ZFS on our internal server
By avalon on Mar 03, 2006
This week we've seen ZFS being deployed on our internal server, jurassic, for the first time and it is being used in a rather revolutionary manner - or at least revolutionary for Unix. Rather than having a "home directory" filesystem, we have a "home directory pool" and each user has a home directory that is a filesystem within that pool.
The first "big change" that comes about, as an ordinary user, is when you do "df". Now, rather than seeing a dozen or so lines of output there is quite a few more....over 2,000. Does this mean I no longer type "df" to casually see how much disk space is being used/free, filesystems are mounted, etc?
So what are the benefits from adopting this model? If, in the case of home directories, a systems administrator had the ability to delegate ownership of the a ZFS filesystem to a user, the user can then potentially control whether or not their data is encrypted on disk, compressed on disk, and any number of other characteristics.
Another very obvious benefit is you no longer have to do "du ~foo" to see home much space foo is using in their home directory as it is immediately accessible via "df ~foo". Finding the largest consumers of disk space is now a relatively quick and trivial task of sorting df output!