The above reasons are especially important if you are running a virtualised OS where extra flexibility/simplicity helps and the performance difference is not really that big[1], both swap types are SLOW. :-)
Let's proceed. First we need to create a - say - 500M file; the best way to do it is via "fallocate" as it requires virtually no I/O (man fallocate), but you can also use good old "dd" if you're on an old OS:
fallocate -l 500M /swap.IMG
Next we need to format it, add it to fstab and mount it:
chmod 0700 /swap.IMG mkswap /swap.IMG echo "/swap.IMG swap swap defaults 0 0" >> /etc/fstab swapon -a
If you ever get in a situation where you need to increse swap you can simply do the above for a new file or just increase the current file:
swapoff /swap.IMG fallocate -l 1000M /swap.IMG mkswap /swap.IMG swapon /swap.IMGVoilà!
If you're working on a virtual machine you might want to avoid swapping as much as possible (many swapping instances generate significant I/O). This can be done via sysctl:
sysctl -w vm.swappiness=0And also add "vm.swappiness = 0" to /etc/sysctl.conf to make it permanent between reboots.
"vm.swappiness = 0" means it will swap only to avoid an out of memory condition.
HTH.
[1] - http://lkml.org/lkml/2005/7/7/326
