The idea is to put a caching layer between filesystems, that tend to be slow, and the user, who is impatient. This is accomplished by the fscache kernel module, and the cachefilesd user space daemon. The kernel module intercepts what would be disk/network access and redirects it to the daemon. The daemon uses local media, which supposedly is faster, to cache recent data.

The new Linux native implementation is very generic, and can be used to accelerate anything like floppies and CD-ROMs. I am interested in this because I find NFS slow.

Read more about it at Linux Magazine.

I have two boxes (i386 and amd64) in the lab that I use for testing of drivers I work on. Recently another Maxtor hard disk died on me, and I decided to get network booting working. I already have a file server from which I host my $HOME directories and do all backups from. It sounded like a win.

I’ve never done this before, so it took me a few hours to get the first host going, the second took 10 minutes plus the amount of time to build the kernel for it.

Below, I describe steps I took to get pxe-enabled hardware to boot a debian image, from a debian DHCP, TFTP and NFS servers.