[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#245465: tmpfs for /tmp



Le samedi 10 novembre 2007 à 12:52 +0100, Frans Pop a écrit :
> I wonder whether this shouldn't include a check for minimal available RAM 
> size.
> 
> Is tmpfs still a good idea for systems with little RAM?
> I'm thinking NSLU, m68k, older 486/Pentiums, etc.

I think so. The reason is that on such systems, files will be almost
directly written to the swap.

With tmpfs, the situation is very similar to that of a regular
filesystem. On a regular filesystem, they will be cached in the VFS
layer during read/writes, and later if there is enough memory available.
IIRC, in modern VMs this is done with the same heuristics for all kinds
of pages.

Add to that the fact tmpfs has less memory overhead than other
filesystems, and this makes an argument for using it even on low memory
systems.

The only case where this will cause trouble is when there isn't enough
swap. In this case, real memory will start to be used for unused pages,
not leaving enough memory for caching files, which can cause giant
performance issues. By default, the tmpfs size is half of that of the
physical, so with our default swap sizes this is unlikely to happen only
because the tmpfs filled the swap. If you want to be safe, maybe you can
increase the minimum swap size from 96 MiB to 128 MiB.

-- 
 .''`.
: :' :      We are debian.org. Lower your prices, surrender your code.
`. `'       We will add your hardware and software distinctiveness to
  `-        our own. Resistance is futile.

Attachment: signature.asc
Description: Ceci est une partie de message =?ISO-8859-1?Q?num=E9riquement?= =?ISO-8859-1?Q?_sign=E9e?=


Reply to: