129

When someone mentions RAID in a conversation about backups, invariably someone declares that "RAID is not a backup."

Sure, for striping, that's true. But what's the difference between redundancy and a backup?

14 Answers 14

152

RAID guards against one kind of hardware failure. There's lots of failure modes that it doesn't guard against.

  • File corruption
  • Human error (deleting files by mistake)
  • Catastrophic damage (someone dumps water onto the server)
  • Viruses and other malware
  • Software bugs that wipe out data
  • Hardware problems that wipe out data or cause hardware damage (controller malfunctions, firmware bugs, voltage spikes, ...)

and more.

6
  • 3
    Will a backup refuse to copy a corrupt file?
    – jldugger
    May 2, 2009 at 0:53
  • 19
    Depends on what "corrupt" means but normally backup applications have a setting for this... however, the second point of backup is to keep different versions of the file through time - not just a single version - thus circumventing the problem with a newly corrupted file overwriting a fresh version... May 2, 2009 at 1:00
  • 3
    > Will a backup refuse to copy a corrupt file Yes, if you cannot read the blocks of a corrupt file, you won't be able to make a copy of it (backup) May 3, 2009 at 3:22
  • 2
    But what about silent data corruption; if a data block goes bad, most filesystems won't notice, will they?
    – jldugger
    May 3, 2009 at 18:07
  • 12
    Reasonable backup strategies include keeping a history, so that you can go back to before the corruption. The most common handling of the possibility of corruption is to pretend it can't happen. But if you want to protect against it, you can attempt to detect it as soon as possible, and in varying chunk sizes (device block level, database page level, file level). If you detect data corruption fast enough, it isn't "silent" data corruption anymore and you have a chance of recovery.
    – carlito
    Jun 1, 2009 at 20:04
129

Q: Why is RAID not a backup?

A: Because the whole purpose of a RAID is to make sure that nothing in the world can interrupt that accidental rm -rf / (or DELTREE /X C:\), not even yanking the power chord in panic.

Q: But whats the difference between redundancy and a backup?

A: If you accidentally overwrite your PhD thesis with garbage, redundancy ensures that you have multiple copies of garbage, in case one gets bad. A backup ensures that you can restore your PhD thesis.

(And an archive ensures that you can retrieve multiple older versions of your thesis, and a version control system also tells you why you made a new version in the first place.)

1
  • Man, I had to read the first answer three times to actually get its meaning. lol Feb 7, 2021 at 22:55
31

Redundancy protects you against your hardware failing. It does not protect against user error, nor against malicious activity (e.g., crackers getting into your system).

See: Why Mirroring is Not a Backup Solution for a hard-earned lesson.

3
  • 8
    Nor software bugs, which are more common than malicious activity.
    – jhs
    May 3, 2009 at 14:39
  • It's an interesting bit of irony that the article linked from that Slashdot page has now disappeared off the web. Not even the Internet Archive provides a meaningful copy; even though they did crawl the page shortly after the Slashdot article date, their copy simply says the page was not found.
    – user
    Dec 24, 2013 at 15:11
  • Nor memory errors, which why you need ECC.
    – inf3rno
    Jul 8, 2017 at 8:47
24

The number one reason you want a backup is not because the physical media died (this is rare), but because of some error that caused the data to be lost or corrupted.

RAID doesn't protect you against a file being deleted.

RAID doesn't protect you against a file being overwritten.

RAID doesn't protect you from your system being compromised and all of your data being overwritten, deleted, or corrupted.

RAID doesn't protect you from your ops team accidentally paving a machine with important data on it.

RAID doesn't protect you from a foolish DBA running a drop command on the production server (mistaking it for a test environment).

RAID doesn't protect you if the building burns down.

P.S. http://ma.gnolia.com/. This is what can happen if you don't have good backups. Your site is snuffed out of existence (note: this tends to be bad for business).

4
  • 1
    So you need to build another building just for the backups. Trolololo. :D
    – inf3rno
    Jul 8, 2017 at 8:49
  • 2
    @inf3rno it turns out that others have already built many other buildings.
    – Wedge
    Aug 25, 2017 at 22:24
  • 2
    I don't think http://ma.gnolia.com/ is quite what you meant to link to...
    – user
    Oct 1, 2017 at 13:51
  • @inf3rno someone already done that, and made a business of it.
    – RonJohn
    Jul 15, 2023 at 12:48
14

Redundancy is great if one of your disks fails. It's no so great if your computer gets a virus, or you mistakenly delete a file, or you need to restore the disk to a previous version for some other reason. That's when you need a backup.

RAID helps you recover from failures, but backups let you go back in time.

10

It should also be mentioned that a hardware fault in the raid controller can easily corrupt the data on all attached disks. So while you reduce the danger from disk failures you add the danger of raid controller failures.

7
  • Multiple rotating copies
  • Geographic redundancy

Asked in a comment to the accepted question:

Will a backup refuse to copy a corrupt file?

Even if a backup copies corrupt or bad data, the point of a backup is that you can and should have multiple copies. For instance, last hour, yesterday, last week, etc. You can get a similar effect from using rotating snapshots on your storage device.

But the other reason for backups is geographic redundancy. You should certainly keep copies of critical data in two different geographic locations. How separate those locations are depends on how critical the data is; keeping copies in two different buildings in the same city protects against fire or theft. Keeping copies in two different countries protects against bigger problems.

2
  • +1 for the value of geographic disparity.
    – killermist
    Mar 17, 2015 at 22:35
  • Great answer but I would really like to dig deeper the "bigger" problem :) What exactly are clasified as the problems
    – Wolf G
    Jul 5, 2018 at 12:29
6

RAID can be a great way to mitigate risks due to hardware failures, but RAID won't help you when your users delete (accidentally or otherwise) their data. To recover data you need some archival facilities, either through local snapshots or online/offline backups.

3

In a RAID5 array, consisting of disks over 400Gb, if you lose a disk there's something like a 75% chance of having an unrecoverable read error while the array is being rebuilt. Think about that for a second and it becomes pretty obvious why someone will always remind you that "RAID is not a backup".

RAID gives you higher reliability and performance, but it's not infallible.

1
3

Fire, theft, RAID controller fault, human error, the list goes on

3

What's the difference between redundancy and backup? Ok, configure a RAID 5 disk set. Store some business-critical stuff on it. Pull a disk out. Everything still works! That's redundancy. Now delete all the data (don't cheat with the recycle bin). Now restore it from the most recent backup. You don't have one? Oops. Well at least you can tell your boss your disks are using RAID 5 redundancy (as you get marched out of the building...)

1
  • The flaw in your argument is that your boss is the one who refused to pay for backups.
    – RonJohn
    Jul 15, 2023 at 12:50
2

RAID helps you to eliminate downtime in case of limited, but most probable scenarios, of HDD failure scenarios. Usually it's one drive failure at a time.

RAID does not protect you from having stored invalid data on drives. Application or system software bug causing wipe of some or all data from drives, or human mistake deleting wrong data, or malicious users, or viruses. In such scenarios, RAID ensures, that data loss happened also on redundant drives.

RAID does not protect you from lossing whole array at the same time. Fires, floods, or other catastrophes destroy it all at once. Similarly thiefs can stole whole NAS at once, or very drunk roommate in a very bad mood can play "throw it as far away as possible" with NAS.

Backups help you get back in time. Restore what was once stored as current/live data.

Backups help you to restore previous versions of lost data in case of catastrophic failure.

Mirroring of data helps you to protect from catastrophic loss of single physical location, but doesn't necessarily prevent hackers or viruses or other means of data loss, or corruption, propagation to mirrors.

4
  • Actually I am very disappointed the correct explanation comes here 12 years after the question was asked. The elimination of downtime is the only real purpose of the RAID. Dec 12, 2021 at 13:55
  • It's not truly only real purpose of RAID. It also prevents short term data loss, changes since last backup, in case of single drive failure (or more, depending on RAID level). Similar can be achieved with very often incremental backups, or some audit log, or mirroring or by other means, but, RAID does that job better. However, the point is similar, that one can't rely on RAID as a way to ensure, that data will be kept intact and available in future, as a system with RAID itself is the single point of failure.
    – kravemir
    Dec 13, 2021 at 16:37
  • 1
    If you need to have guaranteed data recovery at lesser interval than you do backups, the last technology you'll be relying upon is RAID, right? You'll use application-level replication, filesystem-level snapshots, block-level snapshots, etc. RAID doesn't enable data recovery per se, it only enhances all those technologies to be able to survive HW failure. All it does is to be less affected by HW failure. Any of those technologies will do its work in absence of RAID (except for HW failure case), but RAID alone won't allow any such recovery possibility. Dec 14, 2021 at 6:52
  • You're right. Software architecture designed for strong data retention guarantee, would also include data replication on multiple different devices, therefore also covers failure scenarios that are handled by RAID. So, RAID is not strictly necessary then, but a very useful convenience.
    – kravemir
    Dec 14, 2021 at 13:46
1

Also consider with raid that you have multiple hard drives probably build at the same time and then exposed to the same conditions for years .... what are the chances that they will all fail about the same time .... pretty high

2
  • 3
    MTBF != expected lifespan of gear May 7, 2009 at 20:38
  • This isn't really an issue with RAID, though. Well, the "same use patterns" might be exacerbated by RAID, but multiple drives exposed to the same conditions isn't a function of RAID.
    – user
    Dec 24, 2013 at 15:17
1

[Not an answer, I already know. But an instructive tale nevertheless. Feel free to not upvote it. I'm posting it as an answer simply because it's too long for a comment.]

"We have redundant web servers on a load balancer, redundant database servers in a cluster and redundant hard drives in every server. So how did this happen? According to our server company there was a manufacturers bug in the firmware of the specific model that 6 of our 8 hard drives were on. That bug caused the disks to die after a certain number of hours running."

https://tvtropes.org

https://tvtropes.org/pmwiki/posts.php?discussion=15941624520A37147500

0

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .