A NAS is NOT a Backup!

  • Thread starter Deleted member 480461
  • Start date
D

Deleted member 480461

Guest
A backup is where one has a second instance of there data. This second instance could be on the same hard disk as the original, or onto a completely separate disk. For example, one may backup their documents to a second folder, or a second partition, on the same drive as the original. This is still a form of backup.
Is it a good form of backup? No. No it's not.
Why? Because if that hard disk should fail you stand to lose your original, as well as your backup - everything. Gone.

How can we solve this?
We backup to a different HDD, or a different media, to the original source.

Ideally you would backup your data onto a separate drive, be it an external hard drive, or onto tapes via a tape drive. You could even backup onto a second hard drive within the same machine if you wanted - it would at least cover you from your source HDD failing.
With each of these solutions we would be able to restore our source disk and retrieve our data.

More and more households are moving towards centralised data storage. This in itself opens many new doors and opportunities.

We have those who use their storage devices, such as NAS boxes, as a centralised backup medium. All computers/laptops in the household would be backed up to that NAS/server. In the event of a system failure, data would be secure and could be retrieved.

On the other hand, we have those who store all their movies, music, documents, precious photos, and so on, onto their NAS boxes so that all data can be accessed by all clients who are granted permission to access the resources - be it locally, or remotely. Since many of these NAS units, such as those from QNAP or Synology, incorporate RAID arrays, the data is completely protected should the system fail, right?

WRONG.

There are too many people on these forums who choose a NAS unit with the thought of complete data safety. Don't get me wrong, RAID is a great concept. You choose a level of RAID to suit your needs.

Need the fastest access possible? Stripe.
Need a duplicate of your data? Mirror.
Want a mix of both? RAID5.

RAID 5.
That's the big one. That's the one everyone is after. That's the one that keeps your data safe, right? Again, wrong.

So you buy your NAS. You buy your 3+ HDDs so you can set up a RAID 5 array. You put all your data onto the NAS and relax, knowing your data is safe.
But it's not.

RAID 5 is fault tolerant. Without getting into the workings of this level of RAID, it will allow a SINGLE disk in the array to fail. So that's one out of your 5 HDDs that can fail, and you can relax knowing your data is still there. You still have access to it. You can still use the NAS as normal. You replace that single disk and the array rebuilds, and you are back to where you started.

Great, huh?

Now let's say you have a major power outage and the NAS is abruptly switched off with no battery backup (UPS).
You switch the NAS back on to find 2 or more disks have failed. Worse still, the NAS won't even power up.
What now?
Has the super duper RAID5 array saved you now?

No it hasn't.
BECAUSE RAID 5 IS NOT A COMPLETE BACKUP IN ITSELF! All your data. Them precious photos. The memories. GONE.
If you use the NAS as a means to backup the data on your main computer then you can rest knowing you still have the main source of your data safely on your PC. If, like me, the NAS is your data source, then you're in trouble!

Hopefully I have proven my point here and have shown you that although it's a great concept and it's use is encouraged, its not a fail safe solution. Although most disk failures do not usually occur in 2's or more, it's usually a single disk at a time, it can still happen. And we've seen it happen through many members here in this very forum.

So what should we do?

Most NAS units have either USB or eSATA ports in the rear of the unit. These can be used as storage expansion or as a means to backup. You could even purchase a second NAS to backup the first! Simply add a drive to your NAS locally, and your NAS will be able to backup your important data straight to it. Hence we create a copy of the data on our NAS to a separate external unit. .

Is this enough?
For most, it is. But what if, God forbid, your house burns down in a fire? That's still all your data gone.

Hence we need to store an OFFSITE copy of our data. An offsite backup.
Take your backup disk, the backup of your NAS, and store it at your parents house, at work, in the garden shed, at a friends - anywhere that's away from the main data source.

Why stop there? Create more backups!

I personally have all my data stored on my NAS. I regard my most important data to be my documents, my photos, and my home videos. Movies can always be redone or repurchased. Hence my important data is backed up to a separate external drive that sits right next to my NAS. Should I lose any data, or should my NAS fail, I still have all my data ready to be restored.
But wait, that's not enough is it? I just said it sits next to my NAS!
Therefore I have ANOTHER backup disc with another copy of the backup, which stays in my desk at work!

Covered from all angles (unless both places burn down simultaneously, which is unlikely, right?)!!

I cannot stress enough. RAID is NOT a backup solution. THINK. Just spend that little bit extra on a backup drive and cover yourself from any unfortunate situations!


NOTE: I've written the above while hunched over my iPad. My back is hurting. I may have made several typos, and for that I apologise in advance. But if I have encouraged at least 1 person to get that backup drive and do the right thing, it will have been worth it!
Right, now off to straighten my back! :smashin:
 
Last edited by a moderator:
That's a BIG soapbox. :boring:
 
He not wrong though - the house burning down is an extreme example but burning your most vital data to disc and leaving it at a parents or mates house or even in a draw at work is a good idea, If you havent got it in the cloud somewhere already. If the worse happen imagine all your wedding photos, baby pictures, holiday photos or whatever gone forever - screw the movies that recoverable in long term its your personal stuff thats going to leave a whole in your life if you lose it all.
 
I learned the hard way, but fortunately at an early stage with only a small amount of data...........

All my NAS content is now backed up on a different moveable HDD....:D
 
I felt my eyelids dropping reading that :laugh: but he does have a point.
I myself store all my data on the Nas and access it from the laptop and desktop machines.

but i have a plan, i just ordered a HDI Dune Duo that has 2 drive bays :D maybe i can set one of them as a backup target for the NAS.

Its not here yet so i cant try :facepalm:
 
Hi,

It may be boring and mundane, but you really need to consider your backup strategy because sooner or later you will lose a drive, along with all the data on that drive. It may not even be drive failure. I am currently doing a recovery job for a client who accidently formatted his HDD which contained documents which weren't backed up, so I now need to recover the documents which he has worked on since his last full backup (3 months ago).

You really do need two copies of everything, for example you need the originals, then you need a copy on another physical drive (preferably a drive which is only used for backups and turned off when not in use so it cannot be damaged by viruses and malware) and finally a copy on optical disc at least stored in another room, but another building is ideal.

I have been involved in enterprise IT for more than 20 years, and I have had to resort to backup restoration on more occasions than I want to think about.

I agree that RAID 5 helps, but it is falible, and the error correction algorithms break down once you start dealing with volumes over 4TB, so you need RAID 6 at that point, and even that can fail if your controller card fails and you don't have a readily available replacement.

I can only reiterate the original point, a RAID NAS is not a backup, and even a mirrored system is not a true backup, you need a guaranteed uncorrupted data copy, so you need to think near line disc (one which is turned on only during backup or restoration) or optical disc.

Regards,

Dave
 
I am suprised people are criticising - it is a good knowledge source for some less experienced people.
 
Those that critizise the original poster are obviously niave to just how NAS devices are vulnerable, and not completely reliable.

I have a QNAP 509 Pro, not a cheap product - the first unit was replaced because of a inherent design fault with it's Marvell HDD controller chipset, and I have only jet received the replacement back from them, after the internal power supply crapped out on me.

My drives contain all the usual suspects - movies, photos, music albums, all taken for granted until they are gone.

So, for those who are trying to look clever on this public forum - think on!

I'll climb off my soapbox now, before I get accused of being on one! :rolleyes:
 
Ahhhhhhhhhh. Someones lost some data! :p

Well if it pleases the OP. My server is in my garage off site of the house. Linked by wired cat5 to switch/router in loft (it's a bungalow :) )

My mirrored backup (for every drive I have a clone :) ) is kept offsite at my parents house!

Oh and I don't use RAID at all. Nope. Nada. Zip! :) trust it as much as deskstar hard drives ;)

Mental how a 4gb hard drive was "loads" in 1999 and I'd "never fill" 120gb in 2003 and now I'm on the look out for another 2tb lol!

Oh and IMO I think that 2tb should be the limit of a drive size. A 3tb drive Besides windows bein awkward about it, is way tooo much data to lose in one hit! What do you think?
 
jammyb2004 said:
Ahhhhhhhhhh. Someones lost some data! :p

Well if it pleases the OP. My server is in my garage off site of the house. Linked by wired cat5 to switch/router in loft (it's a bungalow :) )

My mirrored backup (for every drive I have a clone :) ) is kept offsite at my parents house!

Oh and I don't use RAID at all. Nope. Nada. Zip! :) trust it as much as deskstar hard drives ;)

Mental how a 4gb hard drive was "loads" in 1999 and I'd "never fill" 120gb in 2003 and now I'm on the look out for another 2tb lol!

Oh and IMO I think that 2tb should be the limit of a drive size. A 3tb drive Besides windows bein awkward about it, is way tooo much data to lose in one hit! What do you think?

It all depends on what you want, and how much the data is worth to you.

At work we use dozens of storage arrays, all with massive amounts of cache, storage tiering, various flavors of RAID, internal mirroring, disk sparing etc. On top of that we use snapshots internal to the array, mirroring at the host level to two or more arrays, automatic scheduled nightly tape backups onsite and offsite, automatic scheduled simultaneous/delayed physical replication onsite and offsite, and the occasional manually-triggered backup prior to major actions. All arrays have "call home" functionality and 24x7 four-hour repair contracts. All of these are designed to cope with different scenarios of both physical data loss and logical data corruption. But then if we lost that data most of the airlines in Europe would grind to a halt, and I'd probably be out of a job, so it's worth it :)

At home a Qnap 419 P+ using RAID 5 and a weekly backup to bluray (to be stored in my desk at work) suffices. I only back up the personal photos/data because the films/music would be re-creatable (although it would be a right royal PITA).

The OP is right, RAID is not a backup solution, but for most people it probably seems like it is. At least it's better than nothing, and posts like this one might persuade a few people to do something more, so well done fella!
 
Last edited:
But then if we lost that data most of the airlines in Europe would grind to a halt, and I'd probably be out of a job, so it's worth it :)


The big difference is that in the business world, loss of access to data can be just as crippling as the loss of that data, even if it's not as permanent.

Businesses pay a lot of money to try to ensure continuous (and fast) access to data - that's really where RAID comes in.
Backup/restore is another problem with a different solution.


For the home user, RAID is not a backup, but as you said, it's better than nothing - the biggest (though not only) cause of data loss in non-RAIDed storage systems is the failure of the storage devices themselves (ie the disks) - at least you are protecting against that to some degree - RAID won't protect your data against anything else though!
 
MikeK said:
But then if we lost that data most of the airlines in Europe would grind to a halt, and I'd probably be out of a job, so it's worth it :)

The big difference is that in the business world, loss of access to data can be just as crippling as the loss of that data, even if it's not as permanent.

Businesses pay a lot of money to try to ensure continuous (and fast) access to data - that's really where RAID comes in.
Backup/restore is another problem with a different solution.

For the home user, RAID is not a backup, but as you said, it's better than nothing - the biggest (though not only) cause of data loss in non-RAIDed storage systems is the failure of the storage devices themselves (ie the disks) - at least you are protecting against that to some degree - RAID won't protect your data against anything else though!


Well, for a real-time business, loss of access to data is effectively the same as loss of data, in the short term at least - customers don't care which problem youre facing. RAID is simply a mechanism to give you time to replace failed hardware, because for sure it will fail. But it most definitely isn't used for fast access to data - that's what caching, striping, storage tiering and fast networks are for.

As I said, the various mechanisms are used to protect against various scenarios, as interlocking components in a proven business continuity strategy. Belts, braces, string and both hands to hold up your trousers :)

But for the home user, you don't need anything more than RAID and an offsite backup, in my view. That combination will get you through 99.9% of any issues you're likely to face, at reasonable cost.
 
Last edited:
RAID is simply a mechanism to give you time to replace failed hardware, because for sure it will fail. But it most definitely isn't used for fast access to data - that's what caching, striping, storage tiering and fast networks are for.

Is striping not RAID0? :p
And is RAID1 not also faster than single drives in most cases?
Even RAID5 is faster than single drives in most cases.

RAID, in several of it's various guises, most definitely is used to improve speed of access to data - though it's not the only method employed.
It's arguably not the primary purpose of RAID either (well, apart from RAID0), which is why I put the "and fast" bit in brackets.


But for the home user, you don't need anything more than RAID and an offsite backup, in my view. That combination will get you through 99.9% of any issues you're likely to face, at reasonable cost.

Fair enough - though you may be surprised at how many businesses adopt this approach too :)
 
MikeK said:
Is striping not RAID0? :p
And is RAID1 not also faster than single drives in most cases?
Even RAID5 is faster than single drives in most cases.

RAID, in several of it's various guises, most definitely is used to improve speed of access to data - though it's not the only method employed.
It's arguably not the primary purpose of RAID either (well, apart from RAID0), which is why I put the "and fast" bit in brackets
True :) But with kit like SSD and FusionIO cards, the disk throughput is no longer the primary bottleneck. Usually (for our setups anyway) the storage front-ends are the slow part, and that includes the RAID controllers. When I said striping I meant deliberately (at an application or OS level) spreading data across spindles to reduce hotspots, not letting RAID just randomly spread it across multiple disks. But most new storage arrays now do this automatically anyway, so it's less necessary than it was, and solid state has totally changed the game. In any case, we used RAID for fault tolerance, not performance, and I don't think we're alone in that in high-end business.

MikeK said:
Fair enough - though you may be surprised at how many businesses adopt this approach too :)
That's absolutely true, especially small businesses, who probably don't realize how valuable their customer list is, until it's gone :-(
 
I'll be honest that I've skim read most of this thread as some posts are very long, but my comment is that most of the posts seem to be talking about RAID and NAS not being a back up for home users. The only thing you need for a backup is a separate copy (preferably more than one, and preferable kept off site) of the data you want keep safe. Whether that second (or third, or fourth) copy is on another PC, DVDs, USB hard drive, thumb drive or NAS (RAID0,1,10,5 or 6) it doesn't really matter.

The short of it is that to back up data there needs to be at least one copy, preferably more and preferably kept off site and tested regularly. The next discussion is about reliability/performance of that backup, and/or the original data store and that is where the arguments come in over single/multiple disks, RAID this or the other etc.

The title is also wrong imo as a NAS can be a backup, but it isn't if it's the only copy. When people ask about NAS for home use in this forum it is usually mentioned that a backup is also required. Maybe we've been slipping so it's goog that people have been reminded out it again.
 
I'd certainly agree that flash technology is rapidly becoming a game changer in the storage arena, certainly from an overall performance perspective.


In any case, we used RAID for fault tolerance, not performance, and I don't think we're alone in that in high-end business.

Perhaps not alone, but others might consider both to be important - hence why many would configure with eg RAID1+0, and take the capacity hit rather than go with RAID5 or RAID6 (and derivatives of such) etc.

Of course, in high end storage arrays, the cache, and more recently the second level cache (flash) can often "hide" the underlying performance of the disk subsystem anyway, and such arrays also often have tools to assist in identifying and alleviating potential performance problems (this kind of thing is partly what makes them "high end" though)
 
Hi,

KeithO: I have to ask, are you using Sun / Oracle ZFS in that mix, or are you using some dedicated SAN's? Is your storage tiering automatic or does it require manual intervention for you to tier between spinning rust and SSD?

I ask simply because I manage a much smaller enterprise installation where we were faced with similar problems to you, and we're using Sun's ZFS system which I find to this day amazing in what it can do, and the ease with which it can do things such as tiering, snapshots and even management and repair of arrays.

Thanks and best wishes,

Dave
 
I'd certainly agree that flash technology is rapidly becoming a game changer in the storage arena, certainly from an overall performance perspective.


In any case, we used RAID for fault tolerance, not performance, and I don't think we're alone in that in high-end business.

Perhaps not alone, but others might consider both to be important - hence why many would configure with eg RAID1+0, and take the capacity hit rather than go with RAID5 or RAID6 (and derivatives of such) etc.

Of course, in high end storage arrays, the cache, and more recently the second level cache (flash) can often "hide" the underlying performance of the disk subsystem anyway, and such arrays also often have tools to assist in identifying and alleviating potential performance problems (this kind of thing is partly what makes them "high end" though)

Well, we're coming from a largely mainframe-based environment where we painstakaingly placed every record on the platter, and used less than 10% of the surface so we could keep head access close to the spindle, so "capacity hit" is relative here.

You're right in that cache can mask a lot of the disk performance, but only in a burst situation. If you fill the cache, you're back to disk i/o throughput as the bottleneck.

Despite the tumbling "per Gb" prices, storage is far and away our biggest investment these days. SSD helps, and will continue to get quicker, but we're already looking at building our own appliances based on FusionIO and Infiniband to alleviate bottlenecks on high-traffic apps.
 
Hi,

KeithO: I have to ask, are you using Sun / Oracle ZFS in that mix, or are you using some dedicated SAN's? Is your storage tiering automatic or does it require manual intervention for you to tier between spinning rust and SSD?

I ask simply because I manage a much smaller enterprise installation where we were faced with similar problems to you, and we're using Sun's ZFS system which I find to this day amazing in what it can do, and the ease with which it can do things such as tiering, snapshots and even management and repair of arrays.

Thanks and best wishes,

Dave
We have some Sun machines, but not by choice :D

Primarily it's HP Superdomes running HP-UX and Oracle for critical databases, but more and more we're moving to Linux and Oracle on x86. Partly that's because the x86 architecture is catching up fast on performance, but also because Oracle are clearly no longer supporting HP-UX like they used to.

Storage-wise it's dedicated SANs, based on various kit. As I mentioned above to MikeK, we're coming from mainframe so that's IBM DS8K, because the OS isn't certified for anything else! For the rest it's a mix of IBM XiV, HP XP, and EMC VMAX.

Pretty much all of them do automatic striping now (tho they don't call it that, to distinguish it from the basic striping of old - IBM for example use the term 'spread' but they all use it to avoid hotspots on the disk).

Tiering is also automatic, within reason. We define certain data we want to stay on SSD, or on SATA, or can move dependant upon access, and the array manages the movement.

Also as mentioned to MikeK, we still see limitations of the array front-ends, and the size of available cache, so we are looking into building our own storage appliances based on FusionIO cards and Infiniband. This will be used for really performance-dependant stuff, like Oracle logs which are sequential and synchronous and therefore ultimately a block on throughput.

If you're interested, take a look here http://www.amadeus.com/amadeus/x144757.html
 
Last edited:
Well, we're coming from a largely mainframe-based environment where we painstakaingly placed every record on the platter, and used less than 10% of the surface so we could keep head access close to the spindle, so "capacity hit" is relative here.

Cor blimey, do people still do this ;)


You're right in that cache can mask a lot of the disk performance, but only in a burst situation. If you fill the cache, you're back to disk i/o throughput as the bottleneck.

[best Yorkshire accent] - then add more cache lad :D


Despite the tumbling "per Gb" prices, storage is far and away our biggest investment these days. SSD helps, and will continue to get quicker, but we're already looking at building our own appliances based on FusionIO and Infiniband to alleviate bottlenecks on high-traffic apps.

It's always an option, but a DIY approach can go pear-shaped for a variety of reasons - product testing (in terms of the appliance) will likely be limited in comparison to system/storage vendor options.

Overall though, you sound like an Exadata customer-in-waiting ;)
 
MikeK said:
Well, we're coming from a largely mainframe-based environment where we painstakaingly placed every record on the platter, and used less than 10% of the surface so we could keep head access close to the spindle, so "capacity hit" is relative here.

Cor blimey, do people still do this ;)

You're right in that cache can mask a lot of the disk performance, but only in a burst situation. If you fill the cache, you're back to disk i/o throughput as the bottleneck.

[best Yorkshire accent] - then add more cache lad :D

Despite the tumbling "per Gb" prices, storage is far and away our biggest investment these days. SSD helps, and will continue to get quicker, but we're already looking at building our own appliances based on FusionIO and Infiniband to alleviate bottlenecks on high-traffic apps.

It's always an option, but a DIY approach can go pear-shaped for a variety of reasons - product testing (in terms of the appliance) will likely be limited in comparison to system/storage vendor options.

Overall though, you sound like an Exadata customer-in-waiting ;)

Yep, people still do that if you need the throughput - which we do :)

In many cases the cache is maxed out, so we can't add more.

Your comments on testing of a self-built appliance being limited compared to vendors is interesting though. In our experience, kit never (and I mean never) performs as per specification. The vendors always have excuses (sorry, I mean explanations) but the upshot is that lab testing never matches real-world conditions. A purpose-built appliance, designed to meet a specific set of parameters, and tested and verified in the environment in which it will be used, will always be better than generic kit. The problem of course is support, but we work very closely with vendors on that.

Exadata is interesting. We haven't got any yet, but we've looked closely at it. I'm sure we will get some, but for less-critical databases. The major problem with Exadata for customers is that it's a black box. You can't independently patch the DB or the OS, you can only apply Oracle-supplied patches. You can't tune the DB for specific workloads and you can't decide to not apply irrelevant OS patches. That's a big benefit for Oracle of course - they know that any customer calling their support has exactly the same hardware, software, firmware and patch levels as every other customer, and as their lab. But it's a big problem for us, because that "one size fits all" approach just doesn't cut it. It also has a lot of problems architecturally with cross-firecell clusters and the like.

If you break open an Exadata you'll find an x86 processor, running OEL, with an Oracle 11 DB engine and Sun storage, all connected by Infiniband. What we plan is to effectively build our own Exadata, but one we can tweak. We're not the only ones doing it.
 
I agree with the original poster, I've had 3 1.5tb seagate hdds fail on me within 2 days. My games, videos were gone as well as my photos and family vids these were really important and I made copies on 50gb bdrs discs.

I Learnt the hard way 8 years similar thing happened and I lost every thing. If I were to lose my photos and vids I would have be very unlucky
 
Your comments on testing of a self-built appliance being limited compared to vendors is interesting though. In our experience, kit never (and I mean never) performs as per specification.

I meant product testing from an operational/reliability perspective, not really performance.
On that subject though, many customers miss the "up to" bit in the performance specs ;)



The major problem with Exadata for customers is that it's a black box

That depends on your viewpoint I suppose - a lot of customers might say that's a major advantage.
I'd agree though that it may not always be the best solution in all circumstances.


What we plan is to effectively build our own Exadata, but one we can tweak. We're not the only ones doing it.

You can certainly build equivalent hardware wise, but Exadata is more than just the hardware.
Support might be the primary obstacle though.


How the hell did we arrive at discussing Exadata anyway, in a thread about backing up a home media NAS? :confused:
It's easy to get carried away on here!! :D
 
Last edited:
MikeK said:
How the hell did we arrive at discussing Exadata anyway, in a thread about backing up a home media NAS? :confused:
It's easy to get carried away on here!! :D
Yeah we did take it a little off-topic didn't we ?! Interesting conversation though, thanks! :)
 
that's why i backup/sync some of my important folders on the nas to the cloud such as dropbox
 
Last edited by a moderator:

The latest video from AVForums

Is 4K Blu-ray Worth It?
Subscribe to our YouTube channel
Back
Top Bottom