SSD-only array: JBOD vs RAID

Currently reading
SSD-only array: JBOD vs RAID

6
0
NAS
DS216+II, DS620slim
Operating system
  1. macOS
Mobile operating system
  1. iOS
Hi everyone,

As a Christmas present to myself :p, I've ordered a DS620slim with 3 x 4TB WD Red SA500 SSDs in order to retire my DS216+II as a backup NAS. I'm really looking forward to the almost-silence and the performance gain, but am a bit unsure on how to configure the volume. Let me explain:

With 3 spinning disks I would most probably go for a RAID 5 or SHR setup, but with a traditional RAID configuration that distributes parity evenly together with the limited write cycles of SSDs, there seems to be a real chance that multiple SSDs will probably be worn out and fail at the same time. For exactly that reason, Synology has developed RAID F1 - however, this RAID type is not (yet?) available for the more consumer-oriented models such as the DS620slim (yes, I knew that before ordering, but the enterprise models seemed overkill to me plus they are usually louder and don't offer HW transcoding) .

So I'm asking myself whether it's actually worth it to go for a SHR or RAID 5 setup (especially considering the price of SSDs and the risk outlined above) or just leave it as a JBOD volume using the full storage capacity. Or to put it another way: If the probability of two SSDs failing in a RAID array is roughly similar to one SSD failing in a JBOD array, then there seems to be no real gain for going "redundant", just added cost. Finally, I can't say much about the reliability of SSDs, but from what I've read so far they are regarded to be at least as reliable as spinning disks (or even much more).

Regardless of going redundant or not, I'm planning to make a complete weekly backup of my new NAS to my old NAS, plus a daily backup of my most important data to the Synology cloud. Downtime to restore from backup is not a big issue for me, it's for personal use.

Thanks for sharing your thoughts!
 
Nice setup, may it serve you well.

Regarding volume choice, considering that you have 3 identical drives, you will not have a different r/w/paratiy setup compared to the conventional RAID5 vs SHR.

I guess the only real question now is raid or basic/jbod. The only benefit with SHR will be putting a larger drive in the array and having the benefit of maximizing storage.

You seem to have done your homework, so the decision is definitely on your end. With a backup plan you have, and if continuity is not an issue, going without a raid setup will be just fine.
 
just few add-ons to your considerations:
1. RAID0 rather than JBOD. JBOD's panning is for the SSD scenario a killing architecture (you will use for write operation just allocated SSDs space and unallocated will "sleep unused". Instead in RAID0 you will utilize all the SSDs.

2. on the other side, when you will lose single disk from the RIAD0 you will lose entire RAID0. Same issue could be happen when you lose first of the JBOD disk, where you can find all necessary filesystem metadata. But in both sides, there are possible way how to recover the data (when there aren't a physical damages).
Btw I have in usage two RAID0 pools, older is from 2015 with Seagate Constellation HDDs. No troubles.

3. BTRFS: not sure (never tried) if it's possible to use btrfs with JBOD
 
Good choice, I do have a DS620slim with SSD as well.
I had the same kind of thoughts as you and came to different conclusions, and set different priorities.
Let me explain and feed your thoughts a bit..
  1. My write load is not that high, I calculated I would not reach the write cyles in 10 years, so did not regard the life time of the drives as relevant. My samsung in raid 5 is still at 97% life time after 1 year.
  2. I hate down time, so the redundancy is important for me to keep the system alive when a disk fails. The extra costs are, depending on your setup 30% to 25% if you have 3 or 4 disks.
  3. Both in JBOD an RAID0, as indicated by @jeyare, you loose all data if one disk fails, so I prefer single volumes above RAID0 and JBOD at all times, reducing risk for 100% data loss with factor 4, and risk for 50% data loss with a factor 2.
  4. About sound: as SSD temperature is higher than a HDD, the fans may spin up and increase noise. There is a post on this forum that describes how tackle that. Wattage is low, so there is low risk of overheating. DS620 is less silent than I thought it would be.
Do what is best for you!
 
Gents, as was written by @1nn1k0 :
Regardless of going redundant or not, I'm planning to make a complete weekly backup of my new NAS to my old NAS, plus a daily backup of my most important data to the Synology cloud. Downtime to restore from backup is not a big issue for me, it's for personal use.

then any advice to redundancy is out of scope. He understand the basic advantages/disadvantages of redundancy.

What is more important:
- people don't have experiences with RAID0, they can see just disadvantages. With proper backup plan (rsync, ...) is RAID0 better. And @1nn1k0 has a backup NAS for such purpose. Don't hesitate.
- people don't have clue how JBOD makes heavy impact to the SSD lifespan. It's about physics.
 
Thanks for pointing that out. @1nn1k0 has prepared very well and I just wanted to give extra thoughts, no advice.
Unfortunately the Synology backup seems to miss several parts (it may have improved since last 2 years, since my latest full restore, i know hyper backup tasks are now restored). Many apps, settings of firewall, white and blacklists, DDNS /quickconnect settings etc. to mention a few. So still prefer to have at least one disk running in the system and not loose the full array and redo those settings. But again, that is personal.
 
@EAZ1964 - agree, then we need understand all of the possible RAID0 failures:

1. Most common failure of RAID0 is due disk logical corruption (bad sectors) or physical damage
2. Second possible failure is due software (disk firmware) failure or host controller failure

The first one is (for me) easy manageable by single or more tools - e.g. great opensource TestDisk (Win, Lnx, Mac), Hard disk sentinel (covers also SSD). For Btrfs it needs btrfs own "check --repair" procedure (but you need check first LVM in Syno NASes).

The second one needs more attention, but a repair is still applicable, sometime. Amount of such failures is few percent from total possible failures.

Finally I must agree with the proposed point of view:
- with HB you can't backup bootloader of DSM. But isn't necessary, because you can download the bootloader from Syno pages. Unless you lose bootloader HW, but this issue you can't cover by RAID (no way).

- with HB you can't backup rest of Syno subsytem, saved in special volume. But you can do it with ReaR tool. Yes this is one of Achilles tend of Syno NASes, what I don't understand more than 10y.

- with all the rest RAID (1,5,6) architecture you can save a time vs. the both of the kind of disk failures repair steps, till you lose last redundant disk in such chosen array. And here we are:
The proper backup plan and new RAID create, then Restore is better from any point of view than time consumption from RAID5/6 rebuild, especially for tenths TB. For a skilled person. Ofc for newbies is better to use redundant arrays.
Then is necessary to decide data storage architecture first and try think about all possible scenarios. Also a periodical maintenance of the HW/SW part of the NAS is necessary. Include checking of the SMART (advanced) statuses, better from cli with smartctl. Then your NAS can give you a responsible and secure data operation. All the hurray action as we can read also in this forum is a highway to hell. NAS is a complicated engine, needs an attention and right operation. NAS isn't additional external USB disk.
 
Last edited:
just few add-ons to your considerations:
1. RAID0 rather than JBOD. JBOD's panning is for the SSD scenario a killing architecture (you will use for write operation just allocated SSDs space and unallocated will "sleep unused". Instead in RAID0 you will utilize all the SSDs.
Interesting, thank you @jeyare - so you mean that with RAID0 striping I would divide the write load equally, which might increase the total lifespan of the SSDs? Or to put it differently: Both RAID0 as well as JBOD will fail with one SSD lost, however RAID0 might be beneficial because it distributes the write load equally, delaying SSD wear out.

Another advantage might be that I would not fill up one SSD entirely, which might also be problematic (however, I'm not sure if that's tackled by over-provisioning). In my case, I have ~ 6 TB of data, which would mean in a JBOD (3x4TB) setup: SSD1: 100% capacity used, SSD2: 50% capacity used, SSD3: 0% capacity used. In a RAID0 setup, all SSDs would be equally filled up to 50%.

One disadvantage that I see, is that JBOD allows for easy volume expansion, while RAID0 does not.

3. BTRFS: not sure (never tried) if it's possible to use btrfs with JBOD
Never tried as well, but I found some people using btrfs with JBOD.
-- post merged: --

  1. I hate down time, so the redundancy is important for me to keep the system alive when a disk fails. The extra costs are, depending on your setup 30% to 25% if you have 3 or 4 disks.
Thanks @EAZ1964 - yes, I understand, however, I'm unsure if a RAID setup would provide real redundancy after all, as it seems that multiple SSDs might fail at the same time (due to be worn out equally).

Citing from Synology RAID F1: Rethinking RAID in all-Flash Environments
Current software RAID solutions work well for HDDs, but transitioning to SSDs provides a new challenge for data protection solutions. With the knowledge that SSDs have a set number of read and write operations in their life span, writing data evenly and concurrently to all SSDs in an array could result in all of the drives “expiring” at the same time. There is no RAID technology available that could survive a simultaneous failure of every member disk, this scenario would be a data loss disaster.
For that reason, Synology developed RAID F1. Since that's only available for enterprise models, I think there's (currently) no good redundancy solution for users of consumer models with all-flash arrays. And that's one of the main reasons why I'm actually considering RAID0 or JBOD (together with a good backup strategy).
  1. About sound: as SSD temperature is higher than a HDD, the fans may spin up and increase noise. There is a post on this forum that describes how tackle that. Wattage is low, so there is low risk of overheating. DS620 is less silent than I thought it would be.
Thanks, I'm not so much bothered by fan noise, but still good to know!
 
@1nn1k0 I have a similar situation that I've been trying to figure out. I was originally planning to run F1 on my DS1621xs+ so that I could plan on one disk reaching its lifecycle before the others. Since my write cycle won't be heavy, I have opted to run 3 x 4TB SSDs in RAID 0 and will put 3 x larger HDDs in the remaining bays (6 bays total) as the first target for backups. After calculating my approximate usage over the next 5-7 years, I think they'll barely hit their halflife if all else is well. I also plan to have current work projects backed up to a cloud location daily as well. Good luck.
 
@1nn1k0 I have a similar situation that I've been trying to figure out. I was originally planning to run F1 on my DS1621xs+ so that I could plan on one disk reaching its lifecycle before the others. Since my write cycle won't be heavy, I have opted to run 3 x 4TB SSDs in RAID 0 and will put 3 x larger HDDs in the remaining bays (6 bays total) as the first target for backups. After calculating my approximate usage over the next 5-7 years, I think they'll barely hit their halflife if all else is well. I also plan to have current work projects backed up to a cloud location daily as well. Good luck.
Thanks @Whitefxdesign, that confirms my thinking - even though I might have opted for F1 if available on the smaller DS models. May I ask how you've configured your HDD array for backups - did you choose a RAID0 or RAID5 for that?

As stated above I'm planning to use my old DS216+II for backups - but since it has only 2 bays I would be limited to 8TB in a RAID1 setting with my current 2x8TB disks. So I'm thinking whether I go for a JBOD setup having 16TB of backup capacity for the future (currently I have 7TB to backup, so I'm near the 8TB limit soon). However, I would feel better with a redundant HDD setup, especially since I'm relying on that backup in case my SSD RAID 0 would crash. So I guess I might go with a RAID1 or SHR setup on my existing 2-bay NAS (with 8TB redundant storage) and if I'm running out of space, I'll get a 4-bay such as the 420+ or the 920+, pop in my existing two disks and add one or two new ones. Not sure if that's the smartest way though.

Cheers and a happy new year to all of you!
 
good to know, that we can find here also critical thinking from new members. Not just shoot into dark based on first line from Google search.

Gents, everything is in your own hands. Have a great experiences with your data. Don’t forget, just bad experiences can move our attitude to future.

PS: I have all my home data in RAID1 or RAID0 based on data tiers and usage purpose. With proper backup policy you can’t lose any critical data.
Same in my business operation. Driven by data availability more than cost of 1TB. To be sure when cost of 1TB storage space in SME is critical point, don’t purchase a car over 2000Eur in total. Then something is wrong in consideration about data value.
 
As stated above I'm planning to use my old DS216+II for backups - but since it has only 2 bays I would be limited to 8TB in a RAID1 setting with my current 2x8TB disks. So I'm thinking whether I go for a JBOD setup having 16TB of backup capacity for the future (currently I have 7TB to backup, so I'm near the 8TB limit soon). However, I would feel better with a redundant HDD setup, especially since I'm relying on that backup in case my SSD RAID 0 would crash. So I guess I might go with a RAID1 or SHR setup on my existing 2-bay NAS (with 8TB redundant storage) and if I'm running out of space, I'll get a 4-bay such as the 420+ or the 920+, pop in my existing two disks and add one or two new ones. Not sure if that's the smartest way though.
After some consideration I decided to go for a RAID1 or SHR setup on my backup NAS. However, to add some backup space I would add an existing external disk through the eSATA port. Can anyone confirm that I can use an external disk connected to a remote NAS as a backup target for HyperBackup?
 
Can anyone confirm that I can use an external disk connected to a remote NAS as a backup target for HyperBackup?
Yes you can:

Screen Shot 2021-01-05 at 08.55.47.png
 
Great, thank you! Ordered a eSATA cable and will try that out (never thought I might need one, but better than the USB 2.0 on my external disk :LOL:)
Don’t think esata will show your external drive. Only usb ports. esata is for expansion RX units. I could be wrong but double check that. I do have esata/usb bay but it’s usbA port that goes into the nas and esata into the case. Not sure if that’s the combo you ment.
 
Don’t think esata will show your external drive. Only usb ports. esata is for expansion RX units. I could be wrong but double check that. I do have esata/usb bay but it’s usbA port that goes into the nas and esata into the case. Not sure if that’s the combo you ment.
I haven't heard of such a limitation so far. And on DiskStation Manager - Knowledge Base | Synology Inc. USB as well as eSATA drives are mentioned:
By connecting an external drive to the system, you will be able to share its drive capacity through a system-created shared folder named usbshare[number] (for USB drives, if applicable) or satashare[number] (for eSATA drives, if applicable).
I hope that's correct...
 
eSATA and also USB external drive will do their job as "Shared folders" and not as "Volumes" = because DSM will propagate them just as Shared folders. There is the main difference.

Then for the backup purpose (HB) both of them you can use, because in HB you need setup target in another NAS defined as Shared folder (not as Volume). Also valid for HB rsync backup tasks.

The Volume you need just for a plain rsync, what is different backup purpose, independent from HB.
 
@1nn1k0 I have a similar situation that I've been trying to figure out. I was originally planning to run F1 on my DS1621xs+ so that I could plan on one disk reaching its lifecycle before the others. Since my write cycle won't be heavy, I have opted to run 3 x 4TB SSDs in RAID 0 and will put 3 x larger HDDs in the remaining bays (6 bays total) as the first target for backups. After calculating my approximate usage over the next 5-7 years, I think they'll barely hit their halflife if all else is well. I also plan to have current work projects backed up to a cloud location daily as well. Good luck.
Hi, just wondering how that 3x SSDs in RAID0 has panned out for you over the last 2 years?
 

Create an account or login to comment

You must be a member in order to leave a comment

Create account

Create an account on our community. It's easy!

Log in

Already have an account? Log in here.

Similar threads

  • Question
Good to hear. You could also clone the drive and expand the data partition using Clonezilla/gParted.
Replies
7
Views
877
Replies
9
Views
1,435
Thanks - Good thinking on the ps. Both drives happened at the same time - so you may be 100% correct. I...
Replies
2
Views
969
Hi all, What are thoughts on adding Intel Optane SSD which maxes out at 64GB as SSD cashe to my Synology...
Replies
0
Views
2,646
I do not see this graphical cache hit rate either. This obviously has been dropped by Synology, running...
Replies
8
Views
1,972
  • Solved
This is a common question here, and there are answers to those questions. Still, it is always a "per case"...
Replies
1
Views
1,142
  • Solved
Thanks Rusty, I've been reading a few articles and guides, and been trying to find a better way to make...
Replies
2
Views
1,768

Welcome to SynoForum.com!

SynoForum.com is an unofficial Synology forum for NAS owners and enthusiasts.

Registration is free, easy and fast!

Back
Top