摘要 |
PROBLEM TO BE SOLVED: To quickly continue processing without losing data stored in a main storage or a cache at the occurrence of a fault in a cluster computer system having four or more nodes. SOLUTION: The cluster computer system is constituted of four or more nodes. The system is provided with an inter-node network of a first type for coupling all the nodes and an inter-node network of a second type for coupling every two nodes, and the inter-node network of the second type is used exclusively for information communication which causes processing being performed in a faulty node to be handed over to the other node coupled with the faulty node by the inter-node network of the second type when a fault occurs. When a fault occurs, processing can be taken over to a destination node without newly retrieving the takeover destination node. Write data in the main storage is allowed to be copied during normal operation, so that data on the main storage can be taken over to detailed takeover processing when the fault occurs. COPYRIGHT: (C)2004,JPO&NCIPI
|