[arch-general] Zramswap for /tmp

Mark Lee mark at markelee.com
Thu Mar 13 09:47:00 EDT 2014


I've had excessive swapping (high cpu usage with kswapd) and the
termination of running programs (like Skype) due to zram swapping. I used
zramswap from the AUR to start the service, but it only calls on the kernel
to produce the block devices. A compressed zram device (like maybe F2FS
partition zram block) for /tmp should eliminate the risk of /tmp files
taking precedence over running programs with a zramswap. It would limit
/tmp performance to the max decompression speed of the zram block.


On Thu, Mar 13, 2014 at 8:06 AM, Simon Perry <arch at sanxion.net> wrote:

> On 13/03/14, Mauro Santos wrote:
> | I have had it enabled for a while and I haven't yet seen any nasty crash
> | due to it. Hibernation also works fine and seems to go faster than
> | without zswap.
> How much RAM do you have?
> zram is probably solving a problem you don't have.
> Enable it on a VPS that has 256MB RAM (which is supposed to be its use
> case) and get back to me.
> --
> Simon Perry (aka Pezz)

More information about the arch-general mailing list