Core Concepts¶

Erasure Coding¶

For more complete documentation on MinIO Erasure Coding, see Erasure Coding. This section summarizes erasure coding in context of MinIO tenants on VCF.

MinIO protects data with per-object inline erasure coding written in assembly code to deliver the highest possible performance. MinIO uses Reed-Solomon code to stripe objects into data and parity blocks with user-configurable redundancy levels. MinIO’s erasure coding performs healing at the object level and can heal multiple objects independently.

At the maximum parity of N/2, where N is the total number of volumes in a MinIO tenant zone, MinIO’s implementation can ensure uninterrupted read and write operations with only ((N/2)+1) operational volumes in the deployment. For example, in a 16-drive setup, MinIO shards objects across 8 data and 8 parity drives and can reliably write new objects or reconstruct existing objects with only 9 drives remaining in the deployment.

MinIO Erasure Coding is supported by Erasure Sets. MinIO automatically calculates and creates Erasure Sets in each zone. MinIO then splits objects on write into data and parity blocks, distributing those blocks among drives in the erasure set. You can view the erasure set distribution when Deploying a Tenant or Expanding a Tenant.

A zone may have multiple erasure sets, all of the same calculated size. MinIO uses the EC:N notation to define the ratio of parity blocks to data blocks for objects stored in a given erasure set. The number of parity blocks determines the Tenant’s ability to continue servicing read and write requests in the event of drive or pod failure. Specifically, for M = Erasure Set Drives:

• For write operations (write quorum), the MinIO tenant requires at least (M-EC:N)+1 drives.

• For read operations (read quorum), the MinIO tenant requires at least (M-EC:N) drives.

A MinIO tenant with enough erasure set drives to satisfy write quorum implicitly has enough drives to satisfy read quorum.

MinIO tenants default Erasure Coding parity to 1/2 of the drives in the Erasure Set(s) for the initial zone. For example, a zone with 16-drive Erasure Sets defaults to EC:8, while a zone with 10-drive Erasure Sets defaults to EC:5.

Since parity blocks require storage space, higher levels of parity provide increased tolerance to drive or pod failure at the cost of total usable storage capacity. You can specify your preferred erasure code parity setting when Deploying a Tenant.

The following table lists the outcome of varying EC levels on a MinIO Tenant with 4 pods and 4 1Ti drives per node. MinIO creates a single Erasure Set consisting of 16 drives for this tenant.