Erasure coding in CORTX

[mshaio]

Hi,

I'm wondering if there is somewhere I can find out more details about the erasure coding being used in CORTX. For example:

1. What's the upper limit of failures it's able to handle at any instant?
2. What erasure coding algorithm does CORTX use?

Thank you.

[mukul-seagate11]

@madhavemuri @huanghua78, can you address the queries asked?

[johnbent]

Hello @mshaio thanks for your question! I don't know the answer to the upper limit but I know it can do N+K+S and N can be 1 or more and I know we are using K=2 and S=2. Here is the code: https://github.com/Seagate/cortx-motr-galois

[mssawant]

@mshaio, Just extending John's answer,

What's the upper limit of failures it's able to handle at any instant?

This depends on your configuration of the cluster. Motr uses parity declustered layout, also checkout Reed solomon algorithm. Fault tolerance is equivalent to the supported redundancy. In this case which is denoted by the value of K (no .of parity units). This can be extended to various failure domains based on your cluster's pool size.

What erasure coding algorithm does CORTX use?

Presently CORTX-motr uses Parity declustered layout.