Messages in this thread | | | From | "pacman" <> | Subject | Re: Out Of Memory in v. 2.1 | Date | Sun, 4 Oct 1998 23:51:50 -0500 (EST) |
| |
Brandon S. Allbery KF8NH writes the following: > >In message <Pine.LNX.4.05.9810042005290.276-100000@tahallah.demon.co.uk>, >Alex >Buell writes: >+----- >| On Sun, 4 Oct 1998, Alan Cox wrote: >| > And to boot a machine with 512Mb of RAM you'll need at least 600Mb or >| > so of swap file. >| >| Why would a machine with that amount of memory need that much swap space? >| I'd have thought it wouldn't need more than 128MB at best as swap. >+--->8 > >You missed the context: this is what would happen without swap-overcommit, >so every active page of memory has to have swap allocated for it at the same >time.
Why?
Here's how I see it: in a machine with a lot of RAM and no swap space, if the RAM is full all malloc()s (brks and mmaps and mremaps and reallocs etc.) are just refused. And that is not a problem, because correctly written programs can actually deal with those errors. The kind of error we CAN'T deal with is the random process deaths that can be caused by overcommitting. I think those are completely unacceptable. No amount of speed increase justifies that blatant unreliability.
-- Alan Curry
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |