Help on raid 5

[the darronator]

Can someone explain to me what raid 5 actually does?
I know how it is redundant, but how does it do this and how the bloody hell does it work?

[tomhumphrey]

[the darronator]

Cheers mate

[TonyStark]

Also, RAID five needs 3 disks or more to be in effect. in a 3 disk RAID 5 array you can have up to 1 drive failure....for each extra disk in a RAID 5 array you add one extra drive that can fail to the equation.

[the darronator]

18 minutes ago, TonyStark said:

Also, RAID five needs 3 disks or more to be in effect. in a 3 disk RAID 5 array you can have up to 1 drive failure....for each extra disk in a RAID 5 array you add one extra drive that can fail to the equation.

Thanks a lot

[TonyStark]

Don't mention it....

[beavo451]

3 hours ago, TonyStark said:

Also, RAID five needs 3 disks or more to be in effect. in a 3 disk RAID 5 array you can have up to 1 drive failure....for each extra disk in a RAID 5 array you add one extra drive that can fail to the equation.

This is wrong. A RAID 5 requires a minimum of 3 disks to create and can only withstand 1 drive failure regardless of the number of disks in the array.

[Ramaddil]

11 minutes ago, beavo451 said:

This is wrong. A RAID 5 requires a minimum of 3 disks to create and can only withstand 1 drive failure regardless of the number of disks in the array.

What Tony Stark was trying to say, was that yes it does require a minimum of three disks in order to create a Raid 5 Array.  But with each additional drive you add you increase the chances array failure, just the same of any other raid set.  A Raid 5 array can only withstand one failure and the chance of a second failure is high (during a rebuild) relative to the number of drives in the array.  In other words if practical either go with a Raid 6 or Raid 10, they are more resilient to drive failures during the rebuild process. 

[spellmanuk]

 

 

[KuJoe]

26 minutes ago, Ramaddil said:

What Tony Stark was trying to say, was that yes it does require a minimum of three disks in order to create a Raid 5 Array.  But with each additional drive you add you increase the chances array failure, just the same of any other raid set.  A Raid 5 array can only withstand one failure and the chance of a second failure is high (during a rebuild) relative to the number of drives in the array.  In other words if practical either go with a Raid 6 or Raid 10, they are more resilient to drive failures during the rebuild process. 

I misread what @TonyStark said also, I thought he meant adding 1 drive would give 1 more drive redundancy but after reading it again with my glasses on it makes sense.

 

RAID10 is different though because it really boils down to luck, if you have 50 drives in a RAID10 then 25 of them can fail if they are the "right" 25 drives. On the other hand if the "wrong" 2 drives fail then data is lost. Parity based RAID (5, 50, 6, 60) have better odds of surviving multiple drive failures than RAID10 but lack the performance of RAID10 (both in use and rebuild). Another thing to note is that if you don't have a hardware RAID controller that handles the parity calculations it'll offload to the CPU which isn't a huge issue if you have multiple CPUs with multiple cores each, but something to consider for a home NAS with a consumer CPU if you're going to be doing CPU heavy stuff.

 

Lastly, I only run RAID5 and 50 on a few production servers, I run RAID10 on everything else (including my personal servers).

 

EDIT: RAID is NOT a form of backup and proper backups should be taken regardless of what RAID you use.

[beavo451]

53 minutes ago, KuJoe said:

I misread what @TonyStark said also, I thought he meant adding 1 drive would give 1 more drive redundancy but after reading it again with my glasses on it makes sense.

 

RAID10 is different though because it really boils down to luck, if you have 50 drives in a RAID10 then 25 of them can fail if they are the "right" 25 drives. On the other hand if the "wrong" 2 drives fail then data is lost. Parity based RAID (5, 50, 6, 60) have better odds of surviving multiple drive failures than RAID10 but lack the performance of RAID10 (both in use and rebuild). Another thing to note is that if you don't have a hardware RAID controller that handles the parity calculations it'll offload to the CPU which isn't a huge issue if you have multiple CPUs with multiple cores each, but something to consider for a home NAS with a consumer CPU if you're going to be doing CPU heavy stuff.

 

Lastly, I only run RAID5 and 50 on a few production servers, I run RAID10 on everything else (including my personal servers).

 

EDIT: RAID is NOT a form of backup and proper backups should be taken regardless of what RAID you use.

Going back and reading it, it can be read either way.

[Ramaddil]

1 hour ago, KuJoe said:

I misread what @TonyStark said also, I thought he meant adding 1 drive would give 1 more drive redundancy but after reading it again with my glasses on it makes sense.

 

RAID10 is different though because it really boils down to luck, if you have 50 drives in a RAID10 then 25 of them can fail if they are the "right" 25 drives. On the other hand if the "wrong" 2 drives fail then data is lost. Parity based RAID (5, 50, 6, 60) have better odds of surviving multiple drive failures than RAID10 but lack the performance of RAID10 (both in use and rebuild). Another thing to note is that if you don't have a hardware RAID controller that handles the parity calculations it'll offload to the CPU which isn't a huge issue if you have multiple CPUs with multiple cores each, but something to consider for a home NAS with a consumer CPU if you're going to be doing CPU heavy stuff.

 

Lastly, I only run RAID5 and 50 on a few production servers, I run RAID10 on everything else (including my personal servers).

 

EDIT: RAID is NOT a form of backup and proper backups should be taken regardless of what RAID you use.

I agree Kujoe, I would never run Raid 5. Unless it is a backup of another Array, and even then I probably wouldn't. I Have 24 (2TB) Drives in a Raid 6 with a hot spare. It is currently backed up to 4 (6TB drives in a JBOD) until I can afford a few more 6TB drives.  And a copy of all my data is saved in the cloud as well (<---- Paranoid about my data)

[the darronator]

17 hours ago, KuJoe said:

I misread what @TonyStark said also, I thought he meant adding 1 drive would give 1 more drive redundancy but after reading it again with my glasses on it makes sense.

 

RAID10 is different though because it really boils down to luck, if you have 50 drives in a RAID10 then 25 of them can fail if they are the "right" 25 drives. On the other hand if the "wrong" 2 drives fail then data is lost. Parity based RAID (5, 50, 6, 60) have better odds of surviving multiple drive failures than RAID10 but lack the performance of RAID10 (both in use and rebuild). Another thing to note is that if you don't have a hardware RAID controller that handles the parity calculations it'll offload to the CPU which isn't a huge issue if you have multiple CPUs with multiple cores each, but something to consider for a home NAS with a consumer CPU if you're going to be doing CPU heavy stuff.

 

Lastly, I only run RAID5 and 50 on a few production servers, I run RAID10 on everything else (including my personal servers).

 

EDIT: RAID is NOT a form of backup and proper backups should be taken regardless of what RAID you use.

Yeah, knew that. I already have a NAS I made, but its really slow... I was planning on making the raid 5 nas my main one, and nightly backup into my current NAS.

Everything in my house is backed up at least 1nce.

[TonyStark]

On 2/21/2016 at 1:29 AM, beavo451 said:

This is wrong. A RAID 5 requires a minimum of 3 disks to create and can only withstand 1 drive failure regardless of the number of disks in the array.

I think, if I'm not mistaken that the parity was distributed to so many parts as the disks in the array, that is  why it could withstand more than 1 drive failure if the array drives were over 3...

[beavo451]

4 hours ago, TonyStark said:

I think, if I'm not mistaken that the parity was distributed to so many parts as the disks in the array, that is  why it could withstand more than 1 drive failure if the array drives were over 3...

RAID5 places the parity block on one drive and then distributes the data block evenly across the remaining drives in the array (striping). There Is not a dedicated drive for parity blocks. So, say you have 8 drives. The parity block would be written to one drive and the remaining data blocks are striped across the remaining 7 disks. The parity block is written to a different drive for each set of data.

 

RAID6 does the same thing except that it calculates 2 parity blocks and thus is able to withstand 2 drive failures regardless of the total number of drives.

 

If you were to stripe the parity block across multiple drives, it would not add any redundancy because if you loss any part of the parity block, then you loose the whole block.

 

So, taking the same 8 drive array, if you were to stripe the parity block across 4 drives and the data is striped across the remaining 4 drives, then yes, theoreticallyft you could sustain up to 4 drive failures IF all 4 drives were the drives holding the parity block. BUT, since the parity blocks are distributed amongst all the disks, a disk failure would take out data blocks as well as parity blocks. So such a scheme could still only withstand a drive failure. There is no point in creating such a configuration.