| raymor |
05-01-2012 08:16 PM |
Quote:
What's LVM like vs ZFS? It seems like most home file servers are run with ZFS these days
|
They are very similar, especially now that LVM has snapshot pools. Before, you'd set a maximum size for a snapshot. LVM is more flexible, Zfs more integrated. A big difference is that LVM ZFS is both volume manager and filesystem in one, which is convenient for the common case, but limiting. LVM separates volume management from filesystems. LVM gives you volumes as block devices on which you can use any filesystem - regular filesystems like ext4, cluster filesystems like gfs, or you can build raid atop LVM, or even use LVM volumes as the devices on which you build other volume groups. We do that with virtual servers - the virtualized machine mounts volumes it builds from it's "disks", and those disks are volume on the host system, which are built atop volumes from the SAN.
That is, like mdadm, iscsi, etc. LVM takes some block devices, pools them, snapshots them, etc. and then exports - block devices! That means you can stack physical drives, SAN LUNs, LVM, raid, etc. in any order you choose. You can put a filesystem on a volume, or a volume can be a virtual disk image, or ... anything you can imagine. You don't have the filesystem forcibly inserted along with the volume manager like Z file system.
Adding and especially removing drives is a lot simpler in LVM, simply:
vgreduce mygroup /dev/sde
All of that flexibilty comes partly from being less integrated than ZFS. For example, flexible because the filesystem is separate, and that imies you need to create the filesystem (format the volume) separately from creating the volume. |
