HW Raid5 as Parity Drive?


Recommended Posts

Buy a 6/8TB HDD fast speed is too expensive.

 

I'm thinking to do a Raid5 3x2TB as parity drive. Can do some problem?

 

I understand that the speed writing should increase because raid5.

 

3 2TB drives will almost certainly cost as much as a 6TB drive  :)    If, of course, you've using drives you already have, or are buying used drives, then you can indeed get 3 of them for a lot less ... and if you have a RAID controller that will work; then yes, you could create a RAID array for your parity "drive".

 

I'd make it RAID-0, however, as you'll get notably better write performance than you would with RAID-5, which is slowed down by the parity calculations.

 

If you're buying new drives, and already have a RAID controller to use; then the least expensive way to get 6TB is probably a RAID-0 array of 2 3TB drives.

 

Link to comment

Buy a 6/8TB HDD fast speed is too expensive.

 

I'm thinking to do a Raid5 3x2TB as parity drive. Can do some problem?

 

I understand that the speed writing should increase because raid5.

Raid5 3x2TB will only give you 4TB. Actually, even less, because most RAID cards do "capacity truncation", see here.

Link to comment

Four seconds late  :'(  :'(

 

Slowpoke  :)

 

I actually thought about the truncation issue -- which would also impact RAID-0 setups; but that's not an issue with MOST RAID cards, so I didn't mention it.  It was, of course, an issue with that nifty little gadget [Addonics AD2HDDHP6G]  you found in that thread, which unfortunately is useless due to the truncation (at least in the basic 2-drive setup).

 

 

Link to comment

Four seconds late  :'(  :'(

 

Slowpoke  :)

:)

I actually thought about the truncation issue -- which would also impact RAID-0 setups; but that's not an issue with MOST RAID cards, so I didn't mention it.  It was, of course, an issue with that nifty little gadget [Addonics AD2HDDHP6G]  you found in that thread, which unfortunately is useless due to the truncation (at least in the basic 2-drive setup).

BTW, it would be nice to have list of hardware RAID cards, capable of no-truncation. I see more and more people are interested in  RAID-0 parity setup.

Link to comment

Correct me if I am wrong, but doesn't using Raid-0 as parity increase your risk of array failure?

 

No.  It does slightly increase the risk that the parity "drive" fails, since if either of the drives in the RAID-0 array fails, then your parity "drive" just failed.    But the consequences at that point are no different than if a single parity drive had failed.  You simply replaced the failed "drive"  (in this case ONE of the component drives of the RAID-0 array); and then do a rebuild of the failed parity drive.

 

 

Link to comment

Correct me if I am wrong, but doesn't using Raid-0 as parity increase your risk of array failure?

 

No.  It does slightly increase the risk that the parity "drive" fails, since if either of the drives in the RAID-0 array fails, then your parity "drive" just failed.    But the consequences at that point are no different than if a single parity drive had failed.  You simply replaced the failed "drive"  (in this case ONE of the component drives of the RAID-0 array); and then do a rebuild of the failed parity drive.

 

That sounds like a Yes to me. Sure if nothing else goes wrong you can just rebuild the parity drive... but since you've increased your risk of a single drive failure, haven't you also increased the risk of a double drive failure?

 

Also I'm going to question the word "slight" since mathmatically the failure rate can be calculated as 1-(1-r)^n (Where r is the disk failure rate, and n is the number of disks in Raid-0), which means the failure rate is just slightly less then doubled with 2 disks and just slightly less then trippled with 3.

 

Edit: Additionally this risk only increases as you add more drives to your Raid-0 set up. 3 x 2tb drives in would have a much higher failure rate.

 

Link to comment

It depends on the failure rates of the drives in question. If the smaller drives are twice as reliable as the larger one, the reliability is the same. Another way to think about this is that it has the same impact as adding another drive to the array. Is that a reason to not add another disk?

Link to comment

Correct me if I am wrong, but doesn't using Raid-0 as parity increase your risk of array failure?

 

Yes, but the risk of data loss is less than adding an additional data drive.

 

For example, the risk of data loss by having a RAID-0 parity plus 7 data drives is a bit more than a single-drive parity plus 7 data drives, but less than single-drive parity plus 8 data drives.

Link to comment

Correct me if I am wrong, but doesn't using Raid-0 as parity increase your risk of array failure?

 

Yes, but the risk of data loss is less than adding an additional data drive.

 

For example, the risk of data loss by having a RAID-0 parity plus 7 data drives is a bit more than a single-drive parity plus 7 data drives, but less than single-drive parity plus 8 data drives.

 

I might have done my math incorrectly, but I don't believe that is true.

 

Assuming a failure rate of 1% per year per drive I calculated the following array failure rates

 

 

 

Math was wrong. I did not expand the polynominal correctly.

 

So two points:

1) Using Raid-0 as a Parity Drive is bad with 2 disks an far worse the more disks you add...

2) Using Raid-0 as a Parity Drive increases your failure rate without giving you any additional storage space, so while you are correct that adding an additional disk to your array does increase the failure rate at least you are getting additional storage space in return.

 

Link to comment

Correct me if I am wrong, but doesn't using Raid-0 as parity increase your risk of array failure?

 

Yes, but the risk of data loss is less than adding an additional data drive.

 

For example, the risk of data loss by having a RAID-0 parity plus 7 data drives is a bit more than a single-drive parity plus 7 data drives, but less than single-drive parity plus 8 data drives.

 

I might have done my math incorrectly, but I don't believe that is true.

 

Assuming a failure rate of 1% per year per drive I calculated the following array failure rates

 

P = 1, Array = 7 (8 disks total, 7 array disks) Failure Rate 0.27%
P = 2 in Raid-0, Array = 7 (9 disks total, 7 array disks) Failure Rate 1.19%
P = 3 in Raid-0, Array = 7 (10 disks total, 7 Array disks) Failure Rate 2.11%
P = 1, Array = 8 (9 disks total, 8 array disks) Failure Rate 0.34%

 

So two points:

1) Using Raid-0 as a Parity Drive is bad with 2 disks an far worse the more disks you add...

2) Using Raid-0 as a Parity Drive increases your failure rate without giving you any additional storage space, so while you are correct that adding an additional disk to your array does increase the failure rate at least you are getting additional storage space in return.

 

Won't comment on the math, but a RAID0 parity increases performance and allows you to leverage older drives rather than spending big dollars on a new very large drive.  Also remember parity only has to be good if one of the other drives fails. My experience is that drives are very reliable without parity and the chance of a drive failure occurring in parity in the couple days I might need it to work to do a rebuild are extremely extremely low.

 

Another perspective - you are better off using 2 3T drives as parity and a 6T drive for data, than the other way round.

Link to comment

That sounds like a Yes to me. Sure if nothing else goes wrong you can just rebuild the parity drive... but since you've increased your risk of a single drive failure, haven't you also increased the risk of a double drive failure?

 

No, it's not a Yes.  There are several considerations ...

 

(a)  It DOES increase the likelihood of a parity drive failure (but not by a lot, since most failures are due to unrecoverable read errors ... and THAT likelihood is not increased).    But a DRIVE failure is NOT an ARRAY failure --- that's the whole idea of a fault tolerant system.

 

(b)  Note that if one of the component drives of the parity "drive" fails, it's NO different than if a single drive of that size had failed.  So from an UnRAID perspective, it has NOT increased the risk of an ARRAY failure [A single drive failing is NOT an "array failure", since the array is fault tolerant].    And at that point the risk of a second drive failure is identical to what it would have been if the parity drive was a single drive.

 

©  Next, a high percentage of drive failures are due to unrecoverable bit errors.  The probability of this happening is the SAME for a RAID-0 array as it would be for a single drive, since virtually all consumer drives have bit error rates of 1 in 10^14th => so a disk with 1/2 as many bits will be 50% less likely to have an unrecoverable error ... thus the two (or three) drives in your RAID-0 array will have the same overall probability of an unrecoverable error as a single large drive with the same capacity as the array.

 

(d)  It is, of course, possible that a failure due to some other issue could happen, but the probability of this is very low; and if it happens it's effectively like simply using a less reliable drive in your array.  As I noted above,  it really doesn't matter anyway, since the array is still functioning just fine => and your next action should be to replace the failed drive, just as you would for any other failed drive in the array.

 

(e)  Note also that you could actually have a dual drive failure without losing data (!!) ... as long as the 2 drives that failed were both part of the RAID-0 array  :)

 

(f)  If you want to increase the reliability of the parity drive, you could, of course, use a RAID-5 array ... but this will have a significant write performance penalty relative to RAID-0.  It would, of course, then provide fault tolerance for the parity "drive" in addition to the overall array fault tolerance ... so you could have two drives fail without data loss as long as one of them was a member of the parity array.

 

Bottom line:  In the event of a parity drive failure, there is NO difference in the probability of a subsequent array failure during the rebuild.    If a data drive fails, there is a very small increase in the probability of a 2nd drive failure -- but NOT in the likelihood of an unrecoverable bit error rate, which is the most likely scenario that would cause a rebuild problem.    If you're concerned about the increased likelihood of failure, you could always use enterprise class drives for the parity array  :)  [These typically have an order of magnitude better unrecoverable bit error specs -- 1 in 10^15]

 

 

 

 

Link to comment

Assuming 1% annual failure rate per disk and RAID0 parity.

 

Annual probability of data loss:

 

                        # of parity disks               

                        0            1            2            3

# data  1          1.000%  0.010%  0.030%  0.060%

  disks  2        1.990%  0.040%  0.079%  0.139%

            3        2.970%  0.088%  0.148%  0.237%

            4        3.940%  0.155%  0.234%  0.352%

            5        4.901%  0.240%  0.338%  0.485%

            6        5.852%  0.342%  0.459%  0.635%

            7        6.793%  0.462%  0.597%  0.801%

Link to comment

Assuming 1% annual failure rate per disk and RAID0 parity.

 

Annual probability of data loss:

 

                        # of parity disks               

                        0            1            2            3

# data  1          1.000%  0.010%  0.030%  0.060%

  disks  2        1.990%  0.040%  0.079%  0.139%

            3        2.970%  0.088%  0.148%  0.237%

            4        3.940%  0.155%  0.234%  0.352%

            5        4.901%  0.240%  0.338%  0.485%

            6        5.852%  0.342%  0.459%  0.635%

            7        6.793%  0.462%  0.597%  0.801%

 

Don't agree with this at all.    This indicates, for example a tripling of the probability of data loss if you use a RAID-0 array for the parity drive with 1 data drive, and a doubling with 2.    This is NOT accurate.  A 1% annual probability of failure means there's a 1% chance a drive will fail in a YEAR.    The probability it will fail in the next 8 hours (or whatever time you want to assume for a rebuild) is FAR lower than that => it's far more likely any 2nd drive failure during that time will be due to an unrecoverable bit error; and the probability of that isn't impacted by the # of drives used to create the parity "drive".

 

Link to comment

I had a typo in my spreadsheet. These numbers confirm comments by bjp999 and bubbaQ.

Assuming 1% annual failure rate per disk and RAID0 parity.

 

Annual probability of data loss:

# of parity disks

                0              1         2         3

# of  1 1.000% 0.010% 0.020% 0.030%

data  2 1.990% 0.059% 0.078% 0.136%

disks 3 2.970% 0.117% 0.174% 0.230%

        4 3.940% 0.193% 0.268% 0.342%

        5 4.901% 0.287% 0.379% 0.470%

        6 5.852% 0.398% 0.507% 0.614%

        7 6.793% 0.525% 0.650% 0.774%

Link to comment

Very Interesting all statics info, thanks.

 

But If I use raid5, I have performance  + extra data protection(like second parity drive, that unraid not support native),

6TB drive 10K RPM is more expensive than 3 normal 3TB drive.

Anyway I could use two SSD 1TB Drive SSD(cache drive) to "solve/ignore" the unraid write speed problem.

 

Link to comment

But If I use raid5, I have performance  + extra data protection(like second parity drive, that unraid not support native),

 

This is not true. RAID5 parity would protect you from a double drive failure only in scenarios where one of the disks is one of your parity disks. Remember that parity is ONLY of value when a non-parity drive in your array fails. I would rather have a RAID5 on my most important data disk, than on my parity disk, at least it is protecting that data from a single failure in its RAID5 universe regardless of what happens to other drives in the array. Having extra redundancy on your parity is not very useful IMO.

 

I cannot tell you how rare these double disk failures have been. Disks don't just blink out like light bulbs. For the most part, they start to act up and show signs that you would start receiving warnings about long before they actually fail. Operating an array for almost 10 years, I have never had a drive fail on me in a catastrophic sense. One or two disks have started to act up and I have replaced them. Rebuilding a disk takes maybe a day. It is only during that day that a failure on the parity disk impacts data loss. And if it happened, and a read error had developed on parity (or a different disk) during the rebuild, I still have the disk that was acting up to recover data. What are the odds that one of the two or three disks in the parity RAID0 will choose such an inopportune time to have this issue AND that the failing disk were dead as a doornail and not just acting up? I loose no sleep worrying. :)

 

I could make a similar argument for the so called dual parity feature. Very very rarely would it be needed. But it would protect you from ANY two drives failing - less rare than a dual failure involving parity, especially in big arrays of say 25 disks. But dual parity has a very useful by-product (at least is should). If a parity error is detected, which might be a sign of bitrot or other cause, with the second parity disk unRAID should be able to triangulate which disk is the source of the parity mismatch and fix that disk. Currently unRAID assumes parity is wrong, which may be likely after a power loss but is not always the culprit. I consider dual parity especially useful for that reason, while acknowledging a small benefit from providing the dual drive failure recovery that might be helpful if your server floods and two drive drown (yes, it happened).

Link to comment
It DOES increase the likelihood of a parity drive failure (but not by a lot, since most failures are due to unrecoverable read errors ... and THAT likelihood is not increased).

 

The AFR of drives is not telling us anything about the failure itself. If unrecoverable read errors are included in the failure rate given by the manufacturer, then the likelihood will increase. The failure rate will double with an additional (identical) drive!

 

"I cannot tell you how rare these double disk failures have been. Disks don't just blink out like light bulbs. For the most part, they start to act up and show signs that you would start receiving warnings about long before they actually fail."

 

This is again, nothing else than apportioning the failure rate of the drive. Let's call them minor and major faults. But after you have done all that, all the "majors" and "minors" will be doubled again!

 

And in the end you still have a doubled failure probability of a raid0 parity drive!

 

The time at risk when you have to rely on the parity drive, namely in the event of a data drive loss, is not much higher though.

 

A far more serious issue I fear is the fault tolerance of such a build.

We know, an improperly seated cable can cause our array to go into "degrded mode" needing a lenghty parity rebuild/check.

How is a RAID0 array/controller acting when a drive is detected "failing soon" due to one of those "minor" faults you mention?

Finally you may have many drives in your box and the more complexity you add, the more risk you take.

If you need the performance, than you have to do it. If not, keep it simple.

Link to comment

Ya'll be missing the point (and the math) by a mile.  You didn't read my post carefully.  I said  DATA LOSS, not drive failure.

 

DATA LOSS requires a 2-drive failure.  Losing 2 data drives is a greater loss than losing 1 data + parity.

 

A 2-drive RAID-0 parity increases ONLY the likelihood of failure of parity.  In an N-drive system (N-1 data drives plus 1 parity) converting to a RAID-0 parity increases the DRIVE FAILURE rate exactly the same as adding a data drive and creating an N+1 size system.

 

Losing either drive in a RAID-0 parity or a single data drive, does NOT result in data loss.

 

With RAID-0 parity, you introduce a special 2-drive failure scenario that does not result in any data loss---loss of both parity drives in the RAID-0 array.

 

So the difference in DATA LOSS between single-drive parity and RAID-0 parity is the additional risk of a second drive failure.  Because there is one special case in the RAID-0 parity system that has no DATA LOSS in a 2-drive failure (both members of parity), the risk of N+1 verses an N-sized system with RAID-0 is slightly less than an N+1 system with a single parity.

 

But you "buy" that small decrease in risk by having less storage space for data since you are using 2 drives for parity.

 

But you don't do this for risk reasons... you do it for speed reasons.

 

I had a bunch of 2TB drives in unRAID, and when I needed to expand, I decided to redo my unRAID with 4TB data drives.  So I used HW RAID-0 for all those 2TB drives.  So Parity is a 2-drive RAID0.  A couple of data drives are 2-drive RAID0.  Cache is a 3-drive RAID0.

Link to comment

The AFR of drives is not telling us anything about the failure itself. If unrecoverable read errors are included in the failure rate given by the manufacturer, then the likelihood will increase. The failure rate will double with an additional (identical) drive!

 

If you look at failure statistics, you see failure rates of individual drives differ by 4-5X those of other brands. You can't say that a RAID0 of say HGST drives is more likely to fail than a single WD or Seagate drive with much higher failure rate.

 

Doubling of failure rate sounds bad, but when dealing with very small percentages, sounds much worse than it is. You might say that eating that candy bar or drinking that scotch is going to double, triple, heck be a 10x greater risk you will die today. But the chances are very small, so even an order of magnitude increase is an acceptable risk you take with your life every day. Now if your chances of dying today were 20%, doubling or tripling that risk, even adding only 50% additional risk, is a whole different ball game. We have to be careful that we are not over-sensationalizing here.

 

This is again, nothing else than apportioning the failure rate of the drive. Let's call them minor and major faults. But after you have done all that, all the "majors" and "minors" will be doubled again!

 

And in the end you still have a doubled failure probability of a raid0 parity drive!

 

The time at risk when you have to rely on the parity drive, namely in the event of a data drive loss, is not much higher though.

 

A far more serious issue I fear is the fault tolerance of such a build.

We know, an improperly seated cable can cause our array to go into "degrded mode" needing a lenghty parity rebuild/check.

How is a RAID0 array/controller acting when a drive is detected "failing soon" due to one of those "minor" faults you mention?

Finally you may have many drives in your box and the more complexity you add, the more risk you take.

If you need the performance, than you have to do it. If not, keep it simple.

 

I generally agree with keeping it simple. But a 2, 3 or even 4 drive RAID0 drive that is still smaller than a single commercially available disk, it is still keeping it relatively simple. If any of the drives in the RAID0 fail, the owner can rebuild it onto a single replacement disk and recover simply. I do agree that creating RAID5s of data disks is not a good idea, and hope my earlier response didn't imply it was. Mixing RAID5 with unRAID doesn't make much sense to me.

Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.