Question Is SHR better than RAID and When?

[jeyare]

There is still open question from my side: Is SHR better than RAID an when?
To be sure I will use SHR1 for specific type of SHR (mentioned in Synology web as SHR), otherwise SHR is general term. Also valid for RAID.

1. SHR1 vs RAID1 - two identical disk - single disk redundancy for both of them. Same, no mater what do you want SHR1/RAID1 (see in picture below)

2. SHR1 vs RAID1 - two different size disk - single disk redundancy for both of them. Same, no mater what do you want SHR1/RAID1 (see in picture below)

3. The unanswered question in my mind is when you use 3 identical disks - same capacity (see in picture below)

What the hell is better in SHR1 vs RAID1?
Basics: RAID is primary about - Redundancy.
In this case RAID1 is more redundant because - the array continues to operate as long as at least one drive is functioning. Then you have still rest of 2 disk as "backup" during degradation process till you rebuild the array to "normal" stage". Then if you lose second disk, you still have third one. Perfect.
In this case the SHR1 is just hanging with single disk as was described in two previous points. Then if you lose second disk, you lose your data.
So, in this case there is an advantage of SHR1 to use twice more available space for Volume vs. RAID1 - but you have just half of the space for redundancy . It is mathematically wrong for me.
The final question is: did you pay cost for the redundancy? Then RAID1 is wort than SHR1. Am I wrong, that any data lose is painful, than cost of single disk in advance?

Of course I need to confirm, that from 3 disks usage the SHR1 becomes (seems to be) relevant to RAID5. But there is a another one problem. RAID1 is faster in case of Read/Write than RAID5. And usage of 3 different disk in SHR1 (different capacity) means your SHR1 will perform than slowest disk in the array.
Then RAID1 golden rule - use identical disks is "must" for every RAID1 build.

And this problem continues in all configuration even in mixed capacity in SHR1 (see in picture below):

I hope, that this topic will open a discussion . Maybe I'm wrong, but hungry for information.

PS: do not forget to backup all worth data. Otherwise data restoration is costly than two disks in small independent NAS for backup, if you are not specialist in this area.

 

[Rusty]

My point is simple. Same drives, convetional RAID. No room for this "uuu I want more space and I'm gonna do it with mix batch". Same drives, same model, same space. Problem solved. If I need to update the array storage vise, I get the new drives and rotate them out, simple. Been doing this for over a decade on all my boxes, no problem.

 

[Telos]

In this case RAID1 is more redundant because - the array continues to operate as long as at least one drive is functioning. Then you have still rest of 2 disk as "backup" during degradation process till you rebuild the array to "normal" stage". Then if you lose second disk, you still have third one. Perfect.

Incorrect. SHR example (3x500GB) has full disk redundancy
You are confusing "redundancy" with unit disk size alone, and ignoring parity. Look at this, for example.

RAID Calculator | Synology Inc.

Use this interactive tool to estimate the storage space of your Synology NAS when paired with different size hard drives and RAID levels.

www.synology.com

More on SHR under the Knowledge Base...

 

[notrui]

I am a SHR novice .
I have a ds418 in SHR mode with two hdds , so far so good.

I have space for more two hdds.
If i have to buy more hdds i am not limited and that is for me one if the reasons that i bought a Synology NAS.

For a home user like myself hybrid raid is great, of course i make backups in other two bay NAS that i have and other places(external hdds, cloud....)

I had before a deskop with 3 hdds in raid5 a long time ago ☺

 

[jeyare]

Thx, Telos I need to update my knowledge , but for business still RAID fan more than SHR

 

[jeyare]

Some questions after studying of Syno KB and rest of Google resources. The KB is written so simple for newbies, but I rather need to understand the participial differences (deep dive info).
Then, here is still open question - SHR(1) vs RAID1:
1. Syno KB statement:
If a Classic RAID array is created with a 500 GB drive, all the other drives within the array can only contribute 500 GB for storage each, therefore creating a 5 x 500 GB RAID volume and wasting the other unused 4.5 TB.
My point: "wasting" is incorrect in case of the RAID fundamental principle - Redundancy
Reason: as I wrote above - in RAID1 - the array continues to operate as long as at least one drive is functioning. Then all of the disks are used for saving of the Redundancy and painful consequences (accidental lose 4 from 5 disks). Specially when there is critical time of RTO = Recovery Time Objective.
I can't find clear information about situation what happen when you lost 4 from 5 disks in SHR(1)

2. Syno KB statement:
Unlike Classic RAID, SHR divides each drive volume into smaller chunks and creates additional redundant storage. By using SHR, you can use the unavailable 4.5 TB volume as smaller usable chunks, therefore maximizing storage capacity of each drive.
My point: then it is not RAID1 from begging, much more RAID5 = because (seems to be) some kind of parity principles. Then where is stored the parity, and how is driven the redundancy reliability?
Performance: then performance from basic point of view must be worse than RAID1, when you use large amounts of write operations = because of parity overhead.

If you have deep inside info, I'll be glad to know more.

PS: Also, a rather complicated structure may become a hindrance to retrieving critical files in case of data loss.
Then I like easy solution, even with wasting of capacity = RAID1.

 

[Gerard]

What would you use for disaster recovery if you lost a drive or the array in a regular raid?

I have the one disk that I pulled out of the SHR array that was throwing an error. I’d be happy to try and test plugging it in and try to recover or see some sort of files, but I don’t know what tools/applications could read from that SHR disk.

 

[jeyare]

@Gerard - I will split your problem to two answers, that can help to everyone also, I hope.

In case of DR we need start from a basics (for DR newbies), except terminology or cases used in Enterprise environment (out of mass market or SME):

1. Service continuity - is one and only of holly grail of DR. Thus when your data is not available, then you have got a kind of personal/business expected continuity. It really doesn't matter when you lost possibility to store your photos to NAS, or you lost mandatory cloud services for your company. It still could be a disaster for someone. I will use in this description as "business continuity".

2. The Recovery Time Objective (RTO) - described by Wikipedia - is the targeted duration of time and a service level within which a business process must be restored after a disaster (or disruption) in order to avoid unacceptable consequences associated with a break in business continuity. In this case = simple = how much time is accepted from your side to repair the problem, in minutes, in hour, in days, weeks, etc.

3. A Recovery Point Objective (RPO) - described by Wikipedia - RPO measures the maximum time period in which recent data might have been permanently lost in the event of a major incident; it is not a direct measure of the quantity of such loss. For instance if the BC plan is "restore up to last available backup", the RPO is the maximum interval between such backup that has been safely vaulted offsite.

Then we can come back to @Gerard question: "What would you use for disaster recovery if you lost a drive or the array in a regular raid?"
Answer is:
A. Chose right backup policy, based on your own data worth valuation. To get expected RPO time value
B. Chose right data redundancy, based on your expected max. RTO (max. time of accepted discomfort).
C. Chose right investment to to right technology, based on your data value. Low cost isn't a synonym for best. Too expensive isn't a synonym for best also. Balanced decision must be taken into the consideration.

Then from DR point of view (still in line for mass market and SME) is guaranteed RAID1 with 2+ disks better than SHR(1):
- SHR is slower than RAID1 - then make an impact to the RTO
- try to remove single drive from SHR and install it in a new non-SHR RAID (for repair, for check, for any reasons), no way, it will need to be formatted. In case of RAID1 you can anytime remove the single disk and make new independent RAID1 in same NAS or another NAS, specially in new (non Synology environment). Then when you accidentally lose all your backups (or you don't have them) and the last single RAID1 disk is your last option, then you have still better chances to keep RTO then in SHR case.

Maybe I'm still non educated into deep SHR principles, but it's hidden to my Google research. If no, then SHR is really usable for Mass market only, but not for SME. Specialy for cases when the time is your enemy = almost anywhere.

 

[jeyare]

I have the one disk that I pulled out of the SHR array that was throwing an error. I’d be happy to try and test plugging it in and try to recover or see some sort of files, but I don’t know what tools/applications could read from that SHR disk.

Second answer:
SHR is just kind of (customized?) Linux SW based redundant technology, then you need mount the disk file system in Linux environment.

Then you need:
- MDADM - is a Linux utility used to manage and monitor software RAID devices
- Logical Volume Manager (LVM), in this case LVM2
use this KB link from Synology. If you don't have Linux, then you can run Linux in virtual environment:
- download Oracle VM box - link
- read User documentation - it is really simple
- prepare simple Ubuntu machine, like in this YT
- then use this KB link from Synology, to mount the disk.
Then you will see your disk as mounted - Except all imaginable issues in your disk - welcome in this simple disk repair world.
If you don't know more about file system structures or disk operation basics is better to finish such try&fail kind of attitude and ask someone with an experiences. Because you can damage more than you have now.

 

[Gerard]

@jeyare

Let me simplify and clarify a bit. In our discussion of the previous thread you had mentioned your curiousty of plugging a single SHR drive in to attempt to recover data from the drive (at least that’s what I thought I got from it); thus in comparison to regular raid such as raid 1 you can plug a drive in and browse the folder/file structure.

All I was looking for here is an application or program where I can plug in the error prone SHR disk that is now trash and play around to see if there is anyway to extract files off of it. This would be experimental and is not needed. I have disaster recovery protocols in place utilizing onsite/offsite backups and even regional country backups in case a severe storm knocks power out.

 

[jeyare]

@jeyare
All I was looking for here is an application or program where I can plug in the error prone SHR disk that is now trash and play around to see if there is anyway to extract files off of it.

Sorry @Gerard, don't know about such simple tool, with GUI and successful usage.
The KB link from Synology is simplest known way how to do it, regarding of FS with SHR mount.
Except situation, when you have damaged HDD/SSD data, e.g.:
A) data in damaged blocks. Solution for etx4 FS is here , but I like TesDisk for all know partition used (still CLI only). For BTRFS is a checker here and guide how to recover the BTRFS partition.
B) damaged HDD parts - heads, read/write arms, ... then you need open the disk, and really know what to do
C) damaged SSD parts - never tried by me

PS: one of the frequent HDD HW problem is, that head arms aren't in Parking position. It can be done be sensitive fingers and tweezers. But as I wrote, be aware - Not all fails are identical, or have identical fixes.

 

[Telos]

I have the one disk that I pulled out of the SHR array that was throwing an error. I’d be happy to try and test plugging it in and try to recover or see some sort of files, but I don’t know what tools/applications could read from that SHR disk.

Answered in your other post

 

[durin]

The original post seems to reflect a misunderstanding of how SHR works. An SHR volume based on two HDDs is effectively RAID1. Once you expand SHR volume to three or more HDDs, it becomes analogous to RAID5.

Moreover, unlike RAID5 you can use different size HDDs without wasting space provided you have at least two of the largest HDDs in an SHR array. E.g 2x2TB and 2x1TB HDDs in a RAID5, it will treat all drives as 1TB drives (I.e 3TB of storage, 1TB for parity and 2TB of wasted space), while SHR will effectively treat them as 4x1TB in RAID5 + 2x1TB in RAID1 (i.e. 4TB of storage, 2TB for parity).

That to me is SHRs main advantage over RAID - greater flexibility. The flexibility of SHR was one of the reasons I moved over from a QNAP NAS many many years ago.

And why anyone would use a RAID1 volume with more than three HDDs is beyond me. Redundancy is not a substitute for backups.

 

[jeyare]

welcome @durin, to be sure:
- the RAID1 to RAID5 transformation, it has been mentioned and it is clear
- no doubt, there is no mention about backup, there is redundancy driven topic only
- much interesting is RTO driven target of such discussion between RAID1 and SHR(1) mentioned in this thread

 

[durin]

I’m not aware of much convincing evidence that there is a difference in RTO between SHR and RAID - you should have backups that you can rely on for recovery, rather than recovering data from a failed array.

 

[jeyare]

@durin you still hanging in case of data recovery, but the RTO is in most cases about how much time you can afford with service disruption repairing (read carefully #8 post). Service disruption pain is different from each point of view:
- for home user with Plex library it can be from one day to one month
- for home user with more important data in days
- for SME with high data loads by Drive up to hours
- etc.

Then back to the main difference between RAID1 and SHR(1) with setup of 3+ disks - in case of RTO:
1) when you lost one disks, your array is degraded
2) it means, the performance of data R/W is seriously injured (partially service disruption/degradation based on this array availability)
Then you have to start spare disk initializing and Raid restoration = transfer data to new disk from rest of them. It takes a time:
a) RAID1 just clear data transfer
b) SHR(1) data transfer include parity calculation
from general point of view is RAID1 faster in such array restoration. But the time, time can save your mental healthy when you during array restoration lose another disk.

As you can see, there is still not mentioned backup, because the backup you will use in last point, when you lose all the array restoration steps. Reason = backup restore time (still about service unavailability, is RPO driven time) in this case is really longer. Because you have need fully operational array, not just degraded. Then you can restore the data to valid array.
This is the story about RTO.
If you have degraded array with Docker containers for daily SME operation, it is more important - thinking about the RTO.
PS: of course you can solve it by HA. But it still about time and cost.

 

[jeyare]

in any cases of RTO driven cases, my recommendation:
- if you don’t invest to HA (additional NAS), it is not a shame
- you can rebuild existing NAS operation architecture with single or few spare disks (up to your critical service), initialized and prepared for immediate action (degraded array repair). But you have to count with point, that you lose one or more bays in the NAS.
- or in future replacement of your old NAS don’t forget for the spare disk bays.

Still need to thinking about not just today’s, but future operations expectations.
Sufficient preparation is base of better sleeping.

 

[durin]

At this point this is not really an issue about SHR vs. RAID1, but rather a more general discussion about multi-HDD RAID1 vs RAID5 (since that is effectively what a multi-HDD SHR drive is).

I have managed small business networks in my past life, and to me talking about a HDD failure as a 'disaster' that needs detailed recovery planning is overkill. HDD failure is a commonplace occurrence, and the point of RAID (and by extension SHR) is about ensuring there is continuity of service after such a failure thanks to redundancy.

When you start talking about real disaster recovery planning that involves planning for RTO, particularly for small businesses that don't have huge amounts of resources to devote to such planning in the first place, we are talking about something far more serious than a trivial HDD failure. It's about recovery after a major disruption (e.g. flooding, fire) where the presumption is that the primary site (which for a small business is often where the server is located) is severely disrupted and you will need alternative ways to restore operations - either using remote mirrored servers (or these days cloud services) if it is absolutely essential to ensure continuity of operations, or recovery from backups for services that are not as time critical.

Sure, there may be some cases where it may be worth to spend far more money to build a multi-HDD RAID1 array. But the reason for the popularity of RAID5 (which is effectively what a multi-HDD SHR array is) is that it offers an relatively inexpensive way to offer a storage solution that still offers redundancy which (combined with backup) is more than adequate for most uses.

Now, if your mental health is going to suffer from waiting for parity calculations while restoring an SHR/RAID5 array rather than a RAID1 array, then by all means go with the RAID1 solution.

 

[jeyare]

Thx @durin. Maybe the forum is not right medium for a long philosophical discussion, but still like to see many point of views. That was a reason of topic in this threat = Is SHR better than RAID and When?
and back to the topic - to be sure - for 3 and more disks, because in case of two disks it is same as we know. The main question was changed, when you read carefully in my first post:
What the hell is better in SHR1 vs RAID1?

Summary:
- data availability (physical/logical availability, speed of data store) is driven by business service continuity expectation (one of them is max. disruption time = RTO)
- then disaster recovery is part of the business service continuity plan. No doubt.

There is still valid previous wording about the service disruption:
RTO is the targeted duration of time and a service level within which a business process must be restored after a disaster (or disruption) in order to avoid unacceptable consequences associated with a break in business continuity.
Yes of course, we can speak about reasons of the disruption:
- natural hazards, (a political is worse )
- ...
- and also any cases of the array degradation what make an impact to unacceptable consequences of business service continuity.

As was mentioned in previous, there is the HA option as BC/DR solver (is running between my two 1813+ for my SME). In many cases the HA is must, no doubt. And of course we can speak about geo-clusters also. But it is base for another topic.

Last point:
- yes, the SHR1 with 3+ disks is RAID5 alternative
- yes you can get more available space vs RAID1 (for 80% of users who can't make data grow plan, or differentiate volume operation from purpose point of view)
- but you lose the previous SHR1 W/R operation speed (with 2 disk array)
- and also you lose time of the array recovery (volume) = perception of un/acceptable service availability, what is different perception for each of us (from clear target to naivete).
And this is maybe (from my point of view) one an only advantage or disadvantages list of the SHR1 vs RAID1.

 

[itsjasper]

The main benefits of SHR (other than set and forget):

1) Flexibility in being able to use differing drive sizes in the group.
This isn't really a good idea because of different drive speeds, ages etc,. So it is mainly a benefit to the severely cash-starved, or those wishing to utilise existing sets and upgrade over time

2) Flexibility in enabling slower drive replacement timeframes whilst still enabling upgrades.
A long time ago I ran a DS1814 / DS1813 pair with both filled with 4TB drives. My budget allowed for me to replace two drives a month (back then 4TBs were around the price of 12TB drives today). At that cost/pace, it would have taken most of the year to use any capacity uplift, had I chosen RAID over SHR.

Totally agree with Rusty's point about swapping out entire drive sets, though that can get expensive real quick with larger drive sets / larger capacities.