4 ms·
The important bit for NAS is redundancy. If you can afford 2+ of these drives to run in RAID1, RAID5 or RAID10, then go on ahead. For me personally, my 5TB x 2
by dragontamer 6y ago
The important bit for NAS is redundancy. If you can afford 2+ of these drives to run in RAID1, RAID5 or RAID10, then go on ahead.
For me personally, my 5TB x 2 RAID1 (ZFS really, so mirrored) configuration with 5TB usable is enough for my personal uses. If I need more space, I'll configure my ZFS with +2 HDDs for another pair worth of space+redundancy.
- duncanawoods 6y agoIf you can only afford 2 then I decided against RAID/ZFS. The issue is that "raid is not backup" and I fear user error as much as hardware error so instant mirroring would be just as bad. For my home bulk storage, I just have two normal 14TB drives and a nightly cron to rsync archive (new files only) from the primary to the backup. This gives me enough of a backup for both user error and hardware failure. It doesn't protect against a house-fire but that is ok. I should maybe have considered BackBlaze but I can't believe they are going to last long if they continue offer flat-rates for multi-TB attached NAS.
- dragontamer 6y ago> If you can only afford 2 then I decided against RAID/ZFS. The issue is that "raid is not backup" and I fear user error as much as hardware error so instant mirroring would be just as bad. Mirroring is for preventing hardware failures. ZFS Snapshots are one protection against user-error. Backups are another.
- duncanawoods 6y agoI was tempted by ZFS but it just felt like too much trust to put in something complex as a novice. Just reading about it led me to too many data-loss stories. I have had enough corporate IT teams tell me their RAID went down with total data-loss and their tape drive recovery isn't working to find a lot of virtue in "dumb-as-hell" infrastructure design!
- dragontamer 6y agoSpend an hour setting up a ZFS on old hardware. Play with snapshots and manually delete test documents until you are confident on how the feature works.
- npongratz 6y agoHighly recommended! One can even learn about, and experiment with, ZFS using files sitting on a disk by setting the files up as loopback devices. For example: https://dev.to/nikvdp/https-nikvdp-com-post-zfs-part1-31fj https://dev.to/nikvdp/https-nikvdp-com-post-zfs-part1-31fj
- dragontamer 6y agoHmm... that's a good idea. It may be simpler to play with ZFS on VMs. The VM-disks will be small and inefficient, but "good enough" to prove whether or not the system works. Not that I've tried any of this before. I happened to have an old computer lying around with ~4 small hard drives that I messed with to learn the ins-and-outs of ZFS the first time I used it. (Shuck hard drives, pull them out of laptops, etc. etc. Its not really that hard to get 4 hard drives to play with, especially if you ask your friends for ancient laptops with 200gb drives or something)
- mbreese 6y agoZFS is very robust, especially in a mirrored setup. Compared to your sync setup, with ZFS you could have snapshots, which would let you rewind back to earlier versions of files. This is a good mitigation for ransom ware attacks. You also get bit-rot protection, where you know what you read is what you wrote. Without this protection, if a file gets corrupted on your primary drive, this corrupted file will get synced to your backup drive. ZFS helps with both of these and a mirrored setup would give you a boost for read IO speeds. Of the many ways storage can get complex, I've found ZFS to be one of the most straightforward.
- kalleboo 6y agoYeah I've run a home fileserver at home before and it's just not fun continually worrying about updates and having to maintain a bunch of monitoring if some cron job broke or something. They're not exactly cheap, but I'm another one who went the Synology route. They slap a user-friendly UI ontop of linux mdraid and BTRFS, so you can RAID together a few drives, set up automatic BTRFS snapshots with retention rules (keep 12 hourly snapshots, 7 daily, 3 monthly, etc), set up disk scrubbing schedules etc all just by clicking around.
- Latty 6y agoTo add to that, one will very quickly find that RAID5 really isn't a good idea for big drives if one reads around. The reason being that for larger drives, resilvering can take so long that secondary failures become probable or—in bad cases—even virtually guaranteed (This is the same probably the WD SMR drives had, but it exists even on normal drives if they get big enough). For this reason, for modern large drives, most people recommend at least RAID6 (known as RAID-Z2 on ZFS), which means you have two drives worth of parity as compared to RAID5's single one. Of course, this means that you need at least four drives for it to be worthwhile, and most people realistically end up going for around six disks, otherwise you are spending a lot of your total cost on parity.