Why is RAID 1 considered a bad idea for data integrity?

Factual Questions
1

Or is it that it’s not enough?

Here’s my thoughts and configuration.

Our family pics n’ videos are stored on a Win 7 Home machine. It has a boot drive and two data drives in a RAID 1 configuration. I have Windows set to regularly and automatically backup data on the RAID array (is that usage redundant—as in PIN number?) to a NAS device. Though off-site would be best, it’s in the opposite end of the house, in a basement, protected by rabid jaguars.

Thanks to a failed power supply, one of the RAID drives is failing the S.M.A.R.T. check, so Western Digital sent me a replacement. Prior to installing it, I’ve been poking around the net to make sure the swap goes smoothly. And once again I’m coming across page after page that says RAID is a bad idea except for mission critical situations (i.e., uptime is very important). Example.

Note again that with backups running, I am protecting myself from accidental deletions or (hopefully) malware.

My thought is that given the relative price of storage space, an extra $60–$70 adds another layer of security. I could turn to the backup (also on a RAID 1) to rebuild, but keeping the RAID 1 mirrors is adding another layer of protection–as it is now, four drives have to fail before I’m in trouble.

Or am I missing something? Are RAID mirrors more likely to suffer faults? Am I overreading and they’re fine if you still have a backup?

2

Make a list of all the ways you can lose data. Now cross off everything that RAID 1 solves. Depending on how creative you are, you’ve probably only crossed off 1 thing and have at least half a dozen left. Is solving that one problem worth the added cost and complexity? Especially since you’re still going to need a backup solution that solves the other half a dozen problems?

That’s basically it. I actually like and use mirroring as a way to reduce downtime, but I still do offline backups.

3

It’s not…it’s just sort of old school. Basically, RAID 1 is disk mirroring…so, if you have a 500 GB hard drive, you’d need a second 500 GB hard drive to mirror it…which means you are using 1000 GB of total disk space to protect a maximum of 500 GB of data. Later technologies cut down significantly on that, and disk imaging backup software (stuff like Acronis) gives you the ability to image your data and restore it really easily.

But RAID 1 is still better than no redundancy at all, at least if you want to try and ensure some level of up time. It’s not really a backup methodology, but it will ensure that if you lose a drive you are still up and running.

Based on your OP, I’d say it’s worth it to replace the failing drive and continue with using RAID 1 (if I’m understanding what you are getting at there). Depending on how important the data is to you, you might want to consider one of the myriad off site backup services available, and invest in something like Acronis to image your system periodically and upload it offsite (or you could buy a NAS and backup to that, which will make restoring a bit easier but makes your data a bit more vulnerable). That should cover all of your bases.

-XT

4

For crimes against the understanding of computers, I would put the person who dreamt up the idea of RAID “levels” second only to the marketing droid who named a certain programming language “Javascript”. Because, you know, Java was cool at the time, so calling things Java-this and Java-that seemed to make sense. RAID levels are not levels, one on top of the other. They’re fundamentally different types of redundancy, that for some reason were grouped together by the idiot in question.

Anyway, the guy in the link is complaining about RAID in general, not mirrored drives (RAID 1) in particular. He has a point about RAID controllers adding complexity and failure points, for little real gain. The performance benefits of RAID in lightly-used servers have always been dubious.

5

That would be Patterson and Chen. In Chen’s original work I think he did try to think of them as an increasing sequence of technologies, with some level of incremental increase in complexity. That was a theoretical description, and (I think) part of his thesis. It wasn’t intended as anything other than a structured way of understanding the ideas. Market droids came later.

Personally I like mirrored disks as a base, but the reasons are not critical to most people. The main reason is that in the face of a disk failure you can keep working. For something mission critical , or when it is just plain really inconvenient to have to drop everything and recover the system, being able to run (unprotected) off one of the mirror drives can be a good thing. The downside is that human nature will put off the time when you replace the disk. Leaving you unprotected.

What mirrored disks do not do is provide a substitute for backups and archiving. Once you have a solid backup and archive solution in place you can think about the marginal value of adding mirroring. For most people it is indeed a waste of time.

6

Single and multiple drive failures. I’ve been building machines since the 386 days, and though I keep to reputable manufactures (e.g., WD, Seagate) and don’t stress the machines (no overclocking, well cooled, and dust is occasionally removed) I’ve had a small handful of drives go south. This one didn’t exactly fail, but the S.M.A.R.T. warning, as vague as it is, made me antsy. Since the whole house is wired for sound and video (kitchen, den, parlour all have PC’s hooked to stereos) and the library is located in that machine, it is kind of critical to have as little interruption as possible. Not quite like running a hospital; much more for convenience. Yes, I could rebuild a single drive from a single drive’s backup–and yes the interruption might be nominal. But having a level of failure to go through first (both main drives crapping out) before turning to that solution is (subjectively) important to us.

It’s really a product of how cheap storage space is. Working in a retail store (remember Software Etc.?) in the late eighties, I sold my first hard drive–a 10 MB monster for $350. Prices are ridiculous now.

So, fire or other natural disaster? Each box (PC and NAS) has two drives in it, in some ways (godzilla attacks notwithstanding) giving me two chances to recover. Lightning strike that somehow makes it past the UPS devices? Again, four chances for data to be recoverable. Barring a super-complete disaster, the two locations are in very different areas of the house. So come what may, I’m (in my mind; hence the nature of the thread) counting on two locations, and two drives in each. Just one has to be readable right away, or intact enough to go to a data recovery technician.

This is really our next-best solution. Given the amount of data that we have, uploading over our slowish connection would be obnoxious. Then a handful of movies gets saved, and the time to upload starts again. We work from home, so off-site dropping off of a drive is also not a practical option. So it’s not the only thing that lets us sleep at night, but having data mirrored across four drives gives us the sense that in the worst case scenario, at least one of those drives will be recoverable.

An underlying premise to this feeling is the core of the OP: could RAID 1 add to the possibility that the whole system will fail? Could the RAID part of the OS actually fry both drives simultaneously–and is it more likely to happen in a RAID array than a single drive? Oh, note that the PC is Win 7, the NAS is a Linux machine.