Ubuntu Puts Out A ZFS Reference Guide

[jrch2k8]

Copy&Paste:

For more pros/cons of deduping, refer to http://constantin.glez.de/blog/2011/...-or-not-dedupe. Deduplication is almost never worth the performance penalty.

For most part that is true but in certain scenarios like High Density Virtualization among others is extremely useful, like everything the right tool for the job

[kreijack]

Relating efficiency:

Well, why isn't reiser4 then the default filesystem in the Linux kernel ?

it has a pluggable design, is very efficient on the resource with its dancing trees, tail packing, etc. and also supports checksums & more

Relating to the checksum aspect, on the basis of this[*], it seems that Reiserfs support checksum only for small files (the ones that can be put in the metadata). It is not capable to protect every files by checksum.

This is not a causality: both BTRFS and ZFS are able to protect the data by checksum because these are COW filesystem. For a not COW filesystem is very difficult to ensure that the checksums are aligned to the data content.

Also ext4 has the capability to protect the metadata with checksum; but not the data.[*] http://www.spinics.net/lists/reiserf.../msg04786.html

[kreijack]

[...]
BTW, I was mistaken about resilvering. That looks to represent a particular implementation of scrub which skips over unallocated blocks. I don't know, and I haven't found anyone who claims to know, if btrfs' scrub works than way.
[...]

The scrub process, check the data against a checksum, which is calculated when the data is written. So it is impossible to perform a scrub process to an unallocated block , because the checksum ... simply is unavailable. I.E. if you have a 1TB filesystem with only one 1MB file, the scrub is performed only on the 1MB file.

[liam]

The scrub process, check the data against a checksum, which is calculated when the data is written. So it is impossible to perform a scrub process to an unallocated block , because the checksum ... simply is unavailable. I.E. if you have a 1TB filesystem with only one 1MB file, the scrub is performed only on the 1MB file.

Pardon. Resilvering is the process of adding a new disk to an array (so, requires a resync), and skipping over unallocated areas. According to the Oracle docs it's a form of scrubbing.
It's advantage is that it is able to recover from a degraded array faster because it's able to avoid the unallocated blocks.

[squash]

Having just hit the "no space left on device, while you clearly have tons of free space" absurdity in btrfs, I'm switching back to ext4. I love zfs and use it on bsd and solaris, but not having support for it in the installer is a real turn-off. zfs isn't just for multi-device pools, even if it makes those things easy.