[Knowledge Hall] [Storage] Concept of RAID

RAID can contains phyiscal arrays/drives and logical arrays/drives. However, this doesn't matter as the concept will still be same.
Drives are collected in one group and RAID level determines how the data is split amongs this group of drives.
There are 3 core concepts of RAID must be understand, which all the other RAID technology also derived from here.

1. Mirror - Mirror is a method to write the data simultaneously to two or more drives. This is to obtain 100% data redundancy. Either disk
is containing the duplicated data which protect against the failure of disks.
2. Stripe - The data is broken into various block-sized/byte-sized pieces and then written distributely to two or more drives. This is to obtain
greater I/O performance but no protection against the failure of disks.
3. Parity - Parity is to take N pieces of data and use mathematic algorithm to calculate one extra piece of data. This N+1 pieces of data will store
in N+1 drives. Regardless of which piece of data is lost, it can always be recovered from the remaining N pieces. Hence it is to protect against the 1 disk
failure.

1 Minor concept is Concatenation. The data will written sequencely to the next concat drive when the current disk had been written full. Therefore, this technic doesn't have
protection and performance increase but only the capacity increase.

Some questions.
1. How is disk failure tolerance amongs Mirror, Stripe, and Parity?
2. How is the read performance amongs Mirror, Stripe, and Parity?
3. How is the write performance amongs Mirror, Stripe, and Parity?
4. What could be the possible reason why customer want to use Mirror, Stripe or Partiy?

No comments:

Related Posts with Thumbnails

Saving The Seahorse Means Saving The Sea