Software RAID performance is greatly reduced when layered on LVM.
Links to some concrete benchmarks/research/articles/documents showing this would be nice.
Updates to pros and cons
It seems by 2.6.37, all of LVMs current cons should be squashed. (Discard support, so SSDs can perform TRIM)
LVMs are still geared toward storage experts, though. The tools to properly manage LVMs are all TUI based. None of the features of LVMs, such as snapshots, are used in Desktop Fedora.
There still could be future problems if LVMs don't have support for Future Feature A and it takes 5 kernel releases to see support for Future Feature A.
What benefit does a Desktop Fedora user gain by having an LVM? Reduction of 1 MS-DOS MBR record? (swap and / combined)
The advent of GPT and Btrfs will make LVM obsolete - sorry. (Yes, GPT will become the default in 2011, motherboards will be coming with UEFI)
--Mooninite 19:33, 15 November 2010 (UTC)
BTRFS hasn't stabilized for _years_ and generally ends up eating users' data - sorry. DM thinp and DM cache both offer more compelling features when paired with a stable filesystem like XFS. Energy should be devoted to working with technology that has proven stable (DM/LVM with XFS ontop) rather than pinning the hopes of Fedora Desktop/workstation on BTRFS. systemd could easily be trained to work with DM-thinp snapshots for container capabilities, etc. It makes little sense to splinter the Fedora storage layers deployed across Fedora variants. Making DM thinp the default for the root XFS filesystem would be more prudent as it will leverage technology that the upstream kernel developers are working to support in the enterprise (XFS on DM thinp is already used by Project Atomic and Red Hat Storage - aka Gluster).
--msnitzer