Nodes and drive enclosures are interconnected into a clique configuration for fault tolerance. One or more nodes in a clique can be configured as an HSN. This eliminates the degradation of database performance in the event of a node failure in the clique. Tasks assigned to the failed node are completely redirected to the HSN.
Each Teradata processing node and HSN has redundant connections to the disk arrays. All links are point to point.
Total number of HDD arrays, SSD arrays, nodes, and HSNs per clique is variable and is configuration-dependent. Combinations within a clique must not exceed the port count of the first-level InfiniBand switches, including at least one uplink port available per switch for every three nodes.