Peer Fusion Clusters Are Simple Operate.


A Peer Fusion cluster functions mostly through peer quorums:

  • A Peer Fusion cluster offers simpler operations, resiliency and scalability.
    • The cluster owns the data and the namespace.
    • In a Peer Fusion cluster the namespace is replicated across all peers and the data is striped across the cluster and onto every peer equally. Each peer has a very simple configuration file used mainly to identify the resources to use (i.e. NICs, volume, and cluster identification).

      Peers use the configuration information for network discovery to join a cluster. Peers are plug-and-play as new peers join the cluster automatically through network discovery and are immediately able to contribute to the cluster.

    • The gateways export file system semantics.
    • The cluster gateway nodes have no persistent content besides a very simple configuration file used mainly to specify the FEC, and the cluster identification. They have no persistent knowledge of the peers and must also perform network discovery to join a cluster. They rely on the cluster for namespace lookup and traversal. While gateways are necessary to access the cluster, they are expendable with no loss of cluster data.

    • Cluster healing and restriping is simple.
    • Healing is the process of refreshing stale data on failed peers once they re-join the cluster. Restriping is the process of modifying the stripe layout of a cluster when new peers are added. Healing and restriping can be configured for automatic or manual initiation. A priority can be established by specifying any combination of files and directories.