Richard Fish wrote:
On 2/23/06, Dave Nebinger <[EMAIL PROTECTED]> wrote:
This is never true.  Swap is *always* called for, and for a good reason.

No, it isn't.  For my single-user laptop with 2G of RAM, I actually
prefer that the OOM kill any runaway process that is gobbling up RAM. My laptop disk (even at 7200rpm) is too damn slow for swap to be at
all useful.  The system _will_ be dead until swap is exhausted and the
OOM kicks in anyway.  The only reason I have a swap partition at all
is for suspend2 hibernation.


But again you have shown that swap is *always* called for. You've got 2gb ram, yet you still need swap for hibernation.

Your example of having a real-time responsive app requiring memory
residence is a determining factor of how much physical memory you'll
need to keep the app resident.

But the truth of the matter is this will not be your only app running on
the system.  Throw some big memory hogs into play, i.e. an active X
session running locally and that remote X session you've started from
work, and pretty soon you can find yourself eating up that 1gb that you
thought would be fine.

No one would ever place a real-time responsive app on a desktop system.

So if your argument is that it would only go on a server, are you also arguing that it would only go on a dedicated server? Or is it a multi-function server that's also running perhaps a web server, an app server, an email server, ftp server, etc.?

The addition of any sort of server which spawns threads in response to incoming network connection means that you've got a variable memory consumer which could, should incoming load require, a potential chance to overwhelm physical memory.

Same situation, just a different scenario.


--
gentoo-user@gentoo.org mailing list

Reply via email to