Introduction to EC

Ceph storage pools are classified into replicated-type pools and erasure-coded pools. This document describes how to optimize the read/write performance of a 4 KB block cluster in EC mode with the EC Turbo feature.

Erasure code (EC) adopts parity code to correct data loss. For example, in EC 4+2 mode, two pieces of parity data x and y are calculated based on four pieces of data a, b, c, and d according to a formula, and the six pieces of data are stored together. If one or two pieces of data among a, b, c, and d are lost, parity data x and y can be used to calculate the lost data through the formula. The EC mode involves data block encoding and decoding. Therefore, the read/write performance of this mode is inferior to that of the replication mode, especially in small-block write scenarios.