Recommendation for expandable 'Backup' NAS box..say 40TB +

[squidman]

So I've researched freeNAS (not going to work not being expandable), unraid, (yes could work but at most 2 parity drives), and the other unraid equivalents (sorry forgot the name!).

Have currently workstation with minimal redundancy (try and copy over working files to a Windows mirror 1 Storage spaces volume), and not even sure (8x4tb drives at the moment!) if it tolerates more than 1 drive failure! But after a previous raid 5 failure (off mobo) and windows complete failure, not willing to go back to hardware raid or even mobo hardware/software raid kind of thing. Apart from scratch disks.

Current only backup is in the cloud, now 35TB and counting!

So Guess have to build a NAS as some kind of local backup, as am on 150/20 Mbps cable internet, and fiber isn't going to happen in the near future. Preferably with minimum 2 if not 3 parity disks, with a total capacity of at least 30TB and expandable with different sized disks. That leaves me with what, unraid, and the other one.

Any ideas? Kind of at a loss here, even after researching sth and anandtech and all the usual suspects...(without breaking the bank either!). One option (iff time!) is to simply have 2nd backup on Backblaze or something..using unlimited Jottacloud at the moment. But the upload times for 35TB on 20 Mbps up..well...

Got the box for it..an old Antec with the spaces in front for 4x (3x 5.25 to 5x3.5" drive bays). Miracle! Don't really want a Norco or server rack in my bedroom! See the need for 10g networking, yes that is on the list for the backing up, and researching (fiber optic cables etc! the old ones..well don't reach the 15 meters I'll need).

Any and all advice appreciated.

 

[Twice_Shy]

I've become at least mentally a fan of SnapRAID. (I do not yet have a SnapRAID based system but I changed my plan from using ZFS to using SnapRAID due to it's strengths and not minding it's weaknesses) The only downside by what I understand of how it runs is having to manually run checks and restorations instead of having it automagically do everything. It may not catch a failure in realtime, but it will let you know the next time a sync is run for instance.

I like the fact that I can use up to 100% of the disks capacity instead of maybe 80%. (and that's before RAID losses)

I like the fact I can use disks with existing data on them instead of having to migrate into a proprietary format.

I like the fact that I dont have any risk of total zpool failure - because instead of being migrated into a "black box" that no restore tools work with, it just uses existing windows and linux formats. It works "top down" over the file system, instead of bottom up creating a different type of file system. Have some kind of SnapRAID parity failure? No big deal the disks are as normally readable as they've always been, swap and put in a different chassis or on a different mobo and copy away for instance. Use all existing recovery tools.

I like the fact I can upgrade or downgrade the redundancy pretty much on demand (ie go from 2 to 3 parity disks or back to 2 without screwing up anything) and replace drives in place like a Drobo (as long as the files match it doesn't care) ie to upgrade a 4tb to an 8tb and keep adding to the archive.

I like what sounds like incredibly relaxed hardware requirements - whereas ZFS for the longest time suggested 1gig ram per TB so systems much over 32TB were more demanding to put together although that's less of an issue now years later than when I first researched it.

I like the idea of power savings of only spinning up the drive doing the writing (except when systemwide syncs and checks are run) instead of having what might be a stripe of 8 drives spinning just to update a notepad file somewhere. Wattage to idle and write even small files becomes more and more of an issue when you start having lots of physical drives under FreeNAS. (imagine a 48 drive system like below having to spin them all to update a Firefox cookie)

I like the idea of having up to _six_ parity disks tolerating six total drive failures backing up as many as 42 data drives. (you can use more drives than this, but the six parity maximum is then covering more drives than recommended) This literally matches my original "worst case growth max" which I predicted at 300TB easily using some SAS Expander chassis and already available drive sizes easily without overcharging me for the system design or setup before it's ever grown out to that size. I'm primarily paying to slide new drives in - without much cost overhead-per-drive like it normally grows when one starts setting up really big NAS systems.

It sounds really good for larger file sizes, maybe not so good for disks of millions of tiny files - endless textfiles, low res pictures, .wav files, etc.



Unfortunately i'm not yet using it because I have a few last projects to complete until I can build my NAS... but hopefully I will be soon!


To your other questions, properly filling a 10gig pipe is a nontrivial task for spinning rust unless you just need 'faster than gigabit but not maxing 10gig'. SnapRAID has no part of the performance angle - that would be about RAID stripes along an SAS Expander potentially, which others will know way more about than I do at this point.