RAID Trust Issues — Windows Or a Cheap Controller?

NicApicella writes "My new system has two sparklin' SATA drives which I would like to mirror. After having been burned by a not-so-cheap, dedicated RAID controller, I have been pointed to software RAID solutions. I now stand in front of two choices for setting up my RAID: a Windows 7 RC software RAID or a hardware RAID done by the cheap integrated RAID controller of my motherboard. Based on past experiences, I have decided that only my data is worth saving — that's why the RAID should mirror two disks (FAT32) that are not the boot disk (i.e. do not contain an OS or any fancy stuff). Of course, such a setup should secure my data; should a drive crash, I want the system up and running in no time. Even more importantly, I want any drive and its data to be as safe and portable as possible (that's the reason for choosing FAT32), even if the OS or the controller screw up big time. So, which should I choose? Who should I trust more, Microsoft's Windows 7 or possibly the cheapest RAID controller on the market? Are there other cheap solutions?"

Comments

Your first problem is Fat32

[Blakey Rat]

If you want data integrity, use NTFS. Using Fat32 is like saying you want a reliable car, so you're buying a Edsel because they've been around a long time-- it doesn't make sense. Every other OS on earth can read NTFS (if not write it), so it won't affect your portability requirement.

Secondly, before you make any decision regarding Windows 7 RAID, make sure the edition of Windows 7 you want to buy ships with software RAID support before you put all your eggs in that basket-- early betas and RCs of Vista had software RAID enabled, only to have it disabled before release. I've seen no guarantees about Windows 7 software RAID support, and which editions will have it enabled. (If any.)

If you're planning to move to a server OS after Windows 7 expires, I can practically guarantee software RAID will be enabled, but that still doesn't mean you can necessarily upgrade your Windows 7 software RAID array to a Windows Server software RAID array. Do your homework.

It's *NOT* hardware raid on your motherboard.

[Zarhan]

These motherboard "raids" are called fakeraids.

All that it is is that it writes the metadata on the disk in specific format so that you can see the raid volumes via BIOS. Note: Only "see" their status - in case you replace one drive, the resync is still done by software and you must boot to operating system. One clue is the fact that in Linux the dmraid package uses exactly same driver for accessing fakeraid-mirrored drives and Linux's own software-raids - device mapper just does a bit of magic at init.

However, if faced with choice of Windows-only or motherboard-raid, I'd go with the motherboard-version, because that's at least supported both by Windows and Linux so in case something goes wrong with your Windows installation you can always pop in Knoppix or some other Linux CD for recovery.

Re:It's *NOT* hardware raid on your motherboard.

[Grishnakh]

DO NOT buy a real RAID card unless you have a pretty good budget for your system, and need the highest performance. The problem with buying a real RAID card is that you need to buy not 1, but 2 or better yet 3 of them, so that you can have spares. If your RAID card dies (and they do, more often than you'd think), the only way you'll be able to access that data, because of the proprietary on-disk storage method used by RAID card vendors, is to have an identical card (with the same firmware version, to be safe). And since hardware is constantly being obsoleted, you need to buy your replacements when you buy your card, not hope they're still available later. It's also a good idea to have spares of the same make and model hard drive, because hardware RAID controllers aren't usually that flexible in allowing you to pair up different sized drives like Linux sofware RAID.

For many purposes, software RAID using Linux is really a much better solution, because the on-disk format is open-source and standardized, so it doesn't matter what hardware you have, you can plug the disks into a different Linux system and you'll be able to read the data with no trouble. The only downside is a slight performance decrease since the CPU has to do all the work, but even then unless the system is heavily loaded, it's still faster than hardware RAID because the hardware RAID cards aren't that fast.

With the giant drives that are now common, I think the best solution, at least for home/desktop systems, is to forget about RAID5/6 altogether and just get a couple of 1-2TB SATA drives and mirror them with software RAID 1 in Linux.

Re:Just remember the first rule of RAID 0

[Anonymous Coward]

That's nice, but the submitter is asking about RAID 1.

Re:Just remember the first rule of RAID 0

[Anpheus]

Actually, it depends on the reliability. 95% reliability becomes 90.25% reliability. 50% reliability becomes 25% reliability. 1% reliability becomes 0.01% reliability.

So if your drives are very reliable, it's very slightly less than twice the failure rate. If your drives are not reliable, then it asymptotically approaches an infinitely greater risk of failure.

Statistically speaking. :)

Re:Be Careful

[tomhudson]

Repeat after me: RAID is NOT backup

... and rsync is your friend.

What RAID is good for:

• making a bunch of cheap disks you just happen to have lying around look like one larger drive
• making reads quicker (mirrored disks, and sometimes, but not necessarily, RAID5/6), though you have a performance penalty for writes
• being able to say "I have a RAID"
• destroying much larger datasets since "It's safe - I have a RAID"

Better to just throw a disk in an old machine and back up to it regularly.

Re:FAT???

[jeffasselin]

No. NTFS is not perfect, but to think FAT is as bad is deluded. I've honestly never seen a HD formatted with NTFS that I couldn't repair with built-in tools, unless it had physical defects, and in such a case ANY file system would have problems. But I've seen so many FAT drives get hosed by little problems, it's not even funny.

Seriously, don't trust your data to a FAT partition - not worth it.

Re:Just remember the first rule of RAID 0

[LoRdTAW]

And what allot of people don't realize is if you build a RAID array and a drive fails can you replace the drive with the exact make and model? Raids work best when every disk in the array is the same model and revision. If you plan to build a 5 disk raid array you should also purchase a 6th drive to keep as a cold spare.

I built a RAID 5 array using three 500GB disks via mdadm under Linux. I assembled the array and formatted it. Within minutes of testing I was getting mail from mdadm telling me the array was degraded. I then began to test each disk for defects and lo and behold one disk was bad right from the start. I tried to RMA the disk but newegg had informed me those disks were now obsolete. Great. I was credited for the bad disk and purchased a new one that closely matched the other two. It was a nightmare as during some boots the disks went haywire and I would get a "Could not bd_claim sdaX" And it would hang for a while and I would have no array. It happend once in a rare while until it became a real problem. I kept my most precious data safely backed up on different disks I had spread around. It finally got so bad that I would have to constantly reboot the machine for up to ten times before the disks were synced up and the array worked. I purchased a 1TB disk and copied all the data off the array to it and used the 500gb disks in other systems. RAID is great for big fat storage arrays but it can become very sensitive and then one day POOF its all gone.

This is the reason OEM drives from Dell, Apple, HP etc. Cost four times what a retail drive would cost. The cost is no way associated with quality but rather consistency. Retail SATA drives are constantly changing: less/more platters, faster seek and read speeds and firmware revisions. Those costly OEM drives are the same disk every time right down to the inner workings and firmware. So if you buy an Apple 1TB disk on a sled and it takes a dump in three years you can be confident Apple will replace that drive with the EXACT same one. Its not a magical Apple disk of superior quality but a Maxtor/WD/Hitachi disk that is produced for Apple with no revision changes unless Apple orders it. Unlike retail drives which are changed at the manufactures whim.

So if you are building your own raid plan for failures and try to buy a spare for your array. I don't know disk shelf live but it will save you down the line. Also keep a USB or 1394 disk around for backups. Spread your most precious data around like pictures home movies and documents. If you have a few computers around the house keep a mirror of that data one those machines. Music, and downloaded video can be re downloaded but home movies and pictures cannot. Put all the silly stuff on the raid along with the precious stuff for access but keep backups of the good stuff!

Re:Just remember the first rule of RAID 0

[Trebawa]

He wants to mirror the drives. This means he wants RAID 1. Therefore, the failure rate of the array is 1/2 the failure rate of each disk (more, actually, because they're like;y identical drives that will fail at the same time, but you get my point).

Re:Just remember the first rule of RAID 0

[nxtw]

You may have noticed that some hard drives are marketed as being designed for RAID use. These work slightly differently to most consumer disks. Typically, a small region of a disk is hidden. If the disk discovers a bad sector then it will use one from the hidden region to replace it, so every write to the bad sector goes to one of the spare ones instead. This is very bad for RAID, because two drives writing to the same sector may be writing to two different physical locations (if one is remapped), with the same problems I outlined above.

All modern disks remap sectors as necessary. The main difference between consumer and RAID drives is the timeout for error correction.

Re:You are asking the wrong question.

[julesh]

Having run RAID quite a bit myself one must remember having all your drives in one box is always an invitation for trouble since hardware failures on a higher order will likely hit all the drives.

Not to mention the temptation to use _Identical_ disks in your redundant array... I've had a RAID1 pair fail totally when both drives died within 24 hours of each other because of a firmware bug. This happens a lot more than most people think. Statistical analysis of the reliability of RAID _always_ assumes failures arrive independently of each other, but a large proportion of failures are caused not by random events but by external circumstances and therefore happen either simultaneously or nearly simultaneously.

Re:Just remember the first rule of RAID 0

[mooboy]

Whoa, hold the boat. I've had a lot of experience with Dell & HP/Compaq(Proliant) provided RAID systems and they are not sensitive to disks with vastly different innards. All that matters is block count and software mirroring doesn't even care about that, because you'll simply be limited to the size of the smaller disk. If you're using mirroring or RAID, try to go with different makes of the same size. This article talks about MTBF. It turns out if 2 drives of the same exact model comes off the line and end up in your PC, there is a chance they could fail within a very close time to one another. So your mirror or RAID could fail permanently while rebuilding from the first failure. But if all your drives are of a different make, chances are they won't fail at the same time and you'll get the critical time needed to rebuild your array.

When I'm going to do mirroring or RAID on hardware that doesn't have high-end dedicated server RAID controller, I use Windows or Linux software RAID. Performance is surprisingly good and I'm not married to a specific hardware implementation. I've had _none_ of the issues you've described with Linux software RAID on several servers for several years. Mdadm has only whined after a power outage or genuine disk failure.

Re:Just remember the first rule of RAID 0

[turbidostato]

"That's nice, but the submitter is asking about RAID 1."

I think he's asking the wrong question anyway.

"Based on past experiences, I have decided that only my data is worth saving"

See? He is asking for backup, not RAID. It has been said one thousand times but it seems it must be said again: RAID is *NOT* in order to protect your data. NOT, NOT, NOT and then NOT again.

RAID (not talking about RAID-0) is there in order to enhance your data's avaliability (as in, say, instead of being able to get to my data 99% of the time, I can get to it 99,9%) but when it's hosed, it's hosed. To protect your data you need backups, not RAID.

"Of course, such a setup should secure my data"

Of course not. Of course you will get quite a funny face when you discover it. Quite more or less the one that had the guy from this story, about six months ago, with the very enlightning title "Why Mirroring Is Not a Backup Solution": http://hardware.slashdot.org/article.pl?sid=09/01/02/1546214

"Even more importantly, I want any drive and its data to be as safe and portable as possible"

Then, *even* if RAID could be considered for data security (which is not) you already answered your question: as a general matter, hardware RAID will only work when using exactly the same controller model, possibly up to its minor revision. You can't count to break a hardware-managed mirror, take one disk to a standard SATA controller and get any data out of it. If your controller dies and miracolously doesn't take the disks with it you can't count on buying a different RAID card (as it will most probably be in about a year for consumer-grade hardware) and get any data out of the mirror. So you should go with software RAID.

AND TAKE BACKUPS.

Re:Just remember the first rule of RAID 0

[Sandbags]

Why RAID is not a backup:
1) not fireproof.
2) not mistake proof "oops, didn;t mean to delete that"
3) not immune to file system corruption.
4) not immune to power supply failure/surge/lightning/other destructive forces
5) more expensive than a good backup
6) not protable offsite
7) does not track versionb history or old files (something that should be of critical importance to a programmer...
8) Viruses, mailware, hackers oh my!
9) bad/corrupt install
10) OS failure

I could easily go on. I worked in DR for 4 years...

Nearly all of the above have a higher frequency of occurance over a 5 year typical HDD life. Even if you continually replace drives without a data failure, you're still eventually going to have an issue RAID can not deal with.

My Qnap was a $399 device. The 4 drives in it were $90 each (and the 5th spare too). The HDDs I run the PC off on the RAID 1/0 were $40 each. I only run the RAID 1/0 for performance during video editing. I chose 1/0 vs 1 since 1 halves the reliabiltiy of the drives. Even though I do have a good recovery solution, the downtime, nor the effort involved in recovery, would be welcome, and the extra $80 to mirror the performance stripe was easily spent.

The Qnap is also my iTunes media server, my FTP server, included the price of the DR software, and runs 2 IP cameras I set up at home too (which let me tell the insurance company I have real-time video monitoring, and they knowcked an extra 5% of my homewoners policy cost, which by itself is enough to fund replacement drives as I'll need them).... Oh, yea, and it's a NAS too... It has a lot of value beyond a backup system.

I'm guessing you've not got a child yet, or a large family. You probably don;t value to pictures you take, files you have, and other stuff on your PC. That's fine, someday you likely will.

There are cheaper ways than mine to do backups. I have over a TB, and 3 (currelty, soon to add 2 Macs to the list an decom 1 old laptop leaving me with 4) computers I'm backing up, so centrally makes sense. If you have 1-2 machines, a small amount of data, and don't value most of it, then 2 external USB drives and a safety deposit box (Dad's house) usually suffice... Or, just an online backup account for $5 a month...

RAID 1 might save you from a firmware failure, or a disk going bad, but that's about it... Also, RAID 1 may be cheap, but a backup is cheaper. Also, good luck rebuilding that RAID if your MOTHERBOARD fails... RAIDs are proprietary to a particular controller. Unless your new board usues the same chipset (and firmware too in most cases) you;re screwed without a backup.