Not sure if this is a good category for this, but…
Just curious, is there any good argument one way or the other for jbod vs raid 0 for combining the storage size of a number of equal size drives?
Never raid 0
It’s cool you get a minor performance bump, but a single failure means ALL the data is gone
Either jbod or enable some redundancy
2 Likes
Argument for Raid0: read and write speed is multiplied by number n of disks you use. e.g. 4 disks = 4x read&write speed.
against: You lose one disk → you lose all data
JBOD: no striping, so if you lose a disk, you only lose the data that that disk was storing. But because data isn’t striped across all disks, you don’t get the benefits of performance increase.
2 Likes
But if you jbod how is that any safer than raid 0?
Also, what if for example you raid 0 or jbod 4 4tb hard drives, and raid 1 that with a 16tb hard drive (because you have a 5 bay nas or something and the 4tb hard drives are really cheap per tb)
Oh, that makes sense. Could a single file end up being spread across 2 hard drives with jbod?
Unlikely, but depending on how JBOD allocates data, you might end up with half a directory stored on disk1 and the other half on disk2, which is obviously a problem. If your car has a broken tire, it won’t go anywhere despite all other parts working flawlessly. JBOD also doesn’t have any form of redundancy, so there is no way to restore the data of the failed disk.
JBOD has fallen out of favor for good reasons. That’s because of the availability of cheap/free RAID for basically everyone nowadays, making everyone using RAID10 or parity RAID (like level 5+6) de facto standard for >3 disks.
1 Like
ok, let’s say I got a 5 bay nas or something and raid 1 a 16tb hard drive with 4 jbod 4tb drives (because the 4tb hard drives are really cheap per tb, so it makes for relatively inexpensive redundancy).
Could that make sense?
Flip side - you should have backups anyway, and RAID0 will give you better performance.
If you don’t have any backup plan for when the array fails, machine gets stolen, you have a dumb shit moment and delete the wrong stuff, bit rot happens, house catches fire, etc. - fix that problem as a priority.
And yeah, JBOD still loses your data in case of a single drive failure. Just some of it which may (murphy’s law says: WILL) be the critical stuff… in case of a drive failure.
If you have 5 drives full of data on jbod, one drive failure means you loose the data on the drive that failed, if you had all the drives in raid 0, and a single drive fails all the data on remaining drive will be corrupted
See above: if that one drive has the data you care about, you’re still boned.
Doesn’t matter that you have 4/5 of the less critical stuff, you’re still fucked if the wrong drive fails.
Even if you’re running triple redundant RAID1 - have up to date backups. The same machine/array/location is not a backup.
Counter point would be now you have to restore all of your instead of just 1/5 of it for one point of failure
Or just replace the drive if you have at least one level of redundancy
If data is critical why would you risk complete down time for a single point of failure
2 Likes
You should always buy drives with the same capacity. Different capacities is just asking for trouble when talking RAID. I’d go for a 4-wide RAID5 with the 4T drives and use 16T drive as backup. Without the need for exotic nested RAID configuration.
2 Likes
The situation is that I’d want everything on the 16tb hard drive, and use relatively cheap (per tb) 4tb hard drives as a mirror
Because both types being discussed have a single point of failure 
I’d do neither. RAID10 or RAID5/RAID6.
I haven’t seen a JBOD in a production environment in 20+ years… given the choice of only RAID0 or JBOD though, may as well take the speed.
1 Like
The situation @Exard3k proposed would actually be better to use the raid 5 as the main since it’s striped just like in raid0 just with a parity drive
But I wanna be able to take out the 16tb drive and replace it with a fresh one, and use the removed 16tb drive as-is as a backup (possibly moving it someplace else), or just for any reason, and have the copy all on the one drive
That is literally also true of jbod. Jbod is just raid0 without the speed improvement and without the requirement for equal sized drives. I used to jbod my two SSDs together out of laziness looool
I say raid0 them. You won’t lose data provided you back up regularly like you should be doing anyway
The two drives I jbodded together eight years ago are still in my system now. Raid0 only multiplies the probability of a drive failure by the number of drives you have. Two times a low probability is still a low probability.
1 Like
4x4TB RAID5 giving you 12TB of usable storage.
Make the 16TB an untouched drive by any service other than the NAS and have ther 12TB RAID array back up into as a warm mirrored backup.
If the array fails, you still have the 16TB drive till you fix the array.
If the single drive fails, you have time to replace it to get your backups back up and going.
Unless the failed drive doesn’t report that is failed and your system writes junk days to your mirror. If much rather manually backup on a schedule than mirror.
For sake of data integrity, unless you also need the 16TB of usable space, put the 4x4TB drives in a RAID 5 or equivalent, as everyone else said.
On both Synology and QNAP devices, you can schedule backup jobs to the other drive, which will achieve the desired result without any weird mirror configuration going on, and you won’t be destroying your redundancy every time the 16TB drive is removed, risking the only version of everything being stored on the one drive.