I haven’t read pinderkent’s blog in a while, but reading it always makes me wonder where his perspective actually comes from. Today’s little excursion (actually, last week’s) is how Sun’s ZFS will be supported by FreeBSD and might eventually be ported to DragonflyBSD (not — see below), and that, accordingly, everyone needs to become familiar with it because it will be “widely” used.
Let’s remember how many people actually use Solaris and FreeBSD. Then there’s the CDDL licensing issue that means ZFS will most likely not find its way into Linux distros (see below). This OS News post suggests that Matt Dillon has given up trying to make ZFS work with his Dragonfly kernel. So it’s unlikely that ZFS will be adopted beyond Solaris and FreeBSD anytime soon.
Then there’s the reality of the matter with respect to FreeBSD. One of pinderkent’s links goes to an announcement that ZFS will be available as an experimental feature in FreeBSD 7.0. How many FreeBSD users are going to adopt ZFS straight out of the box? Probably not many, especially in enterprise/critical use. And just how many users actually need ZFS’ features over what’s already stable in their respective OSes? Even fewer.
As promising as ZFS appears, it’s not going to be adopted widely anytime soon because it’s not ready for widespread adoption and because of licensing issues. Its current limitations are still significant, and its implementation isn’t without issues. And the licensing issues mean that ZFS would have to run in Linux userspace, significantly decreasing its performance (and compounding ZFS’ issue of overconsuming CPU cycles on small writes).