> From: François Revol <revol@xxxxxxx> > As long as the (de)compression cost is lower than the > slowdown of > accessing the physical disk I suppose it could be worth > it. > > Then we should have a compress/ kernel module subtree with > an API that > could be used for other subsystems as well (kernel-side > VPN, ...). While the zip libraries can be used that way, a true compression API that support multiple compressors would be nice to have. XFH for the Amiga was great that way. Make the de/compressors work much the same way translators do now and they would be available to all to use. > >> 3) Fake-Raid: > Besides, if that's what you want, then you could just > implement a native LVM instead. I already have my code working, what advantage do I gain by starting over the LVM? Plus it look like I will have to touch the kernel to make it work. My driver sits outside the kernel. > >> 4) Speed: > We have things like Bonnie++ in the test cases, that should > suit better than a simple dd :p I will look into that - thanks.