Erasure encoding protects distributed data without storing multiple complete copies. It splits data into fragments, creates parity fragments with mathematical algorithms, and distributes them across storage nodes. When fragments become unavailable through hardware failure or corruption, the original data can be reconstructed from the remaining fragments and parity. section 5.4.6
The architectural tradeoff is durability and availability with lower storage overhead than full replication. In distributed object storage, the mechanism supports fault tolerance, while self-healing and rebalancing can restore resilience after failures. section 5.4.6 section 5.4.7
Operational implications include:
- Use erasure encoding when data must remain durable across node failures or when centralized replication is unreliable. section 5.5.3
- Plan cluster topology, replication factors, hardware, network bandwidth, and IOPS carefully, especially for Ceph deployments. section 5.4.7
Edge storage
An industrial IoT platform can use MinIO or Ceph to preserve local data integrity across intermittent network connections until synchronization resumes. section 5.5.3
