Just as patients have to trust their doctors, non-techies have to trust the advice they get from techies. My last posting was about an article in a newspaper that offered, what I felt, was questionable advice on setting up a WiFi wireless network. The July issue of PC Magazine recently arrived in my mailbox and it offers some advice on backing up your computer that is also, to me, questionable.
The article is called "Keep Your Data Safe" and doesn't seem to have been posted yet on pcmag.com.
One section of the article discusses external hard drives (page 72), an excellent medium for storing backup files. The sub-topic on "Multidrives" is what prompted this posting. The magazine defines products in this category as external hard drives that internally contain multiple hard disks (separate and distinct from Network Attached Storage).
RAID 0
One of the products is said to "...hold a pair of 1TB drives for a total of 2TB in a RAID 0 configuration or 1TB of RAID 1 storage.." Another product "uses two 2.5 inch hard drives that are internally connected to get you 500GB of speedy RAID 0 storage."
The term "RAID 0" does not belong in an article about backing up files.
Any hard disk can and will fail. Storing files on a hard disk without backup is playing Russian roulette with your data. You probably knew that. Storing files on a RAID 0 device is playing Russian roulette with two or three bullets in the chamber instead of only one.
RAID 0 is designed for speed. Simply stated, it puts half of each file on each hard disk. It does this because hard disks are brutally slow compared to the CPU and RAM so being able to read from two hard disks at the same time to get a file speeds things up. Fine.
But RAID 0 offers this speed at the cost of reliability. You are more than twice as likely to lose data with RAID 0 as opposed to a single hard disk. For starters, if either hard disk fails, you're facing a data loss. In addition, you may lose everything if the RAID 0 controller fails. A controller may fail in a regular hard disk too, but the RAID 0 controller is keeping track of which half of each file is where. Reconstructing data after a RAID 0 controller failure is likely to require serious expertise and cost thousands of dollars.
To me, suggesting the use of RAID 0 in article about backup is a serious mistake.
A RAID Controller
(Credit: Adaptec)RAID 1
As an earlier quote illustrated, the article also talked about "multdrives" using RAID 1. While RAID 1 is designed for reliability, it's a bit out of place in an external hard disk. RAID 1, presents a single hard disk image to the outside world, but internally it uses two hard disks each one an exact mirror of the other.
The protection offered by RAID 1 however is limited. It doesn't help if you accidentally delete a file or logically corrupt it (where's that missing section from that report?). It doesn't protect you from file system errors such as lost clusters and the like. And it introduces the RAID controller as something else that can fail and isn't cheap. Finally, external hard disks are usually sealed, so if one hard disk in a RAID 1 configuration fails, the protection it offered is gone for good.
RAID 1 makes sense in servers that need high availability, such as those running this very website. It also makes sense in a NAS (Network Attached Storage) device where you can replace a failed hard disk easily. But, for external hard disks used for backup, you are safer with two independent devices as opposed to a single device running RAID Level 1. It may even be cheaper.
Western Digital's Raptors in RAID-0: Are two drives better than one? at anandtech.com (July 2004) offers a good introduction to RAID Level Zero and Level One. It also reports on seven different speed tests of RAID 0. The conclusion? Quoting "...there is no place, and no need for a RAID-0 array on a desktop computer. The real world performance increases are negligible at best and the reduction in reliability, thanks to a halving of the mean time between failure, makes RAID-0 far from worth it on the desktop."
My first real posting on this blog was about RAID Level Zero, see Don't get burned by RAID Zero and Following up on RAID Level Zero.
See a summary of all my Defensive Computing postings.
My previous posting (Don't get burned by RAID Zero) on RAID Level Zero was a warning - both that it can be dangerous and that you may inadvertently be using it.
To make an analogy, consider the old joke regarding the purchase of a yacht. The potential buyer asks how much the yacht costs and the seller responds "If you have to ask, you can't afford it." In the current context: if you don't know what RAID Zero is, you shouldn't be using it.
But, the devil is in the details.
The danger with Raid Zero lies in the fact that every file is split between two hard disks and if either fails, you lose all your data. Thus, despite some reader comments to the first posting, your odds of being victimized by a hard disk failure are twice those of someone using a single hard disk. Like a lottery drawing, you have two balls in the hopper.
The hardware device that keeps track of where to store each half of a file is called a controller. If this controller fails, you again lose all your data, even though each hard disk may be alive and well. Your data is there, but without a road map, it can't be found.
A reader disagreed, saying that a Raid Zero controller can simply be swapped out for a new one and your data is not lost. This is not the case.
For one, external hard drives, whether they have one or two internal disks, are not meant to be user serviced. And, it turns out all Raid Level Zero controllers are not the same.
Ontrack Data Recovery
For help with the issue of replacing a RAID Zero controller, I turned to a company famous for being world class experts in hard disks - Ontrack Data Recovery.
If you are not familiar with them, Ontrack is a vendor of last resort. That is, when all else fails and you absolutely must recover the files on a malfunctioning hard disk, call Ontrack. And, to repeat the joke above, if you have to ask how much their services cost, you're data files are not that important.
Jeff Pederson, Manager of Data Recovery Operations for Ontrack Data Recovery, weighed in on the issue of replacing one RAID Zero controller with another:
As a typical engineering response to your question of whether all raid level zero controllers are exactly the same, my answer is that they are and they are not. Obviously the way that they are all alike is in how essentially they stripe data between disks, but they are not all alike in how they go about accomplishing that task.
As far as the next question of difficulty and expense related to recovering from a raid controller failure, Ontrack has developed tools to overcome any major technical difficulty so we can usually overcome nearly every scenario related to raid failures, whether the problem is one or both of the drives have physically failed, the raid controller was reinitialized and it does not recognize the original volumes that were on the disks, or even when data has been re-written to the raid 0 after a re-initialization we have been successful in recovering underlying data as well.
As far as the commenter indicating that if the controller fails, all the data is lost, that is right up our alley and is completely untrue. We have recovered data from Raid 0, 1, 3, 4, 5, 10, 50, etc.
As with most of these situations if customers contact the controller manufacturer to discuss their particular situation with them, there are ways for controllers to be replaced and have them identify the original raid configuration if the drives are still operational.
It's fair to say, that if the RAID Zero controller fails, you're in deep trouble.
This is in stark contrast to hard disks connected to normal ordinary controllers, be they IDE/ATA or SATA. These controllers do indeed all function in the same way and, should one fail, it's a fairly simple thing to connect a hard disk to another one.
Scott Meuller
In the original posting, I mentioned a failed LaCie external hard drive with two internal hard disks, configured as RAID Zero. This was the first time I'd run across an external hard drive, being sold as a single unit (as opposed to a NAS device), with two internal hard disks. Scott Meuller, who was nice enough to add his thoughts on the subject, has been warning people about problems with this design all along (see Got a BIG drive? Then where do you keep the other two?). Quoting:
"... while their designs and shortcomings are obvious to a professional, virtually none of the published product reviews I've seen point out the multiple internal drive/RAID 0 configuration or the potential ramifications."
As for the failure of the RAID Zero controller in one of these units, Scott says it might "be possible to swap the otherwise standard internal drives over to another identical unit in order to restore array functionality (recover the data)."
When it comes to PC hardware, Scott Meuller literally wrote the book on it. Hordes of techies learned the ins and outs of hardware from his line of Upgrading and Repairing PCs books. If he isn't sure how to deal with this problem, it's not one you want to be faced with.
Never mind, that to have an "identical unit", you likely have to have purchased both at the same time.
Other Reader Comments
Many comments on the initial posting mentioned backups as a cure for data loss from a hard disk failure, RAID or no RAID. There will be many postings on this blog in the near future about backing up data on a computer. It's inherent to Defensive Computing.
Thanks to ajhoughton for helping to make my point.
To MC: what is LDO?
Update: July 15, 2007. While browsing the web site of Gateway Computers today, I happened to notice that they offer RAID Zero as an optional feature on the FX530XG computer. Quoting Gateway: "Experience pure power with optional RAID 0 with SATA II/300 drive support for improved performance."
- Original posting: Don't get burned by RAID Zero
- Comments on the original posting
To a computer nerd like myself, RAID refers not to a bug spray but to various ways of hooking together multiple hard disks. The various approaches are referred to as levels. Raid levels one through five are designed to decrease the chances that a hard disk failure will result in lost data. Typically RAID configurations are used in server machines as opposed to personal computers.
Raid level zero, however, is the black sheep of the RAID family. It's goal is performance rather than reliability. I'm writing this posting because two of my clients have been burned by their inadvertent use of RAID level zero. Consider this a word to the wise.
Client one purchased an external hard disk from LaCie. This person needed lots of storage space and, at the time, this particular model was top of the line, offering the most storage capacity. The reason it offered more storage than cheaper competing models is that internally there were two 3.5-inch hard disks instead of the usual one. While it looked to the outside world like one chunk of 500 gigabytes, the internal reality was that there were two 250 GB hard disks.
Client two owned a high end Dell XPS tower with two internal hard disks. This person wanted the latest and greatest and fastest computer. Thus, Dell configured the two internal hard disks for speed--RAID level zero. Like the old saying goes, be careful what you wish for, you may get it.
The hard disk is significantly slower than the processor and RAM. Thus to get the fastest read/write performance, RAID level zero stores half of a file on one hard disk and half on the other. Using both hard disks in parallel should reduce the total time needed to write a file.
To someone designing a computer system, the term single point of failure is like kryptonite to Superman. The term refers to a single point in the system which should it fail, would bring down the whole house of cards (so to speak). In an automobile, each tire is a single point of failure, thus they all have a spare in the trunk.
In a personal computer the hard disk is a single point of failure. However, in a RAID level zero configuration, there are three separate single points of failure involving the hard disks. Should either hard disk fail, all is lost because each disk contains half of every file. Oops.
In addition, both of my clients were also dependent on the RAID hard disk controller, the thing with the intelligence to split files as they are written out to the hard disks and re-assemble them back together when read. If the hard disk controller fails, the files may still be alive and well and happy on the hard disks, but you can't read them. Recovering from a RAID level zero controller error might cost thousands of dollars, as its a job for the most sophisticated data recovery companies.
RAID Zero with a failed disk. Click for full-size.
In the case of the client with the Dell computer, it was fairly easy to determine that the problem was with one of the hard disks. Replacing it, re-establishing the Raid zero environment and then restoring a disk image backup got the machine up and running. Disk image backups will be the subject of an upcoming posting on this blog.
The client with the external hard disk lost everything. Not that it mattered, but I couldn't even determine if the problem was with one of the hard disks or the RAID controller.
Update: July 8, 2007. Originally I had said the hard drive was the slowest thing inside the computer, except for the fan. Someone pointed out that optical drives are even slower, so that sentence now says the hard disk is slower than the processor and ram, which was the point I was trying to make.
Note: See
the comments on the original posting.
Update: July 13, 2007. See Following up on RAID Level Zero
- prev
- 1
- next
