I don't completely understand RAID

[QuantumBit]

I need some help with RAID

 

I've seen videos and I understand things like RAID 1 and 0, but I don't understand the ones with parody disk.

 

I don't know the name of the RAID, I think its 5, but in that setup lets say you have 4 1 TB disks, A,B,C,D If each disk gets 1/4 taken up for a parody, then a total of 1 TB is lost. What I don't get is that if disk B dies for example, how will it rebuild it? How could you back up 3 TB disk on 1 TB?

[Glenwing]

33 minutes ago, QuantumBit said:

I need some help with RAID

 

I've seen videos and I understand things like RAID 1 and 0, but I don't understand the ones with parody disk.

 

I don't know the name of the RAID, I think its 5, but in that setup lets say you have 4 1 TB disks, A,B,C,D If each disk gets 1/4 taken up for a parody, then a total of 1 TB is lost. What I don't get is that if disk B dies for example, how will it rebuild it? How could you back up 3 TB disk on 1 TB?

Parity contains data on the difference between two data sets. If either set is lost, it can be reconstructed from the remaining set and the parity data. Since it can go both ways, you only need one set of parity data to protect two data sets.

 

Lets say you have two 8 bit disks.

 

The data on the disks is:

A: 00100110
B: 11101100

You have a third disk, C, which contains 8 more bits that represent whether each bit of A and B is the same or different from the other (1 if the same, and 0 if different).

C: 00110101

If disk A fails, you can take the data from disk B and use the parity data to determine what A was. If a bit of C is 0, then the lost bit of A is the opposite of B's bit. If C's bit is 1, then the lost bit of A is the same as B's bit. And the same parity data can be used to reconstruct B if B is lost instead of A. So you only need 8 bits of parity to protect 16 bits of data, so if you had a 24 bits worth of data storage, you can use 2/3 of it for data and 1/3 for parity-based redundancy.

[Cyracus]

it's data voodoo

[QuantumBit]

2 minutes ago, Glenwing said:

Parity contains data on the difference between two data sets. If either set is lost, it can be reconstructed from the remaining set and the parity data. Since it can go both ways, you only need one set of parity data to protect two data sets.

 

Lets say you have two 1 TB disks.

 

The first 8 bits on the each disk are:

A: 00100110
B: 11101100

You have a third disk, C, which contains 8 more bits that represent whether each bit of A and B is the same or different (1 if the same, and 0 if different).

C: 00110101

If disk A fails, you can take the data from disk B and use the parity data to determine what A was. If a bit of C is 0, then the lost bit of A is the opposite of B's bit. If C's bit is 1, then the lost bit of A is the same as B's bit. And the same parity data can be used to reconstruct B if B is lost instead of A. So you only need 8 bits of parity to protect 16 bits of data, so if you had a 24 bits worth of data storage, you can use 2/3 of it for data and 1/3 for parity-based redundancy.

Woah thats a really smart system, also thanks for the great explanation!

[Cereal5]

Linus has a video on TechQuickie abou raid 0,1, and 10, and also 5 and 6 I think. Not at my desktop so I can't link them but look it up

[QuantumBit]

Just now, Cereal5 said:

Linus has a video on TechQuickie abou raid 0,1, and 10, and also 5 and 6 I think. Not at my desktop so I can't link them but look it up

Yea I already watched them but I didn't understand how you could repair a disk that randomly failed

[pwn_intended]

Ill leave that here for you. Not super in depth but it should be a good start.

[Cereal5]

1 minute ago, QuantumBit said:

Yea I already watched them but I didn't understand how you could repair a disk that randomly failed

Oh yeah me neither. Except the part that any raid except 0 can be backed up because there's always another drive that stores that one drives information. Even if the main drive failed all the other ones could be used to piece it together

[QuantumBit]

Just now, Cereal5 said:

Oh yeah me neither. Except the part that any raid except 0 can be backed up because there's always another drive that stores that one drives information. Even if the main drive failed all the other ones could be used to piece it together

Ah ok thank you

[fjdhdhcyfhf]

before you implement RAID; RAID is not a proper back up solution. Its redundancy for uninterrupted service

[pwn_intended]

7 minutes ago, Cereal5 said:

Oh yeah me neither. Except the part that any raid except 0 can be backed up because there's always another drive that stores that one drives information. Even if the main drive failed all the other ones could be used to piece it together

Ah, so lets say you have a 4X1TB drive RAID 5. You will effectively get 3TB total storage, and 1TB is being used for parity. If drive 1 fails, a third of the info from that drive was also stored one each of the other drives as well. Once you put a new drive to replace drive 1, your computer (or RAID controller) will start copying all of the drive 1 parity data from the other 3 drives in order to rebuild the first drive, then it will have to create new parity data for the other 3 drives onto itself.

While an array is re-building, it becomes degraded, and it will be much slower to access, and if a single drive fails during this process, your data is gone bro. Unless of course you had multiple redundancy built into your array.

[leadeater]

50 minutes ago, Cereal5 said:

Oh yeah me neither. Except the part that any raid except 0 can be backed up because there's always another drive that stores that one drives information. Even if the main drive failed all the other ones could be used to piece it together

 

43 minutes ago, SCHISCHKA said:

before you implement RAID; RAID is not a proper back up solution. Its redundancy for uninterrupted service

See second quote.