Looking for an Enterprise backup NAS for archived info. Does not need to be cutting edge NVMe as this will be for long term archiving and backup storage. Looking for:
ZFS
250TB usable space
redundant power supplies
2 x 10GB NIC, minimum
redundant controller (This is negotiable, but nice to have)
Pick your favorite manufacturer, type ā4U storage server HDDā.
Any 4U rack server with 24 bays will do 400T raw which results into around 250T with RAIDZ3 (8+3 vdevs + 2 spare).
10Gbit RJ45 is on-board most of the time, but getting a dedicated NIC doesnāt hurt. Especially if infrastructure is fiber.
For HDD and archive+ backup you donāt need to go crazy on CPU and memory.Homelab level is plenty. This is mostly write once - read mayhaps workload. You canāt cache that anyway and unless there is a lot of compression (which is likely for archiving) going on, CPU will idle 99% of the time.
Spending 1000$ on a bigger CPU to save one HDD of space obviously isnāt worth it. So do the math whether compression and CPU-sizing correlate.
Pre-configured systems, like iX Systems, 45 Drives or the big OEMs all have reasonable sizing for those kind of servers.
A few months ago, Linus posted a video on his LTT channel, buying a āsurplus to requirementā NetApp SAN rack locally. LMG discarded the useless (licencing!) NetApp head servers and replaced these with an AMD EPYC system, sporting an external HBA. They basically kept the rest of the system as-is (was?) although I understand the SAS drives were taken out and replaced with high capacity SATA drives (which are of the SMR type, IIRC) by means of an interposer (per drive).
Thanks all for the comments so far. This is for an Enterprise business so support is all important and a DYI solution (as fun as that would be) is out of bounds.
I am looking at 45Drives at the moment. Got a quote from them in less than 10 min of chat on their web portal. Impressive stuff. Also talking to Dell, but they are taking ages (literally weeks going on a month) and still no quote. Lots of emails using flowery language to āengageā me and setting up unnecessary calls with sales people. The Dell web site is voluminous and garbage for reference. Not like it used to be at all.
We have a mix of 24 drive Synology and QNAP at the moment, but concerns over bit rot and long term viability of archives (especially in the face of legal cases) is becoming more important to the business. Thatās why we wanted to kick it up the quality a notch.
We donāt. We have multiple backups across multiple sites. Up until now any issues have been solved with alt backups. This tech will be assessed and if it meets expectations, weāll go forward and replace the other NAS with updated hardware over time.
Spending over a quarter million to replace everything only to find it does not work isnāt a good career move.
A wise move. 1-2-3 rule always applies. And having a (plugged out) cold backup that gets booted and scrubbed periodically doesnāt hurt in a case where you have resting static data over many years.
Cold backup is also resilient to something infecting the network or human error fucking up the pools+snapshots. 3 copies donāt mean shit if all 3 copies can go down for the same reason.
There is always the argument for tape in these scenarios too. So good to see the full range of options available. I know tapes still have a place in archiving, but Iām not familiar with specifics, so I pass on that.
Hardware is secondary as it doesnāt require performance, just rock-solid reliability and safety. RAIDZ3 can certainly deliver. Iāve seen a bunch of 4+3 and 8+3 erasure coding / parity configs for mission-critical data lately, so this seems to be some kind of consensus.
Sucky performance, as mentioned above, isnāt a disadvantage nor relevant. Youāre not writing TBs of data and random read small blocks from a database on these kinds of pools.
Itās just a fancy organized dumpster keeping things in stasis.
edit: What I like from an archival point of view is browsable snapshots on ZFS (and BTRFS). You can browse and see prior versions of a document or media. This can be a gamebreaking feature because you donāt need x times the amount of space. And everything is accessible without any zip folders or having to dig out some tape and bother a technician to restore some photo. Things like shadowcopy is just great for end-users.
Migration can be hard, because your typical documents folder has 10 documents for 10 versions and zip folders left, right and center, because old-school and no CoW. This kind of mess gets solved over time eventually.
Shouldnāt greatness be in quotes? (Iām currently observing multiple simultaneous horror stories at various locations about unrecoverable ZFS disasters, so the āgreatnessā thing made me giggle lol)
Unless you factor in non-ZFS disaster horror stories as well and compare, I call this cognitive bias and selective perception. There is no evidence to back this up.
Defect hardware happens. No software can solve this. Backups and a sound recovery plan do.
Iām not a ZFS-hating crusader, and I also have no interest in a tit for tat playground p*ssing match to plea my case here. Like anything else, itās not perfect (a core principle of objective reality that continually escapes a shocking number of otherwise perceivably and presumably intelligent individuals). Much of the details could qualify as proprietary information anyway, and Iām not interested in NDA violations, either. Cheers.