[Discuss] Thin Provisioned LVM

[markw at mohawksoft.com]

As the storage wars continue, the debate of ZFS vs LVM continues. I have
been dealing with ZFS heavily for about a year now and just don't see it
as a viable file system for a lot of applications that would otherwise
benefit from its feature set.

Specifically "thin provisioned volumes" for virtual machines or iscsi
luns. Yes, ZFS zvols do support thin provisioning and the API is basically
correct. Unfortunately, the implementation of ZFS is too resource
intensive for much hungrier applications. LVM is much more light weight
and has better performance in applications that manage their own
journalling and data integrity (like a database).

LVM has recently gained "thin provisioning" of volumes, but its kind of
broken. You create a "thin pool" as an LVM volume and then sub-allocate
LVM volumes out of that. So, you have the volume group, the "thin pool"
allocated out of the volume group, and the volumes allocated out of the
thin pool.

I am not sure if this even makes sense. It is conceptually no different
than allocating a volume out of a volume group, putting a file system on
it (ETX2, say) and then putting a sparse file on it. The EXT2 file system
is performing the function of the "thin pool" code.

I think its kind of bogus.

Any opinions?