there is a quite technical explanation at LWN
looks to me like it most useful for virtual machines. or maybe as a general backend that you write the kernel's file caching system in. though maybe i have missed something.
or could this be used like bcache? with an ssd for the cache for a large hdd.
Guys, see zcache - compress the cache, thus fit more files into cache, thus avoid more HD trashing > more fluid usage in general.
More fluid usage maybe, but longer 'interupts', meaning faster regular hdd acces time (which is useless given that RAM is cheap and huge) and longer kernel mode time. Doesn't this actualydegrade performance in exchange for little to no gain whatsoever? I mean preload should be way more efficient given HDD size compared to RAM size....
PS: I'm talking about compression.
LZO compression is pretty damn fast on current cpus. They have benchmarks of course, the most gains are with many cores, while single cores gain a tiny bit IIRC.