Sign in with
Sign up | Sign in
Your question

Hard Disk...RAID Land Mines

Last response: in Components
Share
February 12, 2007 2:07:20 AM

My recent experience with ON BOARD NVidia SATA RAID has led me to believe ALL RAID SYSTEMS are vulnerable to complete data loss.

My experience begins with an MSI motherboard purchase 4 years ago with NVidia SATA RAID on the motherboard. I ran my system 2 hard drives mirrored.

6 months after my purchase the motherboard blew up. Not wanting to wait for MSI to repair the motherboard under warranty, I bought a 2nd identical board so I could have my computer running while the original motherboard was being repaired. I simply(that is a misnomer) switched RAM, CPU, DRIVES and up and running.

Last month, about 4 years after my original purchase, the newest motherboard died. So, due to my 4 year rule, I figure it was time to upgrade anyway.

In my mind, the most important motherboard option was SATA RAID. I thought I could just plug in my previous SATA drives, with faster CPU and RAM and I would be up and running.

WRONG. After being unable to get my drives running, and only getting a "YOUR COMPUTER MAY HAVE BEEN INFECTED BY A VIRUS" message from WIN XP, I contacted NVIDIA.

They emailed me back saying the only way to get my drives running was to PURCHASE a board identical to the one the drives were originally in.

WHAT???? ARE THEY KIDDING ME?

Where do you purchase a motherboard that hasn't been made in 3+ years!

Well, fortunately, I still had the motherboard they repaired from 4 years ago. I put it in, with my old CPU, RAM and drives, and up the PC came.

Wow.

I have talked about this with my "tech" friends, and they were astonished to the point of not believing me...until one of their SATA RAID motherboards croaked. They were striping, and could not get the drives to read properly after several motherboard attempts.

Is there anyone out there who can offer a solution to this problem? If not, why is ANYONE recommending a RAID SYSTEM for ANYONE? Wouldn't this be true for an NAS RAID, or external RAID as well?

I understand if you have DOZENS OF RACKS, you probably have other backup systems. But if you are a home/SOHO use like myself, MIRRORED RAID seems like a waste to me. You are protected against a drive failure, but SOL on a board failure.

Thanks...and be careful.
February 12, 2007 3:33:48 AM

This is because almost all of the integrated RAID solutions are not using true hardware RAID, but rather some form of software RAID. True, it is being done in the BIOS, and not in the actual OS, but SW none the less. If they used actual HW RAID, MBs would cost an extra $200-300 - hmmm they're almost there already, aren't they?

If you use something like a 3ware HW RAID Controller, the only piece you have to keep a spare of is the controller, and as long as the controller doesn't use a proprietary format, it doesn't even need to be the same brand.

Do keep in mind, that RAID isn't backup, although some varieties are fault-tolerant. Even RAID 5 will only survive the loss of a single drive.

edit: any system can have a single point of failure, the point is to attempt to minimize these when you assemble yours. Integrated RAID is there because hard-core gamers like to stripe & you get the others for free.
February 12, 2007 11:58:46 PM

As yosfem noted, the on-board RAID controllers are essentially software implentations. And the advice about using a stand-alone hardware controller is very good as well.

I would like to draw the following trivial details to your attention:

1) the on-board RAID controllers on various MoBos, from the same manufacturer have been known to be mutually incompatable friom one revision to a newer one of the same model. There have been a few of posts to this effect in these forums. You may want to consider searching the forums for more info.

2) on-board RAID controllers are a fairly recent addition to MoBos. And until the last year or so, the only options were RAID 0 or 1. RAID 5 and 0+1 or 1+0 are also very recent additions, and are equally subject to the limitations of software controllers constrained by the MoBo and chipset. What is scary in this context is that different manufacturer's products, that ostensibly use the same chipset, are incompatable. If you are going to use a software RAID controller under Windows, you would be wise to try and find a stand-alone piece of software that is not dependant on the MoBo. Such things exist, but are uncommon and fairly expensive. Linux has dealt successfuly with this issue for at least the past 8 years.

To answer your question about why RAID is recommended you should follow up the following link:

http://www.pcguide.com/ref/hdd/perf/raid/

Note that RAID is NOT a substitute for proper data back-up. The main purpose of RAID is to provide harddrive fault tolerance in mission-critical applications like file servers. In the "good old days" all RAID controllers used under such conditions were stand-alone hardware cards. They allowed things like hot-swapping failed drives and real-time re-building of arrays, but they were (and still tend to be) expensive.

In the case of NAS and external RAID, the most important question is the controller used in the box. If it is a real, stand-alone, hardware based controller, then you are good to go. If it is a flavour of software controller, beware, but investigate thoroughly. Note that a full-on hardware RAID controller will add at least $200 to $300 to the price of the system, but performance will be very good.

A RAID 0 setup will provide a significant improvement in read and write times. Tests at THG have shown that the performance of a pair of 7200 RPM drives configured as a RAID 0 setup reach performance levels that rival the native performance of a 10,000 RPM WD Raptor drive. RAID 0 is recommended for applications like streaming AV and video editing due to performance advantages. But see below for the limitations and shortcomings of RAID 0. See also the PC Guide link above.

RAID 0 is not actually a RAID system, because there is zero redundancy or fault tolerance in the specification. In fact, the liklihood of a system failure in a RAID 0 setup, due to problems with the HDD, is greater than the liklihood of an individual drive failing. This is because combined error is not additive, it is multiplicative. In other words, the probability of a failure is equal to X times Y, not X+Y, where X and Y is the individual expected failure rate of the two hard drives in a basic RAID 0 setup.

As a follow-on to the above point, you may wish to consider the fact that essentially ALL HD manufacturers have reduced the warranty on their product from 3 years to one. This was done almost 2 years ago, and was discussed in an article on THG. This warranty length reduction has some interesting and serious implications and ramifications, especially in the context of the increased use of RAID 0 and 1 setups in consumer-level systems. Also note that the reduced warranty length is not applicable to the "enterprise / business-class" lines. These also happen to be considerably more expensive than the consumer-class lines. Co-incidence? I don't think so!

If you peruse the threads in these forums, you will note the majority of posters that are self-professed gamers tend to advocate for RAID 0 and also express lack of concern for the limitations of RAID 0. It is my personal opinion that this is inappropriate and that all sides of the issue need to be considered carefully. That very extreme and radical concept known as "INFORMED CONSENT", don't ya know.

Hope this is of use.
Related resources
February 13, 2007 10:43:56 PM

Thanks for the nod, Wiz. Very accurate post w good advice - except this 1 nitpick. RAID 0 is RAID, it's simply not fault-tolerant - otherwise totally accurate. I was around what the various RAID specs were being nailed down - sorry :D 
February 14, 2007 2:55:57 PM

Quote:
Thanks for the nod, Wiz. Very accurate post w good advice - except this 1 nitpick. RAID 0 is RAID, it's simply not fault-tolerant - otherwise totally accurate. I was around what the various RAID specs were being nailed down - sorry :D 


Thanks for the support.

Just one question about your comment re RAID 0: doesn't the "R" stand for REDUNDANT? Where is the redundancy in RAID 0?
February 14, 2007 3:29:50 PM

Quote:
Thanks for the nod, Wiz. Very accurate post w good advice - except this 1 nitpick. RAID 0 is RAID, it's simply not fault-tolerant - otherwise totally accurate. I was around what the various RAID specs were being nailed down - sorry :D 


Thanks for the support.

Just one question about your comment re RAID 0: doesn't the "R" stand for REDUNDANT? Where is the redundancy in RAID 0?

Well, RAID 0 is considered a RAID level, but it provides no redundancy. It was a bad idea for someone to include 0 in the RAID levels, but what does it matter as long as the user of the box understands their chance of losing data has increased (i.e. more physical parts = increased chance of failure)?
February 14, 2007 4:34:30 PM

Main Entry: re·dun·dant
Pronunciation: -d&nt
Function: adjective
Etymology: Latin redundant-, redundans, present participle of redundare to overflow -- more at REDOUND
1 a : exceeding what is necessary or normal : SUPERFLUOUS b : characterized by or containing an excess; specifically : using more words than necessary c : characterized by similarity or repetition <a group of particularly redundant brick buildings> d chiefly British : no longer needed for a job and hence laid off
2 : PROFUSE, LAVISH
3 : serving as a duplicate for preventing failure of an entire system (as a spacecraft) upon failure of a single component
- re·dun·dant·ly adverb

According to M-W.com, using the primary definition of redundant, RAID0 *is* redundant.

To the guy who doesn't think disks are inexpencive now obviously didn't pay >$300 for their first 10M disk. No typo, Ten Megabytes.
February 14, 2007 4:38:06 PM

By software RAID on the MoBo's, do you mean that it is essentially controlled by a software routine in the BIOS or something to that effect, even though there is a "chip" on board for RAID functions. What about the inclusion of a "Silicon Image sil 3114 controller" as some boards offer? Is this also a form of software RAID? Read in a recent post where someone had an external card and found out even that was incompatible with a different card.

I am suprised anyone who offers RAID solutions, namely RAID 1, does not also include a way (software) to at least recover data from one disk, ie put it in another computer and run the software to either make it non-RAID readable or at the very least be able to back it up to another drive.

Also there are some very affordable drives now with a 3 or 5 year warranty. Each of mine only cost about $20 more for a more durable version and I would say the performance was within about 90% of the standard drive.
February 14, 2007 10:23:25 PM

Quote:
Main Entry: re·dun·dant
Pronunciation: -d&nt
Function: adjective
Etymology: Latin redundant-, redundans, present participle of redundare to overflow -- more at REDOUND
1 a : exceeding what is necessary or normal : SUPERFLUOUS b : characterized by or containing an excess; specifically : using more words than necessary c : characterized by similarity or repetition <a group of particularly redundant brick buildings> d chiefly British : no longer needed for a job and hence laid off
2 : PROFUSE, LAVISH
3 : serving as a duplicate for preventing failure of an entire system (as a spacecraft) upon failure of a single component
- re·dun·dant·ly adverb

According to M-W.com, using the primary definition of redundant, RAID0 *is* redundant.

To the guy who doesn't think disks are inexpencive now obviously didn't pay >$300 for their first 10M disk. No typo, Ten Megabytes.


Thank you for that clarification. But in the context of the purpose and intended function of RAID, the third definition you cite is the applicable one. And using that definition, RAID 0 isn't actually RAID. It would be more correct and accurate to describe it as AID. But as yosfem noted, this is how it was labelled when the definitions were being set up.

I raise the point in order to make it explicitly clear that a Level 0 array is not secure or fault tolerant in any meaning of the term.
February 14, 2007 10:35:54 PM

Simonetti wrote:

Quote:
However, "I" also stands for Inexpensive. There is nothing inexpensive about disks nowadays.


And what perception-altering substances have you been imbibing to make this statement?

I can remember buying a 2.5 GB HD for $600 retail and knowing I was getting a good deal. Today, I can buy a 300+ GB HD for around $150 to $200. On what basis do you say that hard drives are expensive to-day? Even the Seagate Cheetah SCSI 15K RPM drives are cheap at the prices charged for them now.
February 14, 2007 10:58:24 PM

Quote:
By software RAID on the MoBo's, do you mean that it is essentially controlled by a software routine in the BIOS or something to that effect, even though there is a "chip" on board for RAID functions. What about the inclusion of a "Silicon Image sil 3114 controller" as some boards offer? Is this also a form of software RAID? Read in a recent post where someone had an external card and found out even that was incompatible with a different card.

You hit the nail on the head. This is why even different revs of the same model of MoBo from the same manufacturer have compatability issues wrt RAID implementations.

The specific controller chip is irrelevant in this case. The reason is that the different manufacturers apply their own "tweaks" to the BIOS and controller chips on the MoBo. See the different settings in BIOS between different MoBo manufactureres using the same brand of BIOS chip.

Re stand-alone RAID controller cards, see comments above re MoBo manufacturers. Why would you assume the situation here would be any different?

I am suprised anyone who offers RAID solutions, namely RAID 1, does not also include a way (software) to at least recover data from one disk, ie put it in another computer and run the software to either make it non-RAID readable or at the very least be able to back it up to another drive.

Why are you surprised? All of the manufacturers are competing, and pushing the "advantages" of their line. Providing a long term solution to the support of competitor's products is not high on their "to-do" list, especially since they won't evenm support their own products.

Also there are some very affordable drives now with a 3 or 5 year warranty. Each of mine only cost about $20 more for a more durable version and I would say the performance was within about 90% of the standard drive.

Well, you are a lucky so and so aren't you? Don't assume that this is the norm out there.
February 14, 2007 11:04:47 PM

Quote:
WizardOZ,

point taken :!:

I've paid US$250 for 80 Mb Western Digital Caviar, back in the early 90's. I still have the sales slip!!!

OTOH, considering that I am now living in Brazil, things around here are between 150-200% more expensive than they are in the US.


Even $400 for a 300+ GBHD is cheap. I once paid $500 for a 450 MB (not a typo) HD. I have a 40 GB drive that I paid $150 for. Quit your whining.
February 14, 2007 11:22:41 PM

Yup, almost all controllers, be they hardware or software, inbuilt or add-in, use different data structures, and the drives are not transferrable between them. Sometimes, even different firmware revisions are incompatible.

This is why high end RAID systems will have RAID 0/1, or 5/1 or 6/1 nested RAID across two controllers, for controller redundancy as well as drive redundancy.

To be honest though, if you are changing the motherboard, you want to be doing a full windows re-install anyway.
February 14, 2007 11:38:06 PM

The I doesnt stand for inexpensive it stands for independent.

anyway to OP, you are correct in saying every array is prone to data lose, however like others have said as long as you use have a seperate raid card you can move the drives to that card, plus an actual card failure is less likely then a motherboard failure. Also raid is not a subsitute for a backup that can be restored to any system. You can get close though if you buy a couple of raid cards for backup and use a raid 60 then you can handle 4 drives dropping off at the same, though you need a minimum of 8 drives.
February 14, 2007 11:38:36 PM

as i have pointed out in other posts, RAID specs define behavior, but not implementation. every controller (manufacturer) is left to determine the precise disk structures that it will use to accomplish the required behavior. ergo, RAID spindles are not interchangeable. any RAID controller which does not have an exact twin stashed away in your workbench is a single-point-of-failure.

any of those "computer techs" who were shocked by that fact should, perhaps, be in another line of work.
a b V Motherboard
February 15, 2007 12:22:36 AM

this is what you get if you buy MSI, there absolute RUBISH

raid 0 - its not a standard, every raid controller writes data to the hdd's in its own way, its safer to get a raid card so you can carry it between computers etc.

Besides, using raid 0 is a risk in its self, if your using raid 0 atleast have another hdd for backup or better yet another hdd set in raid 1 - i use raid 0 in all my machines and i have a dedicated server with 2 x raid 1 sets because of the risks (although i have never seen a fault with my raid 0 arrays yet).

This fault is on you sorry, you cannot blame raid 0, or the motherboards (although if you buy MSI then you have to expect the worse ;) )

LEARN TO BACKUP PEOPLE
February 15, 2007 2:07:11 AM

darkangelism wrote:

Quote:
The I doesnt stand for inexpensive it stands for independent.


Actually, in the original definition / implementation, it did stand for inexpensive.

The "independant" interpretation didn't become relevant / applicable until ATA-IDE drives became common and relatively inexpensive.

It would be useful if you kept in mind that in the original implementation and, even today, for serious and major applications the interface of choice remains SCSI. In the "old days" SCSI drives came in two flavours - expensive enterprise and much less expensive consumer models. To this day, the interface to geat remains SCSI.

To Apache_lives:

MSI are actually a pretty good brand of MoBos. Your comments may be based on personal experiencce with a particular run. That's fine as far as it goes, but I have seen many brands of MoBo go through bad lots. And as far as it goes, in my own experience, Asus MoBos are garbage. And I make this comment on the basis of dealing with Asus product over masny years and models. The one brand that I will never, under any circumstances, recommend is Asus. And more experienced and knowledgable individuals who I trust have said the same thing about this particular brand. You have your opinion, I have mine. And so it goes.
February 15, 2007 4:15:15 PM

WizardOZ wrote:
Quote:
Just one question about your comment re RAID 0: doesn't the "R" stand for REDUNDANT? Where is the redundancy in RAID 0?


Sorry so long to answer -
RAID 0 has redundant disks, and uses the capacity of both. It's the fact that you have more than a single drive in a single unit or volume that earns it the R for redundant. RAID was revolutionary as it was the 1st open standard for putting multiple disks in a single volume. I remember looking at it and saying "Why would anybody use any of this except RAID 5?" Glad the SIG wasn't as short sighted as I was :lol: 

Inexpensive vs Independent -
The original idea was to put those smaller drives nobody wanted anymore to use now that we had these huge, honking 500M drives & the poor 50M/100M drives were gathering dust in the H/W closets. Inexpensive actually meant "already depreciated" or scavenged :roll: "Independent" was suggested as a joke when it was noted that people were pushing the spec by putting new drives in RAID boxes - the idea was to head off the objections from the bean counters before they even saw the word inexpensive. 8O
February 15, 2007 4:28:26 PM

Quote:
I am suprised anyone who offers RAID solutions, namely RAID 1, does not also include a way (software) to at least recover data from one disk, ie put it in another computer and run the software to either make it non-RAID readable or at the very least be able to back it up to another drive.

Why are you surprised? All of the manufacturers are competing, and pushing the "advantages" of their line. Providing a long term solution to the support of competitor's products is not high on their "to-do" list, especially since they won't evenm support their own products.


The software would be specific to that controller from that manucturer. I think the software would be a selling point knowing that you can recover the drive is the card fails, or you want to upgrade (preferably to their next model.)

With what has been posted and what I have learned recently, there is not much point to desktop RAID for any type of security. Two drives with frequent backups is almost better.

I had been under the assumption that RAID 1 "Mirrored" with only two drives created a Mirror of the Original drive without corrupting (for lack of a better term) the Original drive. I would think that would at least be an option for RAID 1 since it it appears their can be different methods for accomplishing it.
February 15, 2007 9:44:22 PM

Quote:
WizardOZ wrote:
Just one question about your comment re RAID 0: doesn't the "R" stand for REDUNDANT? Where is the redundancy in RAID 0?


Sorry so long to answer -
RAID 0 has redundant disks, and uses the capacity of both. It's the fact that you have more than a single drive in a single unit or volume that earns it the R for redundant. RAID was revolutionary as it was the 1st open standard for putting multiple disks in a single volume. I remember looking at it and saying "Why would anybody use any of this except RAID 5?" Glad the SIG wasn't as short sighted as I was :lol: 

Inexpensive vs Independent -
The original idea was to put those smaller drives nobody wanted anymore to use now that we had these huge, honking 500M drives & the poor 50M/100M drives were gathering dust in the H/W closets. Inexpensive actually meant "already depreciated" or scavenged :roll: "Independent" was suggested as a joke when it was noted that people were pushing the spec by putting new drives in RAID boxes - the idea was to head off the objections from the bean counters before they even saw the word inexpensive. 8O

Thank you for the clarifications and explanations.

Your posts have been very educational.

Man, those bean-counters sure do a rotten job of differentiating when it is good to pinch pennies and when it isn't.
February 15, 2007 10:47:57 PM

Quote:
I am suprised anyone who offers RAID solutions, namely RAID 1, does not also include a way (software) to at least recover data from one disk, ie put it in another computer and run the software to either make it non-RAID readable or at the very least be able to back it up to another drive.

Why are you surprised? All of the manufacturers are competing, and pushing the "advantages" of their line. Providing a long term solution to the support of competitor's products is not high on their "to-do" list, especially since they won't evenm support their own products.


The software would be specific to that controller from that manucturer. I think the software would be a selling point knowing that you can recover the drive is the card fails, or you want to upgrade (preferably to their next model.)

I see where you are coming from, and you are correct. Tragically, this won't happen - it would cost extra money.

Some MoBo manufacturers releaase at least 10 BIOS updates for each revision of a device. This doesn't say anything positive about the QA/QC and programming process they use. One of the worst offenders for this is the largest and most popular manufacturer - Asus.

Some would say that the frequency of BIOS updates shows that Asus is "taking care" of their customers. I would ask, "Why are you doing such a poor job of programming in the first place"?

This is like Redhat Linux - 2 days after a new version is released, they have 10's of security patches up on the website. I contrast this with something like SuSe, Ubuntu, Debian or Slackware, where it takes at least 2 weeks for security patches to show up, and not anywhere as many. So who is actually doing a better job of programming and QC?

If the MoBo manufacturers can't even get the BIOS right, why do you assume that something as "trivial" as supporting the RAID implementation would even show up on their radar?

With what has been posted and what I have learned recently, there is not much point to desktop RAID for any type of security. Two drives with frequent backups is almost better.

Well, there it is: a rational and intelligent assessment of the current state of affairs. You really should check the link I posted earlier re RAID. You may want to follow-up on this link as well:

http://www.pugetsystems.com/articles.php?id=29

I got this from another thread here at THG. Very interesting points raised.

As I noted earlier, and yosefm has also said, RAID is designed to provide - with the exception of 0 level - hardware fault tolerance and some real-time data security. But only under specific conditions and RAID is NOT a substitute for a properly designed and executed data back-up system. The purpose of RAID and Backups are complenetary, but mutually exclusive. Note that the full definition of a back-up includes storage of the backed up data in an off-site location. You may also want to consider the question of media life and obsolescence when setting up true back-ups. Not only from the perspective of physical lifespan of the medium used, but how long the hardware will be supported. As one simple case, 10 years ago, the majority of tape drives used a connection to a 5 1/4" floppy drive-capable cable to work. And the last time you have seen a system that supports a 5.25" FD was? Not to mention a ntape drive.

Floppy drives are endangered, period.

And you thought the failure to support mere variations in RAID controller software was a big issue.

I had been under the assumption that RAID 1 "Mirrored" with only two drives created a Mirror of the Original drive without corrupting (for lack of a better term) the Original drive. I would think that would at least be an option for RAID 1 since it it appears their can be different methods for accomplishing it.

RAID 1 is a mirror setup. But whatever happens on D1 also happens on D2. Malware, physical issues, and so on. Your comment about "different methods" is irrelevant, given the definition of what RAID 1 is.

You really shouild look at the PC Guide url link.

February 16, 2007 2:51:44 PM

RE: Software and 5.25FD's

All very good info. Some BIOS and other changes are for the ever changing market place, but yes a lot of the software today is sloppy. Nostalgia remembers Lotus 1-2-3 on one 5.25FD, and loaded it all in to memory, while MS played the drive swap game, and easily left files open on your data disk.

Quote:


I had been under the assumption that RAID 1 "Mirrored" with only two drives created a Mirror of the Original drive without corrupting (for lack of a better term) the Original drive. I would think that would at least be an option for RAID 1 since it it appears their can be different methods for accomplishing it.

RAID 1 is a mirror setup. But whatever happens on D1 also happens on D2. Malware, physical issues, and so on. Your comment about "different methods" is irrelevant, given the definition of what RAID 1 is.

You really shouild look at the PC Guide url link.



I am familiar with it, though I poked around it again. I was not referring to viruses, malware, etc. What I was referring to above by mirrored and "corruption" was had been under the assumption that if a drive failed, or the controller failed (without damaging the data) a person could place the drive in another computer and easily recover the data, ie as easily as if it had been only a single drive in a PC. From previous posts it sounds like a single drive pulled from a RAID 1 configuration and placed in a different PC may have issues reading the data because the RAID controller is not present, ie the RAID controller manipulated (corrupted) the data so it is specific to the controller or the software version thereof.
February 16, 2007 7:45:24 PM

Quote:
RE: Software and 5.25FD's

All very good info. Some BIOS and other changes are for the ever changing market place, but yes a lot of the software today is sloppy. Nostalgia remembers Lotus 1-2-3 on one 5.25FD, and loaded it all in to memory, while MS played the drive swap game, and easily left files open on your data disk.



I had been under the assumption that RAID 1 "Mirrored" with only two drives created a Mirror of the Original drive without corrupting (for lack of a better term) the Original drive. I would think that would at least be an option for RAID 1 since it it appears their can be different methods for accomplishing it.

RAID 1 is a mirror setup. But whatever happens on D1 also happens on D2. Malware, physical issues, and so on. Your comment about "different methods" is irrelevant, given the definition of what RAID 1 is.

You really shouild look at the PC Guide url link.



I am familiar with it, though I poked around it again. I was not referring to viruses, malware, etc. What I was referring to above by mirrored and "corruption" was had been under the assumption that if a drive failed, or the controller failed (without damaging the data) a person could place the drive in another computer and easily recover the data, ie as easily as if it had been only a single drive in a PC. From previous posts it sounds like a single drive pulled from a RAID 1 configuration and placed in a different PC may have issues reading the data because the RAID controller is not present, ie the RAID controller manipulated (corrupted) the data so it is specific to the controller or the software version thereof.

The RAID controller doesnt corrupt anything. It just doesnt partition the drive in quite the same way. If one drive fails in a RAID 1 array, you boot as normal without putting it into some other computer. You can try this - make a RAID 1 array, unplug one of the drives, and everything will carry on working.

Most "corruption" is actually caused by writes that fail to complete because of power loss tbh, this would affect both drives in a RAID 1 array equally.
February 16, 2007 10:16:36 PM

Quote:


<snip>
The RAID controller doesnt corrupt anything. It just doesnt partition the drive in quite the same way. If one drive fails in a RAID 1 array, you boot as normal without putting it into some other computer. You can try this - make a RAID 1 array, unplug one of the drives, and everything will carry on working.

Most "corruption" is actually caused by writes that fail to complete because of power loss tbh, this would affect both drives in a RAID 1 array equally.


Very good points. Thanks for the clarification.

To pkquat:

I have an idea for a solution to the problem: the MoBo manufactureres shoud release a set og RAID driver disks that contain all versions used, with appropriate patches to enable earlier versions to work correctly on newer MoBos. Sadly, I don't see this happening.

The reason I can't see this happening is because the demand is too small. The only people who routinely use the on-board RAID capabilities are gamers, and they use RAID 0. There is the odd enthusiasat who will use RAID 1 at home, but the majority of consumers aren't interested in (or capable of) setting up and using any kind of RAID. The enthusiasts and gamers are on their own.
February 17, 2007 12:11:46 AM

It is sometimes the case that these things can be moved on.

For example, if you went from an AM2 or S939 nForce 4 board, to an nForce 680i LGA775 board today, you'd stand a good chance of getting the array to work on the new system, as they both use the same MCP.

I do see your point, but imho most people that are doing RAID on a motherboard chipset are doing RAID 0 for pure performance, and keep important stuff on another drive.

If you are changing motherboards, its a damn good idea to reinstall windows anyway, at which point repartitioning a RAID array doesnt make much difference anyway.
February 18, 2007 7:21:54 AM

Your point about selected compatability is interesting, and should be more widely publicized. But on a big-picture scale, all it does is confirm the comments about controller incomapatabilities. On-board and stand-alone.

As for your comments about people employing common sense about backing up data, even a casual perusal of threads in both this forum and the hard drives forum will quickly dispell that fantasy. I am amazed by the number of pleas for help in recovering data from failed RAID 0 arrays. Never mind the comments about problems when a MoBo upgrade has been implemented. Incredible number of potential Darwin Award contenders.

Re changing MoBos - or even brands of video cards - yes indeed the best option is wipe and re-install the OS. The problem in most cases is the detail that so many are applying the "one big drive" approach., They can't wipe and re-install the OS because they would lose all the data files they have. This is why a seperate partition for the OS remains a good idea.

But thanks for the useful information.

Stay Frosty
February 19, 2007 1:43:27 PM

Where I was going with this is if the controller fails, unless you have a 100% identical controller it is not as easy as just placing one of the drives in another computer. From one of my previous posts I used corruption "for lack of a better term" because it makes the drive(s) specific to the RAID controller.
February 19, 2007 2:44:05 PM

Especially given the history of RAID as an open standard, this is an extremely valid complaint, and rather ironic besides. Avoid proprietary RAID solutions like the plague. It's also a very good idea to research the recovery process - if it's not reasonably easy (and sane), pass on that solution.

Sort of like integrated sound, if you truly care about it, you invest in a discrete solution. If you can't justify the expense, you probably don't need it.
February 20, 2007 2:06:03 PM

All RAID solutions are proprietary....

If its 3Com, Promise, nVidia, VIA, Intel, whatever, they all use their own data structures.

Even full software RAID is the same, a WinXP Software RAID 0 set wont be read under Linux, and vice versa...
!