• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Raid 5 & 6 URE limitations

S

Silver565

n00b
Joined
Jan 9, 2012
Messages
41
Hi Everyone

I was hoping to get opinions on the drive capacity vs Raid 5/6 issue.

My understanding is that a URE will occur on a standard disk for around 12TB in a R5.

(source: http://www.zdnet.com/blog/storage/why-raid-6-stops-working-in-2019/805)

Has anyone found issues with this? It's basically put me off R5 & R6 all together. I'm only after an 8TB array, however... It's still not exactly reassuring as I plan to use 4TB drives

:confused:
 
For home use your better off using 2 4tb drives in the computer and 2 4tb removable backup drives than any kind of raid.
If your paranoid manually check sum the data monthly on all 4 drives.
 
Thanks,

Would a raid 10 not be safer? Especially if I were to run a hot spare?

I plan on having an old server that I turn on once a month running storage spaces as the backup(I won't need the performance etc for that one)
 
Your drives are the limiting factor in all cases. It sounds like you need a better understanding of them so I will try to summarize for you.

With RAID 5 you use 3 drives, 2 striped (performance) and 1 parity. Parity (simplified) works by comparing the first two values and says 1 if they are the same or 0 if they are different. So if first value on disk 1 is 1 and first on disk 2 is 0 then parity bit is 0. If a drive goes down then it compares between the 2 drives running to recreate the data, but,if 2 fail then it can't be recreated. RAID 6 uses an additional drive for parity or basically a mirror of parity. So if two drives fail then you can recreate the data. RAID 10 works by having 2 stripped mirrors. So two drives can fail but they have to be a drive from each set, so 1 and 3, 2 and 4, etc. If drives fail in the same set then the data can not be recreated. I won't go into the science but RAID 5/6 gives 2x the drive speed read and 2x write while RAID 10 gives 4x read and 2x write.

The question is what is your end goal? Any drive has a chance of failure and RAID helps keep data accessible during a failure. If you are looking for the highest probability of data to remain available during failures then go with RAID 6. If you are looking for the least likely way to lose data then just RAID 0 with 2 drives and make an offline backup (a drive without power has a lower percentage of failure).

Remember it is always a matter of probability, and there are always other factors. I recently read an article on a major private SVN/ GITHUB company who had top of the line redundancy etc and if everything died they would be able to have amazon web services fire up and keep their customers data available. A hacker got in and dumped all their data and redundancies and they didn't have a backup. They are working their hardest to get their customers data back but will cease to exist afterwards. You can't plan for everything. Hope this helps.
 
My understanding is that a URE will occur on a standard disk for around 12TB in a R5.
Click to expand...

In practise I have (a person with 100 to 200 drives in raid6/raidz2/raidz3) either been very lucky (got to play the lottery) or this is very wrong or maybe a combination of the two.. What I see from my arrays at work is bad drives will have many UREs (a rate much higher than 1 in 12TB). Working drives may experience a URE but not any where near 1 every 12TB.
 
Last edited:
stormy1 said:
For home use your better off using 2 4tb drives in the computer and 2 4tb removable backup drives than any kind of raid.
If your paranoid manually check sum the data monthly on all 4 drives.
Click to expand...

I agree with this.
 
Silver565 said:
Hi Everyone

My understanding is that a URE will occur on a standard disk for around 12TB in a R5.
Click to expand...

No, I suggest reading this.

http://www.raidtips.com/raid5-ure.aspx

Also, you can always buy hard disks with 10^15 and 10^16 URE rates if you are really that worried. They make consumer SATA disks with 10^15, but 10^16 are typically SAS disks.

I've been running a 28TB ZFS array with 15TB od data on it for 6 months and have scrubbed it a dozen times and have not seen a single URE yet while reading through about 180TB of data (15TB of data * 12 scrubs). And this is with WD Reds which I believe are only 10^14.
 
BillytheSkids said:
If you are looking for the least likely way to lose data then just RAID 0 with 2 drives and make an offline backup (a drive without power has a lower percentage of failure).
Click to expand...
Raid 0 is always a bad idea if you care about the data.
Leave them as separate drives.

Get a smart ups.
The most common avoidable cause of data loss are power problems.
All of my computers are on a ups other than the laptops and they are on apc surge strips.
 
Silver565 said:
Thanks,

Would a raid 10 not be safer? Especially if I were to run a hot spare?

I plan on having an old server that I turn on once a month running storage spaces as the backup(I won't need the performance etc for that one)
Click to expand...
Its all a matter of best use of funds.
Offline drives and a line interactive/smart ups are a better investment.
For home users a single drive will be fast enough.
You will rarely have simultaneous users drawing data and if you do it will be a max. of a 2-3 computers simultaneously.

In a work environment I would recommend more spindles of 10k drives to get to 8TB in raid 5 or 6/zfs. I don't like anything over 1.2tb per drive in work raid arrays and even that is pushing my limit. 600GB per drive is better but the economics forces 1.2s sometimes.
 
drescherjm said:
In practise I have (a person with 100 to 200 drives in raid6/raidz2/raidz3) either been very lucky (got to play the lottery) or this is very wrong or maybe a combination of the two.. What I see from my arrays at work is bad drives will have many UREs (a rate much higher than 1 in 12TB). Working drives may experience a URE but not any where near 1 every 12TB.
Click to expand...


I agree with this.
 
If you have 10k or 15k drives (Fast Class), RAID 5 is fine. But if you're using slower 7,200 or 5,400 drives (Nearline), you should go RAID 6. Why?

Nearline drives are slower and larger, which means they take much longer to rebuild. But they are cheaper. This means RAID 6 is an option to protect against a second failure during rebuild because it's cost effective.

Fast Class drives are faster and smaller, which means they rebuild much faster. But they are more expensive. This means RAID 5 is an option because it saves money, and has less chance of a second failure during rebuild due to the speed.

The other opinion is that anything larger than 2TB should be RAID 6 simply due to rebuild time. You don't find FC drives larger than 2TB.
 
Last edited:
BillytheSkids said:
Your drives are the limiting factor in all cases. It sounds like you need a better understanding of them so I will try to summarize for you.

With RAID 5 you use 3 drives, 2 striped (performance) and 1 parity. Parity (simplified) works by comparing the first two values and says 1 if they are the same or 0 if they are different. So if first value on disk 1 is 1 and first on disk 2 is 0 then parity bit is 0. If a drive goes down then it compares between the 2 drives running to recreate the data, but,if 2 fail then it can't be recreated. RAID 6 uses an additional drive for parity or basically a mirror of parity. So if two drives fail then you can recreate the data. RAID 10 works by having 2 stripped mirrors. So two drives can fail but they have to be a drive from each set, so 1 and 3, 2 and 4, etc. If drives fail in the same set then the data can not be recreated. I won't go into the science but RAID 5/6 gives 2x the drive speed read and 2x write while RAID 10 gives 4x read and 2x write.

The question is what is your end goal? Any drive has a chance of failure and RAID helps keep data accessible during a failure. If you are looking for the highest probability of data to remain available during failures then go with RAID 6. If you are looking for the least likely way to lose data then just RAID 0 with 2 drives and make an offline backup (a drive without power has a lower percentage of failure).

Remember it is always a matter of probability, and there are always other factors. I recently read an article on a major private SVN/ GITHUB company who had top of the line redundancy etc and if everything died they would be able to have amazon web services fire up and keep their customers data available. A hacker got in and dumped all their data and redundancies and they didn't have a backup. They are working their hardest to get their customers data back but will cease to exist afterwards. You can't plan for everything. Hope this helps.
Click to expand...



Thank you

I understand the basic raid concepts. It was the URE that threw me off everything as I had been unaware of it before. I was about to create a 4 disk R5 with one hot spare (4 x 4TB) until I read the dangers about it
 
bigdogchris said:
If you have 10k or 15k drives (Fast Class), RAID 5 is fine. But if you're using slower 7,200 or 5,400 drives (Nearline), you should go RAID 6. Why?

Nearline drives are slower and larger, which means they take much longer to rebuild. But they are cheaper. This means RAID 6 is an option to protect against a second failure during rebuild because it's cost effective.

Fast Class drives are faster and smaller, which means they rebuild much faster. But they are more expensive. This means RAID 5 is an option because it saves money, and has less chance of a second failure during rebuild due to the speed.

The other opinion is that anything larger than 2TB should be RAID 6 simply due to rebuild time. You don't find FC drives larger than 2TB.
Click to expand...

Thanks,

I'm basically after a 9+ TB array that I can map to a single VM. I have an ESXi 5.5 host with an LSI 9260-8i raid controller. I would like to create an array that I can use for this VM as a backup location for other computers.

I was going to(due to my lack of understanding) create a R5 with 4x 4TB disks which would have done the trick(on paper) until discovered the URE problem.

My case has 6 disk bays, however I didn't want to use them all as I was aiming for a low power box.

Which leaves me with a couple of scenarios I suppose:

1) Raid 10: 4x 4TB which would give 8TB before formatting
2) Raid 6: 5 x 4TB which would give me 12TB before formatting

Each one would leave me room for a hotspare.

I was going to use standard 4TB desktop drives, as the array would hardly ever do anything. As the drives aren't ever going to be at enterprise level, I guess I should be using R10? As the mirrors are less likely to cause a problem?
 
bigdogchris said:
If you have 10k or 15k drives (Fast Class), RAID 5 is fine. But if you're using slower 7,200 or 5,400 drives (Nearline), you should go RAID 6. Why?
...
The other opinion is that anything larger than 2TB should be RAID 6 simply due to rebuild time. You don't find FC drives larger than 2TB.
Click to expand...

And here I thought FC stood for "Fibre Channel"
 
If your primary concern is URE's then I think what you might be missing is URE's are an issue with the drives, not necessarily the RAID type. The reason why RAID 5 is more susceptible to these issues is if you have a drive failure and 1 of the two remaining drives has issue with it's data due to URE then the data will be corrupt when you replace the drive. So with RAID 10 you can have two drive failures, or 1 failure and 1 corrupt and recover IF they are not on same group. RAID 6 allows for 2 issues of any disk, failure or corruption. For 5 drives RAID 6 should provide the best fault tolerance. Also keep in mind this article was written back in 2010 and was addressing the rapid disk size expansion. The expansion has slowed if not stopped and more reliability and fault tolerance has been built in.

In summary: Stay calm and RAID
 
BillytheSkids said:
If your primary concern is URE's then I think what you might be missing is URE's are an issue with the drives, not necessarily the RAID type. The reason why RAID 5 is more susceptible to these issues is if you have a drive failure and 1 of the two remaining drives has issue with it's data due to URE then the data will be corrupt when you replace the drive. So with RAID 10 you can have two drive failures, or 1 failure and 1 corrupt and recover IF they are not on same group. RAID 6 allows for 2 issues of any disk, failure or corruption. For 5 drives RAID 6 should provide the best fault tolerance. Also keep in mind this article was written back in 2010 and was addressing the rapid disk size expansion. The expansion has slowed if not stopped and more reliability and fault tolerance has been built in.

In summary: Stay calm and RAID
Click to expand...

Cheers :)
 
bigdogchris said:
If you have 10k or 15k drives (Fast Class), RAID 5 is fine. But if you're using slower 7,200 or 5,400 drives (Nearline), you should go RAID 6. Why?

Nearline drives are slower and larger, which means they take much longer to rebuild. But they are cheaper. This means RAID 6 is an option to protect against a second failure during rebuild because it's cost effective.

Fast Class drives are faster and smaller, which means they rebuild much faster. But they are more expensive. This means RAID 5 is an option because it saves money, and has less chance of a second failure during rebuild due to the speed.

The other opinion is that anything larger than 2TB should be RAID 6 simply due to rebuild time. You don't find FC drives larger than 2TB.
Click to expand...

Raid 10, performance, faster rebuilds than raid 6 with the same 2 drive failure rate, pending on which spans it occurs.

Raid 5, i see little reason to use anymore unless you are really broke or your drives as smaller, 500Gig or less and as you said 10k or 15k allowing much faster rebuilds

Any consumer drives or SATA, Raid 10.

Remember Raid is not a backup....backup your data!
 
Raid 10, performance, faster rebuilds than raid 6 with the same 2 drive failure rate no?
Click to expand...

In raid6 any 2 drives can fail and you still can recover. In raid10 if the wrong 2 fail you can have total loss.
 
drescherjm said:
In raid6 any 2 drives can fail and you still can recover. In raid10 if the wrong 2 fail you can have total loss.
Click to expand...

Yep, if two disks in the same mirror fail, then you're screwed.

I think I'll stick with R6, it seems like it should be fine for a small array
 
I have liked RAID10 in the past, but it's expensive, not the best performing, and honestly a dying configuration. Raid 5/6 is superior and RAID 50/60 is even better.
 
bigdogchris said:
I have liked RAID10 in the past, but it's expensive, not the best performing, and honestly a dying configuration. Raid 5/6 is superior and RAID 50/60 is even better.
Click to expand...

I'd love to do raid 60, but I'm trying to keep this box low power.
Dreams nice though haha
 
drescherjm said:
In practise I have (a person with 100 to 200 drives in raid6/raidz2/raidz3) either been very lucky (got to play the lottery) or this is very wrong or maybe a combination of the two.. What I see from my arrays at work is bad drives will have many UREs (a rate much higher than 1 in 12TB). Working drives may experience a URE but not any where near 1 every 12TB.
Click to expand...

Yeah at least outside of RAID my XP is that URE are pretty rare, however RAID cards will drop drives that take a bit long to read relatively quickly, so one URE might be enough for one drive to drop, then if during reconstruction there is another, it's bad. Add to that that the common file systems don't check for data integrity and I stopped my project of going RAID some years ago as soon as I discovered ZFS. ZFS on top of its qualities will not drop a drive for a single URE, and will let it try to read for longer than RAID cards.

Also ZFS offers RAIDZ3 aka "RAID7".
 
Okay, I've decided to go for a 4 disk raid 6 as my requirements have dropped. 8TB will be more than enough.

I was looking at WD red 4TB disks as they're designed to be on 24/7. What I don't like about them is that they can change their RPM rate. How does a raid controller deal with that?

Makes me a little nervous...:confused:
 
You must log in or register to reply here.
Back
Top