To RAID or not to RAID?
February 3, 2009 11:09 AM Subscribe
I just purchased two WD Caviar Black 500GB HDDs and I'm trying to figure out the best backup configuration for my home computer.
Previously, I had two 160GB drives. On HDD-0 I had a partition for my OS and a partition for data. HDD-1 was used solely for backups. I used Acronis to do separate backups of the OS and data partitions. In addition to this I backup most of my data online with JungleDisk.
With the new drives I'm trying to decide if I should continue doing the same thing or set them up in a RAID 1. RAID 1 sounds great, but I keep reading that the the only reason to do RAID 1 is if you're concerned about maintaining server-level uptime. I don't care so much about uptime as protecting against HDD failure. I realize RAID 1 is not a "backup" solution, per se, but I do have online backups to cover that. Performance isn't an issue as long as I'm not losing too much speed compared to a single drive.
The advantage that I can see to my current setup is that if I get a virus or something gets corrupted I still have the backup locally and can restore from that, whereas a RAID will instantly copy the virus/corruption to the mirrored disk. I'm pretty careful so I see that as a pretty darn unlikely scenario.
Other than that it seems like RAID 1 is ideal here, but is there anything I'm failing to consider? I rarely see it recommended for home use so I want to make sure I'm not missing anything.
Previously, I had two 160GB drives. On HDD-0 I had a partition for my OS and a partition for data. HDD-1 was used solely for backups. I used Acronis to do separate backups of the OS and data partitions. In addition to this I backup most of my data online with JungleDisk.
With the new drives I'm trying to decide if I should continue doing the same thing or set them up in a RAID 1. RAID 1 sounds great, but I keep reading that the the only reason to do RAID 1 is if you're concerned about maintaining server-level uptime. I don't care so much about uptime as protecting against HDD failure. I realize RAID 1 is not a "backup" solution, per se, but I do have online backups to cover that. Performance isn't an issue as long as I'm not losing too much speed compared to a single drive.
The advantage that I can see to my current setup is that if I get a virus or something gets corrupted I still have the backup locally and can restore from that, whereas a RAID will instantly copy the virus/corruption to the mirrored disk. I'm pretty careful so I see that as a pretty darn unlikely scenario.
Other than that it seems like RAID 1 is ideal here, but is there anything I'm failing to consider? I rarely see it recommended for home use so I want to make sure I'm not missing anything.
If you're just looking for a method of reducing downtime in the case of a HDD failure, RAID 1 is perfect.
File corruption doesn't care how 'careful' you are. If you lose power at a critical time, or have bad RAM, your files will get corrupted.
Even if you're backing everything up online, the only way RAID 1 is the better of the two options is if you think that a hardware failure is more likely than lost data.
posted by Jairus at 12:19 PM on February 3, 2009
File corruption doesn't care how 'careful' you are. If you lose power at a critical time, or have bad RAM, your files will get corrupted.
Even if you're backing everything up online, the only way RAID 1 is the better of the two options is if you think that a hardware failure is more likely than lost data.
posted by Jairus at 12:19 PM on February 3, 2009
Best answer: Performance: theoretically, reads are twice as fast and writes are twice as slow. Reads because the OS can schedule data to be read from both disks independently. Writes because the OS has to wait for each disk to sync independently. Realistically, I think the performance is somewhere in between.
Failure: good that you know this is not a backup solution in and of itself! On drives they talk about the Mean Time to Failure (MTF). Let's say that your drives get 500,000 hours. Most novices think that RAID1 then gives them the sum, or 1,000,000 hours. Just the opposite! The MTF for the array is (mtf1 + mtf2)/2, or in this example, 250,000 hours. That's because the array is more complex, and more complex systems generally have a lesser MTF.
The benefit is that the failure mode is more acceptable. I wouldn't RAID these drives unless you're (1) willing to monitor the array status (I do once a day) and (2) are willing to run out ASAP and get a replacement drive when one fails.
One more thing: I'd use software RAID over hardware, especially in this non-critical case. With hardware RAID if you controller goes you might have to buy exactly the same replacement controller or rebuild your array from scratch. A major pain in the ass. Not a problem if you let the OS do it, though.
Most consumer hardware RAID cheats and implements it in the software driver anyway. Which sucks, but it's cheaper.
note: my MTF calculation may not be right, but the point is. You get a lesser MTF when you RAID, not greater.
posted by sbutler at 12:20 PM on February 3, 2009
Failure: good that you know this is not a backup solution in and of itself! On drives they talk about the Mean Time to Failure (MTF). Let's say that your drives get 500,000 hours. Most novices think that RAID1 then gives them the sum, or 1,000,000 hours. Just the opposite! The MTF for the array is (mtf1 + mtf2)/2, or in this example, 250,000 hours. That's because the array is more complex, and more complex systems generally have a lesser MTF.
The benefit is that the failure mode is more acceptable. I wouldn't RAID these drives unless you're (1) willing to monitor the array status (I do once a day) and (2) are willing to run out ASAP and get a replacement drive when one fails.
One more thing: I'd use software RAID over hardware, especially in this non-critical case. With hardware RAID if you controller goes you might have to buy exactly the same replacement controller or rebuild your array from scratch. A major pain in the ass. Not a problem if you let the OS do it, though.
Most consumer hardware RAID cheats and implements it in the software driver anyway. Which sucks, but it's cheaper.
note: my MTF calculation may not be right, but the point is. You get a lesser MTF when you RAID, not greater.
posted by sbutler at 12:20 PM on February 3, 2009
RAID 1 is NOT backup. Your drives are mirrored. As long as nothing stupid happens you do have a backup, but if you get a virus on Disk 1, you get a virus on Disk2. If you accidentally overwrite or mangle a file on Disk 1 ...
If backup is your concern, your old way is better. If you are really concerned, you could get a third disk and set up RAID 5.
I always set my OS and Programs on a RAID 0 for speed, use a separate internal HDD for storage, and an external HDD for extra redundacy for really important stuff.
posted by Benny Andajetz at 1:04 PM on February 3, 2009
If backup is your concern, your old way is better. If you are really concerned, you could get a third disk and set up RAID 5.
I always set my OS and Programs on a RAID 0 for speed, use a separate internal HDD for storage, and an external HDD for extra redundacy for really important stuff.
posted by Benny Andajetz at 1:04 PM on February 3, 2009
Coincidentally, I saw on LifeHacker today that Acronis is giving away the last version of their backup software. I've had mediocre results with it (in the months since I paid for it, like a chump), but your mileage may be better.
posted by Shepherd at 1:19 PM on February 3, 2009
posted by Shepherd at 1:19 PM on February 3, 2009
Best answer: RAID is not a backup. You said it. Benny said it. I'll say it again: RAID is not a backup.
It's great for things like servers but for personal use I'm not convinced of its utility compared to the added complexity and other potential problems. I ran RAID 1 on a linux box for awhile and lost some files due to user error. I needed to have that lesson hammered home. Malware, corruption, hardware problems, and, of course, user error are all things that could cause loss of data.
You're much better off running your old way, with a main drive and a dedicated backup drive with the appropriate software for your OS. If uptime was very important you could run two drives as the main drive with RAID 1 and a third drive as the backup drive.
Really, unless you've got a specific need for RAID, don't bother with it.
posted by 6550 at 1:35 PM on February 3, 2009
It's great for things like servers but for personal use I'm not convinced of its utility compared to the added complexity and other potential problems. I ran RAID 1 on a linux box for awhile and lost some files due to user error. I needed to have that lesson hammered home. Malware, corruption, hardware problems, and, of course, user error are all things that could cause loss of data.
You're much better off running your old way, with a main drive and a dedicated backup drive with the appropriate software for your OS. If uptime was very important you could run two drives as the main drive with RAID 1 and a third drive as the backup drive.
Really, unless you've got a specific need for RAID, don't bother with it.
posted by 6550 at 1:35 PM on February 3, 2009
Something few consider: You are not really backed up unless a copy is off-site. Back in the early days of PC's I thought I was very forward thinking by having a tape backup in my office PC and I was very disciplined about backing up my files. Imagine my surprise when I came to work one morning to find someone had broken into our offices and stole our 3 PC's. Mine was stolen with the backup tape in it. Everything gone. If you're not planning for a catastrophic and total loss of your primary data location, then you really aren't planning. It's not pleasant to think about, but offices, data centers, and homes occasionally burn to the ground or otherwise destroyed.
If your backup resources are limited to two 500GB hard drives, I would rotate them on and off site at whatever frequency you think reasonable. I personally prefer 3 nearly identical data stores. Two onsite and one offsite. Most of my computers run multiple Operating Systems and I use tools on each OS to back up the system partition of the other operating systems. For example on my mac that also boots XP I use Winclone from the mac side to back up the windows partition and I have another disk that I can run in target disk mode if Mac partition fails. So data aside, I can also easily get either OS back up should I have a system or disk failure. I also try to organize my stuff into at least 3 backup categories:
1. System Partitions - don't really need a new backup unless I add new software.
2. Disk Images and install software - static and also doesn't need to be backed up too often.
3. Stuff I collect or create - Back up ALL THE TIME.
posted by Rafaelloello at 1:47 PM on February 3, 2009
If your backup resources are limited to two 500GB hard drives, I would rotate them on and off site at whatever frequency you think reasonable. I personally prefer 3 nearly identical data stores. Two onsite and one offsite. Most of my computers run multiple Operating Systems and I use tools on each OS to back up the system partition of the other operating systems. For example on my mac that also boots XP I use Winclone from the mac side to back up the windows partition and I have another disk that I can run in target disk mode if Mac partition fails. So data aside, I can also easily get either OS back up should I have a system or disk failure. I also try to organize my stuff into at least 3 backup categories:
1. System Partitions - don't really need a new backup unless I add new software.
2. Disk Images and install software - static and also doesn't need to be backed up too often.
3. Stuff I collect or create - Back up ALL THE TIME.
posted by Rafaelloello at 1:47 PM on February 3, 2009
Response by poster: @sbutlet, benny, 6550, et. al.: Thanks. I think that about settles it for me. Based on what you said, and the reading I've been doing, RAID is beginning to sound like more trouble than it's worth.
@Shepherd: I'm in the same boat. I bought TI 10 a year or two ago for $50. It's been... okay. It gets the job done but they've been horrible about fixing some glaring bugs. No chance I'm paying to upgrade to TI 11. There are other free options that do basically the same thing these days.
@Rafaelloello: As previously mentioned, I use JungleDisk for my online backups. I keep an image of a fresh XP install tweaked just the way I like it as well as weekly OS backups and daily data backups locally. The OS image and data are also backed up online.
Thanks for all the feedback guys!
posted by jluce50 at 2:09 PM on February 3, 2009
@Shepherd: I'm in the same boat. I bought TI 10 a year or two ago for $50. It's been... okay. It gets the job done but they've been horrible about fixing some glaring bugs. No chance I'm paying to upgrade to TI 11. There are other free options that do basically the same thing these days.
@Rafaelloello: As previously mentioned, I use JungleDisk for my online backups. I keep an image of a fresh XP install tweaked just the way I like it as well as weekly OS backups and daily data backups locally. The OS image and data are also backed up online.
Thanks for all the feedback guys!
posted by jluce50 at 2:09 PM on February 3, 2009
@Rafaelloello: As previously mentioned, I use JungleDisk for my online backups.
Whoops, missed that part. Good.
Also, for planning purposes, I've found that when it comes to computer hardware, if it doesn't fail in the first 30 days, it will usually last 3 years. You'll probably get 4 or 5 years out of most equipment, but I rarely see failures in the 30 day - 3 year timeframe. This hasn't seemed to have changed in decades.
RAID - I use hardware RAID or none at all. I like 3Ware's cards quite a bit.
posted by Rafaelloello at 2:19 PM on February 3, 2009
Whoops, missed that part. Good.
Also, for planning purposes, I've found that when it comes to computer hardware, if it doesn't fail in the first 30 days, it will usually last 3 years. You'll probably get 4 or 5 years out of most equipment, but I rarely see failures in the 30 day - 3 year timeframe. This hasn't seemed to have changed in decades.
RAID - I use hardware RAID or none at all. I like 3Ware's cards quite a bit.
posted by Rafaelloello at 2:19 PM on February 3, 2009
Go for the backup copy first. You can go for the uptime advantages of a RAID1 later, any performance benefits are likely to be minor.
Personally, I like software RAID because it eliminates the dependency on the controller, but most software RAID1 implementations have problems with block-level errors that can lead to data corruption.
posted by Good Brain at 2:51 PM on February 3, 2009
Personally, I like software RAID because it eliminates the dependency on the controller, but most software RAID1 implementations have problems with block-level errors that can lead to data corruption.
posted by Good Brain at 2:51 PM on February 3, 2009
FWIW, I have a dual enclosure from Buffalo set to RAID 1 that my macbook backs up to with time machine (via airport wireless). This particular setup is very low-effort and low-maintenance, but if I had the same money to spend again, I'd buy two separate HDDs and rotate them as Rafaelloello suggested. In your case, I don't think RAID is merited; it doesn't protect against user error, malware or theft, and offers very little that isn't covered by your online backups.
posted by Chris4d at 5:24 PM on February 3, 2009
posted by Chris4d at 5:24 PM on February 3, 2009
Here's my setup, which I recommend because it provides a good mix of convenience and safety:
I have two HDs, one of which (the backup) is external, unplugged from the wall when I'm not using it -- that way, if there's a power surge that fries everything in the computer, at least my backup is safe. Also, the external hard disk isn't spinning needlessly, reducing wear and (I would imagine) extending its life.
My internal HD is partitioned into two: C: is for operating system (plus programs, etc.), and D: is for user data like documents, pictures, movies, music, etc. Once a month or so, I image C: into a single file on the backup HD (using Macrium Reflect, which is free). Once a week, I copy the files on D: (using xcopy /s /d /v) to the backup hard disk. Copy is important to me because I'm not dependent on a backup program to read the most important files, and that stuff doesn't compress well anyway.
Also, every few months (in principle, if I remember) I burn a few DVDs of my data on D: that's changed since last time I burned DVDs, and I give them to a family member to hold. That way, in case of disaster, the most I lose is a couple of months worth of pictures, etc.
And I'm Nthing that RAID is useless for your needs.
posted by Simon Barclay at 6:23 PM on February 3, 2009
I have two HDs, one of which (the backup) is external, unplugged from the wall when I'm not using it -- that way, if there's a power surge that fries everything in the computer, at least my backup is safe. Also, the external hard disk isn't spinning needlessly, reducing wear and (I would imagine) extending its life.
My internal HD is partitioned into two: C: is for operating system (plus programs, etc.), and D: is for user data like documents, pictures, movies, music, etc. Once a month or so, I image C: into a single file on the backup HD (using Macrium Reflect, which is free). Once a week, I copy the files on D: (using xcopy /s /d /v) to the backup hard disk. Copy is important to me because I'm not dependent on a backup program to read the most important files, and that stuff doesn't compress well anyway.
Also, every few months (in principle, if I remember) I burn a few DVDs of my data on D: that's changed since last time I burned DVDs, and I give them to a family member to hold. That way, in case of disaster, the most I lose is a couple of months worth of pictures, etc.
And I'm Nthing that RAID is useless for your needs.
posted by Simon Barclay at 6:23 PM on February 3, 2009
* I just purchased two WD Caviar Black 500GB HDDs
Stupid decision. Why would you buy two of the same drive? Always buy different drives. Sometimes you have drive series with high failing rates. This is why you don't want to have the same brand.
* and I'm trying to figure out the best backup configuration for my home computer.
You seem to have little understanding what a backup is. Mentioning RAID makes things even worse.
Disclaimer: I use a setup of two (different) 500GB too. But I don't use RAID nor do I claim that this is a reliable backup system. I have a master HDD and a copy of it. From time to time I copy the master to the other drive. Once a year a buy a new drive and "retire" one in my parents house as a "backup" where I can always fall back on.
Enough for me? Yes. A reliable backup system? no.
And by the way, encryption is your friend.
posted by yoyo_nyc at 8:58 PM on February 3, 2009
Stupid decision. Why would you buy two of the same drive? Always buy different drives. Sometimes you have drive series with high failing rates. This is why you don't want to have the same brand.
* and I'm trying to figure out the best backup configuration for my home computer.
You seem to have little understanding what a backup is. Mentioning RAID makes things even worse.
Disclaimer: I use a setup of two (different) 500GB too. But I don't use RAID nor do I claim that this is a reliable backup system. I have a master HDD and a copy of it. From time to time I copy the master to the other drive. Once a year a buy a new drive and "retire" one in my parents house as a "backup" where I can always fall back on.
Enough for me? Yes. A reliable backup system? no.
And by the way, encryption is your friend.
posted by yoyo_nyc at 8:58 PM on February 3, 2009
Response by poster: @yoyo: I'll ignore the insulting tone and address your points.
Stupid decision. Why would you buy two of the same drive? Always buy different drives. Sometimes you have drive series with high failing rates. This is why you don't want to have the same brand.
And sometimes you have particularly reliable drives. Like Rafaelloello said, I should have a pretty good idea by the end of the first 30 days. Worst case scenario, you're right and both drives take a crap, I've got my OS and my data backed up.
You seem to have little understanding what a backup is. Mentioning RAID makes things even worse.
Uh, okay. I have redundant backups, including an external drive in a floor safe and off-site, all of which are encrypted. But please, do enlighten me. (Okay, so I couldn't ignore the insulting tone, sue me...)
BTW, "mentioning" RAID doesn't make anything worse. If I would have just gone ahead and done it without research, that would make it worse. I did my due diligence and have decided against it. So what's the problem?
posted by jluce50 at 7:09 AM on February 4, 2009
Stupid decision. Why would you buy two of the same drive? Always buy different drives. Sometimes you have drive series with high failing rates. This is why you don't want to have the same brand.
And sometimes you have particularly reliable drives. Like Rafaelloello said, I should have a pretty good idea by the end of the first 30 days. Worst case scenario, you're right and both drives take a crap, I've got my OS and my data backed up.
You seem to have little understanding what a backup is. Mentioning RAID makes things even worse.
Uh, okay. I have redundant backups, including an external drive in a floor safe and off-site, all of which are encrypted. But please, do enlighten me. (Okay, so I couldn't ignore the insulting tone, sue me...)
BTW, "mentioning" RAID doesn't make anything worse. If I would have just gone ahead and done it without research, that would make it worse. I did my due diligence and have decided against it. So what's the problem?
posted by jluce50 at 7:09 AM on February 4, 2009
« Older Where can I watch German television while in TX...... | References to Video & Web production having... Newer »
This thread is closed to new comments.
Not sure about performance implications but I really like having mirrored hard drives to avoid downtime and buffer you against a single hard drive failure. It's not a backup and doesn't replace one but I've had enough hard drives fail that it's nice to not have a hard drive failure take you out of commission while you restore backups and so on.
posted by pombe at 12:08 PM on February 3, 2009