Abstract: A multi-processor system includes a partition including a selected number of nodes selected from a plurality of nodes provided in a plurality of node groups, each of the nodes including a computer. A failed node in the partition notifies a failure to a corresponding service processor of the node group and other nodes of the partition. The corresponding service processor and the service processors managing the other nodes notify the error log information to a service processor manager, which identifies the location of the failure and indicate the service processors to recover from the failure.

Abstract: A multi-processor system includes a partition including a selected number of nodes selected from a plurality of nodes provided in a plurality of node groups, each of the nodes including a computer. A failed node in the partition notifies a failure to a corresponding service processor of the node group and other nodes of the partition. The corresponding service processor and the service processors managing the other nodes notify the error log information to a service processor manager, which identifies the location of the failure and indicate the service processors to recover from the failure.

Abstract: A loop network without a standby loop is provided, wherein switch devices in plural stages are connected through multiple loops. The network has at least a pair of loops each of which transfers data in the opposite directions. Under normal conditions, the respective loops transfer data separately. On the other hand, when an error occurs in either loop in a switch device, a switch device located immediate upstream of the faulty section executes loop back operation to build one end of a C-loop. Moreover, the switch device in which the error occurs in the loop also executes loop back operation after outputting data in process to build the other end of the C-loop. By this way, the C-loop circumventing the faulty point can be established. Thereby, it becomes possible to transfer data through the C-loop with a guarantee of ordering and to repair, replace and/or maintain the faulty point without any influenced on the network.