Erasure Coding for Data Protection

Each vendor have implemented and derived their own mechanism and algorithm for  erasure coding but the main principle remain same for each vendors. Principle is protecting the failure of drives with help of least additional drives.

Data protection is one of the key aspect while designing any storage product. Fault tolerance had been key part of data protection which means ability of data to recover  at times of drive failures. To achieve fault tolerance data redundancy needs to be done. As for data redundancy there is trade off between fault tolerance and the cost of additional drives. Erasure coding helps to overcome the above mentioned cost overhead of additional drives required.

• Traditional Way (RAID-1)  

Traditional way of redundancy is 1-1 mirroring of drives. Here space efficiency becomes 1/n , where n is number of drives.

 

For protecting 4 drives we need here 4 additional drives. Ratio n:k here becomes 4:4

Where n = given number of drives to be protected

k= additional drives required to protect given number of drives.

 

 

• Erasure Coding Implementation:

 

Before fault

In this approach we are given 4 data symbols  namely A,B,C and D. These data symbols are encoded into two equations

• A+B+C+D
• A+2B+4C+8D

 

After fault

Two data symbol failures. Say B and D.

 

Recovery Mechanism

After two symbol failures( B and D)  we are left with 4 variables and 4 equations. We can easily recover the two symbol failures.( B and D).  

To conclude Erasure coding is helping  to achieve data protection in efficient way with minimum additional drives required to protect given number of  drives.