Why Is RAID Not a Backup
-
We always say it, RAID is not a backups. But people routinely confuse it with one, why?
RAID makes a copy of the data, right? Well mirrored RAID does. Parity RAID, which most people use, does not. So the answer is, at best, sometimes. But that doesn't really matter.
RAID is tightly coupled. Backups, by definition, must be uncoupled. What does this mean? RAID is dependent on the system it is protecting. If the server burns in a fire, so does the RAID. If the server is stolen, so is the RAID. If someone deletes a file on the RAID, all "copies" are instantly deleted. If the data corrupts, all copies are instantly corrupted together. If the original is modified, all copies are modified with it. If the original is encrypted, all copies are encrypted with it. Backup does not carry these risks.
RAID is Disaster Avoidance. The purpose of RAID is to make a system more resilient in the hopes that a disaster does not happen.
Backup is Disaster Recovery. Backups are useful once a disaster has happened and now you have to recover.
Think of RAID like the brakes on your car, their job is to keep you from hitting that tree. Backup is like the seatbelt. It's only there to save you after you have hit the tree. You would never skip either, both are very important.
-
-
@Carnival-Boy said:
@scottalanmiller said:
But people routinely confuse it with one
Really? I have never heard that.
I can't tell if you are being sarcastic or not. It's one of the most common fallacies that I see people state. Saw it once this morning already in a post that I linked to here. It's common enough that "RAID is not a backup" is a mantra that has arisen in response to it.
-
No, I'm being serious. It's not something I've heard. Do these people not take backups then?
-
@Carnival-Boy said:
No, I'm being serious. It's not something I've heard. Do these people not take backups then?
Well, hopefully we catch them before they implement a system with RAID instead of backups, but yes, there are tons of shops out there running RAID believing that it is a backup. As an MSP, we see it once in a while.
The question today was something akin to "Trying to decide between RAID or duplicate disks for backup."
-
I'm thinking it's semantics, which I know is your thing, and you probably know isn't mine
From wikipedia, on redundancy, "In engineering, redundancy is the duplication of critical components or functions of a system with the intention of increasing reliability of the system, usually in the form of a backup or fail-safe."
In that context, I'd argue that RAID involves "backup" disks, so that one or more disks can fail but the array will continue working. So it's a backup in one sense, just not a backup in what I think of as a backup of my data.
I guess what I'm saying is that "backup" can refer to many things.
-
@Carnival-Boy said:
I'm thinking it's semantics, which I know is your thing, and you probably know isn't mine
From wikipedia, on redundancy, "In engineering, redundancy is the duplication of critical components or functions of a system with the intention of increasing reliability of the system, usually in the form of a backup or fail-safe."
In that context, I'd argue that RAID involves "backup" disks, so that one or more disks can fail but the array will continue working. So it's a backup in one sense, just not a backup in what I think of as a backup of my data.
I guess what I'm saying is that "backup" can refer to many things.
Which is why I defined it above. Backup, in the common usage and the legal one, means decoupled. With RAID if you delete a file and want to recover it immediately, you quickly find out that you didn't have a backup. There were two copies, but they were actually the same copy, too.
-
It's effectively identical to the thing that snapshots are not backups. It's become somewhat common for people to think that taking a snapshot of their data is a backup. But when the storage fails, the backup is always gone with the original. Or if it corrupts, both go together as they are literally the same data.
-
Which, in turn, is the same as the really old issue that we had in the 1980s and 1990s that people would just make a copy of a file on the same disk (in the 80s that meant on the same floppy) and call it a backup. Sure, it protects against a few things, but not the things that they were worried about.
-
@Carnival-Boy said:
@scottalanmiller said:
But people routinely confuse it with one
Really? I have never heard that.
We had it come up here about 2 weeks ago...
-
Something I've seen in several conversations is a remote host, extremely limited funds, and a demand for a ton of storage space.
So the result is "I have to configure this in RAID0 as if that provides some protection. RAID0 offers no protection in any way, even if only used as a "backup target".
It presents the disks in the system as a single device that gets written to, but if anything in the world happens to this system, the array and data are likely lost.
-
@scottalanmiller said:
It's effectively identical to the thing that snapshots are not backups. It's become somewhat common for people to think that taking a snapshot of their data is a backup. But when the storage fails, the backup is always gone with the original. Or if it corrupts, both go together as they are literally the same data.
Well with snapshots at least that IS the beginning of the backup mechanism. Just you have to move them and the VM metadata off the host to call it a backup.
-
@Carnival-Boy said:
@scottalanmiller said:
But people routinely confuse it with one
Really? I have never heard that.
I have heard this as well, though not as much as I used to. I think people are starting to learn more in this area, but I do recall at least in the early 2000s RAIDS people would often say "Back ups? It's got a RAID"... then it turns out to be RAID 0, yes I've seen that.
-
@tonyshowoff Now that's just painful to think about.
