Re: swap killer kernels

Mark Lehrer (edge@dux.raex.com)
Mon, 30 Mar 1998 15:41:56 -0500 (EST)


Would it be reasonable to select from two or three different "swap
personalities"?

For example, if I have 128MB and I am certain that I'll never use it,
I don't need any swap logic to even make it into my kernel... it would
just waste a few K of RAM and some CPU cycles. (go ahead and tell me
to RTFM if this is already a make config option).

Or, are we shooting for a one-size-fits-all swapper that behaves
reasonbly in all conditions? It might be useful to be able to give
the swapper a compile-time hint as to how much overcommitment the
sysadmin is planning for the system.

Linus said:

>Andreas Jellinghaus <aj@dungeon.inka.de> wrote:
>>same problem here :
>>i have 64 mb, currently no swap partition (used by a dos based game :-).
>>i never got real problems before with 64mb, but starting with 2.1.91
>>compiling (and X, some xterms, several daemons like exim or inn, all
>>this usual stuff) can kill the system : hard disk is accessed like mad
>>(don't know what the kernel does), and X gets nearly freezed
>>(to slow to do anything - Zap keystroke didn't help for minutes, so i
>> had to use reset).
>
>The 2.1.x kernels almost require a swap area. They should work without
>one, but as you see they can get rather upset if they don't find one.
>
>Essentially, not having a swap area means that the kernel cannot get rid
>of the pages it would like to get rid of, so it has to spend more CPU
>time finding pages that it _doesn't_ like getting rid of in the first

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu