Conclusions
Well, after so many pages, what lessons can we learn?
If you need maximum, stable performance, even in the face of a lesser flexibility and ease to use, go with classical LVM volumes. My performance builds where based on this LVM scheme and I am very pleased of how they runs. However, do not forget to make a judicious use of snapshots, as their legacy implementation is quite slow.
If you are ready to use a somewhat newer and more flexible (albeit less tested) technology, ThinLVM is your prime candidate, especially when coupled with a host-side filesystem: with this later arrangement you can disable zeroing without too much security concerns. Snapshots are quite fast, also. Only remember that fragmentation can somewhat slow-down you system over time.
A similar consideration can be done for ZFS on Linux, which shows very good results. You will definitely like its advanced features, most notably compression (with LZ4 compressor). But if you are thinking to enable on-line deduplication, please think twice: due to how it is implemented, it both require enormous amount of RAM and often give you low space saving (but this is good for another article...)
For VMs storage, stay well away from BTRFS: not only it is marked a “Tech Preview” from RedHat (read: not 100% production ready), but it is very slow when used as a VM images store.
I hope you found this article interesting. If you want, you can email me at
Have a nice day!