Ive installed a 6TB raid array and it runs so quickly. Im loving the smart monitoring and easy to use cifs. They suggest using raid-z. From what I understand this is a better version of raid 5? Was hoping you could let me know a bit about that.
So I'll try keep this as simple as possible.
ZFS is a file system which also offers features that weren't usually handled by file systems (eg. redundancy, error checking, backup, etc.)
Each file in ZFS has a checksum (there is a lot of theory here but that checksum can virtually never clash unless the data is correct). So ZFS will always know if a file is damaged, regardless if you have redundancy setup.
ZFS has two types of redundancy, RAID-Z and Mirror. Mirror offers the same redundancy as RAID-1 but it is implemented completely and entirely different from RAID-1.
RAID-Z offers 3 levels. RAID-z1, RAID-z2, RAID-z3. Each level indicates the failure tolerance. Eg. z1 = single drive failure redundancy, z2 = double failure redundancy, vice versa.
Once again, RAID-Z is no way like RAID-5/6 in how it is implemented. Don't even think they are comparable, they aren't. RAID-z is superior in many ways with none of the weaknesses of RAID-5/6. There is a lot of theory around what is better and why and it would take pages to explain it so I won't go into that now. Suffice to say that RAID-5/6 CAN and WILL have data corruption, even the most expensive kind. It was in fact CERN that published a lot of research showing how much *silent* data corruption they suffered from their RAID-5 systems.
Apart from that Sun Microsystems recognized that RAID is not cheap and the controllers are falling behind development in the CPU & Memory fields. Even modern smart phones now rival some of the most expensive RAID controllers in processing power and memory bandwidth.
Currently there is no RAID-5/6 controller that could come even close to offering what you get with ZFS in terms of performance, price or any kind of measurement I can think of. Once again a lot of theory. If anyone thinks I'm wrong on this, I'll happily elaborate why you are wrong (yes I'm that confident
).
So you get the redundancy of RAID-5 (z1) or RAID-6 (z2) but at a much lower price and it also happens to have much lower chance of data loss or corruption.
Im also curious about how the snapshots work. I assume theyre something like images? If I have 9 TB worth of data is the initial snapshot 9TB in size or is it compressed?
ZFS doesn't compress a Snapshot.
ZFS does offer various forms of compression for the file system, and you should be turning it on. Disk IO is the slowest part of a NAS, so the higher your compression the better your throughput, that is if the CPU can keep up. I have compression enabled (the default compression) and my highest CPU usage is around 40% while maxing out a gigabit Ethernet connection.
So snapshots are very technical to explain but I'll try my best:
Essentially ZFS NEVER allows data to be overwritten. Instead data is written in a new place and the file system table is updated with the new location of a file. This is called Copy-On-Write. This also means if you have absolutely 0bytes left on your file system, you can no longer delete because it cannot modify data in place and you are screwed (so never let it get to 0bytes free). FreeNAS does offer you the option of reserving some disk space, I suggest you use that if you think you are going to use up all your space till you have 0 left.
Ok so a snapshot: It takes the file system table, makes a copy.
Any files that are deleted are still being kept track of by the file system table so it is not really removed from disk, until you remove that snapshot.
Essentially unless you delete a snapshot will take 0bytes of space or near 0 (if you just move files around).
It is very technical but if you are a programmer, it is like a linked list. And you make a copy of a link list and keep it around somewhere. A GC wouldn't collect the objects linked to that list until you get rid of that list. But essentially that list isn't costing you much memory because it just has references to objects.
Hope that explains it somewhat. If you are unsure about anything I have written ask away. And I welcome any hardcore I Love Hardware RAID guys to debate about why ZFS is better than hardware RAID in every way and why software RAID *will* replace hardware RAID (and is already doing so).
So you can bet that once the release is made end this month, it should have sabNZBd.
