we may have to build a storgae solution that holds about 80TB of data - which means we would have to be able to install 240TB of storage capacity in order to have acceptable redundancy.
I’ve never come accross any motherboard where I could install more than 12 drives and even those are too small.
I would love to build this system myself with TrueNAS, but I’m a bit afraid I’m getting too far out of my abilities here.
Looking at some of the commercial cloud storage providers who offer storage at $7 per TB/month this would mean 7*80 = $540 per month (!) for the storage. This is definitely and very much out of our price range.
Or am I better off segementing the data on “cheap” 10 year old hardware (like those office PCs with Core i5 or similar processors and a couple of SATA ports each) each running a TrueNAS instance with 3 or 4 harddrives in them?
I’m at a bit of a loss here, as you can tell.
Perhaps someone can point me to some reading up on creating storage in this capacity range on a budget.
Welcome!
What probably you will need, Is not setup more different system, but a good server grade motherboard where place 1 or more HBA, according to your need of disks.
Certainly not; you’d need a disk controller.[1] A LSI/Broadcom/Avago SAS HBA is what you want, probably in the 9300 series. Combine that with SAS expanders (probably in the backplane of your chassis) and you can manage a practically-unlimited number of disks. See:
Well, not “need.” 12 x 20 TB drives gets you the 240 TB you mention. ↩︎
Is there a reason that RaidZ2 doesn’t get you the redundancy you need? Or maybe draid.
RAID is not a backup.
I mean the mirrors make sense if you’re reading/writing to the pool, but instead of triple mirroring you could use RaidZ3 have more redundancy and use less disks.
This would be purely an archive system - We would archive new projects to that storage about 5 to 6 times a year (roughly 500GB per project) and retrieve material from old projects probably with the same frequency. Transfer speed is not that crucial, because we access the data so little that we can wait a bit. Stability and reliability are more important.
We are not working off of these drives, our working drives are attached directly to our workstations (music production).
I thought for data to be relatively safe it needs to exist 3 times hence the idea of 4 X of 3 20TB drives (each triplet of drives mirroring) to achieve 80TB of usable storage space - so we could lose 2 drives of a triplet and still be relatively safe. But perhaps I’m not understanding modern storage concepts (zfs) correctly in that way… I may be stuck in the old way of me wanting to understand on which drive a specific file actually sits…
4 x 3-way mirrors would indeed roughly give you 80TB of space but here’s another idea.
2 x 6 disk Z2 would roughly give you 160TB with the same protection level per vdev.
Or how about
12 disk Z3 giving you roughly 200TB with a greater protection level per vdev.
Simply from a safety point of view option 2 is the best but if you’re trying to get the most bang for your buck with no plans to increase the system (unless in 12 disk chunks) then option 3 looks decent.
