RAID Array - Which type should I use?

In My Humble Opinion (IMHO)
#1

I’d basically like your input on what I should do with my 4, brand spanking new, SATA (3.0Gb/s) 250 GB Hardrives.

This will be my first RAID setup and I’m a little confused as to what would be the best setup for me. I’ve also heard that it’s not a cakewalk setting up a RAID system to run properly with Windows XP, so any suggesiton on setting it up would be very welcomed!

My original plan so far was:

2 seperate arrays. One type 1 array would use two of the drives to set up a secure place to store my mp3’s and family pictures. Coupled with a DVD backup every 6 months or so I thought this would provide some very decent data security (Array type 1 is data mirroring, right? See why I need help!?).

The second array would utilize my other two drives and it would be setup as a type 0. I’d have my transitional files there (anything I have to edit/modify before sending it to the other, safer array) the OS (windows XP SP2) and my games. This would give me great performance with the caveat that should either hardrive fail I’d loose everything in those disks (am I right?).

Does this sound good to you?Do you have a better suggestion? Also, can two different array types be used at the same time? I’m using the RAID controller of my motherboard fwiw (It’s a new ASUS board - ATI Xpress 3200 Chipset).

#2

Be careful. A RAID array proves backup for the DRIVE, not the DATA. It improves the uptime of the drive because if one drive fails the other takes over, and you can replace the failed drive with no losses. But if you type del ., then your data is just as gone as on a single drive. Your two drives will always match each other…delete or change data on one and it’s deleted or changed on the other.

I see you’re planning a DVD backup of your data so I think you already understand this, but I’ve seen several people who were burned because they thought their data was “backed up” by the RAID array.

#3

In a more reasonable scenario (I assume you’re unlikely to hit del .), if the data become corrupted in one way or another (maybe a loss of power during a read/write cycle, or spyware, etc.) you’re just as hosed.

I’ve never quite understood the use of a type 1 array for home use. As Lemur866 said, it only protects against a phsyical media failure, and even then only when downtime costs money.

I’d suggest just leaving the backup harddrive as a stand-alone and using Microsoft’s SyncToy (free) to schedule regular “mirroring”.

#4

The controller on your mobo is probably run in software. Unless you’re going to be buying a separate controller card, and that controller is real hardware RAID (hint: if you have to load a driver to get basic functionality, it ain’t hardware RAID), don’t bother doing anything other than 0. Performance is gonna be crap.

If you don’t wanna mess with a card, stripe two of the disks, and use the other two for periodic backups. Don’t set the backups disks as a single logical volume - make sure your backup utility can span across multiple volumes instead.

If you do decide to get a controller card, buy two. What happens if your card bites the dust?

Doing RAID right can get very expensive very fast, if you’re truly worried about both performance and fault tolerance.

On the other hand, if you take one of those two factors out of the equation, you can get satisfactory performance on the cheap. I just set up a file server using Linux software RAID1, since the network was gonna be a bottleneck, anyway.

#5

Since you have 4 drives, RAID 5 would probably get you the best combination of performance, reliability and available space - it’s designed to handle any single drive failure at a time. If you have a RAID 0 setup, and one drive goes down, your data is lost; under RAID 1, you’re losing 50% of the available capacity. RAID 5 offers excellent read performance and moderate write performance, and in this case you’ll have the equivalent of 3 drives’ worth of space available to you - 750 gigs of redundant space.

#6

Speaking as someone who did RAID on the cheap, it’s not that difficult to set up- maybe twenty minutes of work. I just used the RAID utility that came with my motherboard. However, my RAID array isn’t my primary drive; I’ve got an SATA drive that serves as a main system drive, and a few old IDE drives that are RAIDed to give me a storage space for downloads I don’t trust and large files that I can replace. The speed is lousy, but it works as a long-term storage space for things I’m not especially attached to.

Quite true. I did it on the cheap, and I’m aware that it’s a bit of a risk.

#7

Except that RAID 5 on a software controller kills the CPU with all sorts of parity checking.

#8

So it seems the concensus is that I won’t be getting any of the benefits of a hardware RAID at all. Which means I might as well RAID 0 and just use the RAID 1 for the two drives in order to improve keep those important files safe. I guess I’ll use one of the toher dirves as my system drive, and the last one as a downloads disk (I’ll also move the system swap file there.

#9

Excellent point. I didn’t read the part about it being a software controller. :wally

I’ve seen RAID 5 on a pretty heavy-duty dual-processor machine, and doing it on a software controller is… well, pretty much a waste. With a good hardware controller, however, it’s pretty damn quick. That, however, seems to be an outlay of around $300, from looking at prices on Google.

#10

I considered RAID briefly when I got my new machine in February. I even ordered a second 250GB drive, the exact model of the original. Then I pretty much figured out the same thing that Jayrot said: I would be tossing out a huge amount of disk space without protecting things beyond the hardware failure level. What happens if I get a virus that wipes everything out?

My current solution: I have a 10 year old clunker that I set up as a Linux file server where I keep all important files. It is accessed as network drives served by Samba. I have cron jobs running that change permissions and ownership on the files every night so that no rogue virus in Windows-land can ever harm my pictures/video/mp3s.

Since there are far too many files to back up to DVD, I finally settled on purchasing an external hard drive (WD MyBook, for the record), and I plug it in once a week and run the previously-mentioned MS SyncToy software to copy any changed photos/mp3 files to the “MyBook” (over the network share). I then unplug it and put it on a book shelf :).

One thing of note: it seems that the free backup software you get with some of those external drives won’t read from network drives, possibly a ploy to get businesses to buy more expensive software. That’s why I ended up using SyncToy.

#11

Oh, c’mon everyone! An EMC Symmetrix with a 4 gig fibre channel controller through a McDATA director backing up to a Storage Tek library packed with T10000 drives.
What’s the big deal? Piece of cake! Sun and EMC would probably even come out and set it all up for you!

I suppose, if you want it on the cheap and you wanna hook it up yourself, HP EVA 4000, an edge switch, ATL tape library. But you’re giving up high availability, can you afford that kind of down time?

#12

Do not, whatever you do, go RAID 0. If you use RAID 0, it means that if any one of your drives fails, all your data is lost. With four drives, consider purchasing a proper RAID 5 card or using RAID 0+1 which is like RAID 0 but the whole thing is mirrorred.

#13

My SO and I set up a new “media” machine a few months ago, and were asking the same questions. We finally decided that a raid really didn’t offer us any value.

A raid 0 would allow us to use all of the HD storage, but we aren’t saving any huge files that need to carry over from drive to drive. We also figured out, as Quartz states, that if one of the drives went bellyup, we would lose it all.

A raid 1 would let us back everything up, but this is music and pictures. It isn’t business critical for us to be up and running 24/7. A standard backup situation makes a whole lot more sense for us, and doesn’t cut our storage space in half.

#14

But if you just want to back stuff up, you don’t want a RAID 1 array.

You want two disks, one main disk and one backup disk. Every so often you backup your data to the backup disk. RAID 1 isn’t backing up your data, it’s just protecting it against a drive failure. If your data gets corrupted on a RAID 1 then you don’t have a backup. The advantage of a RAID 1 is that if one drive fails your drive continues to function without interuption.

This is critical for businesses, but for home use it won’t hurt you to have your main drive down for a few hours/days while you get it fixed/replaced, then restore from backup.

The only risk is that if your main drive fails you’ll lose the data since your last backup. But how much critical data do you have? If you have critical data, save it to backup right away. But you don’t need to worry about 24/7 uptime.

#15

One more point. A RAID 0 array is problematic because if any one of the physical drives fail you lose everything. This is much more risky than having all the data on one drive. Unless you need everything on on partition, why not just have two independent drives?

#16

I’ve set up a RAID-5 array with 8 250GB drives hooked up to a 3Ware 9000 controller. 1.5 TB available space, no CPU hit, and I use the box as a NAS over gigabit.

This was done approximately a year and a half ago.

Nowdays, you can put together an equivalent system (hot swap drives, RAID-5, gigabit NAS, 1.5 TB available space) using a $550 Intel SS4000E enclosure and 4 500GB drives, for less than half the price of the box I built. Time marches on.

The controller on your motherboard, Kinthalis, uses the CPU for its heavy lifting. RAID-0 (striping) is not CPU-intensive at all. RAID-1 (mirroring) isn’t really CPU-intensive either. However, your setup (2 drives in RAID-0, 2 drives in RAID-1) results in 750GB of usable space, of which only 250GB is proof against drive failure. RAID-5 would probably be the optimum way of putting things together (excellent read and good write performance, all 750GB proof against drive failure).

The problem with RAID-5, however, is that it is very CPU-intensive. Also, that motherboard controller probably doesn’t support a RAID-5 mode, so you’d be using it just for its ports, and doing RAID-5 entirely in software. Windows XP does not do RAID-5 out of the box, there’s some tinkering that needs to be done. Also, once you get it running, you’ll take a big CPU hit.

You can use a dedicated controller that can do hardware RAID-5 - these controllers typically run 2-300 dollars. Or, you can just buy the enclosure for $550.