Peer Fusion Cluster Data Resiliency
Peer Fusion clusters interweave resiliency across all the peers:
The repair checksum blocks in a Peer Fusion cluster are striped across the cluster and onto every peer equally. Every peer computes and stores checksum blocks as required to support the resiliency level configured. Every peer has an equal opportunity to assist in performing repairs. Every peer has unique data and no data is replicated. No peer is more important than any other peer in the cluster and all peers are expandable up to the FEC configured.
A Peer Fusion cluster heals rapidly. All the available peers contribute equally to the repairs. The repairs are driven by the cluster-wide namespace so the repair process walks the files and only the HDD blocks containing data are repaired. This is fast. Multiple peers can be healed simultaneously with no measurable additional overhead. This is faster. The Peer Fusion codec is able to encode/decode multiple gigabytes per second per CPU core. This is fastest.
There is no fail-over in a Peer Fusion cluster when a peer fails. When peer failures occur they are detected by the cluster and the remaining peers are instantaneously setup to perform on-the-fly repairs. All I/O transactions are successfully completed and the applications using the cluster are not disrupted as they are never aware that failures have occurred.