxxxxxxxxxxxxx>
  • Date: Wed, 6 Mar 2002 12:27:19 +0000
  • Message-id: <200203061227.MAA38396@xxxxxxxxxxxxxxxxxxxxxxxx>
On Tuesday 05 March 2002 8:53 pm, you wrote:
Sorry, but I think it is confusing the issue even further by not using
<snip>

Thanks for the lecture. The guy asked a question from which I inferred he was
from a Windows background, and was therefore confused about the way things
worked in the UNIX world. I tried to straighten him out using the "everything
is a file or a process" philosophy. I'm fully aware that what happens under
the VFS layer is complicated, but on the top of it, it looks pretty simple.

Not to beat a dead horse, but there is a "hard and fast rule" that you
should have a swap size which is at least as big as memory. Let's assume

Nothing hard and fast about that. As I said, I have 1GB physical RAM and
128MB swap space. Works fine.

that for whatever reason one your applications takes up 150 MB in its data
segment and your other apps use up so much that you need to start swapping.
Where is the kernel going to put that 150MB of data if your swap space is
only 128MB? If memory serves me correctly, this can cause the system to
panic.

You don't understand the workings of the VM. In your example, the kernel
won't try to swap out 150MB, it'll swap enough pages from that 150MB in order
to meet the other application's requirements. If memory is exhausted it'll
invoke the kernel's OOM killer, the semantics of which will lead it straight
to the application which is hogging the 150MB.

--
12:16pm up 21:47, 1 user, load average: 0.02, 0.01, 0.00

< Previous Next >
Follow Ups