Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for three-dimensionally stacked neural network accelerators. In one aspect, a method includes obtaining data specifying that a tile from a plurality of tiles in a three-dimensionally stacked neural network accelerator is a faulty tile. The three-dimensionally stacked neural network accelerator includes a plurality of neural network dies, each neural network die including a respective plurality of tiles, each tile has input and output connections. The three-dimensionally stacked neural network accelerator is configured to process inputs by routing the input through each of the plurality of tiles according to a dataflow configuration and modifying the dataflow configuration to route an output of a tile before the faulty tile in the dataflow configuration to an input connection of a tile that is positioned above or below the faulty tile on a different neural network die than the faulty tile.

A communications network and method for operating the communications network that includes ring devices that are networked to one another in a ring topology, where the ring devices participate in a ring redundancy process in which an administrating ring device regularly sends test packets over the communications network, which received in sequence by other ring devices and successively forwarded by these devices back to the administrating ring device so as to detect faults in the communications network, where a transmission delay of a test packet along the ring topology because of interfering packets within the ring topology is prevented by synchronizing the ring devices and processing at least parts of a data traffic schedule in accordance with the 802.1Qbv standard in the ring redundancy process applied, and where transmission of the test packets is scheduled and controlled such that the packets are forwarded to the ring devices substantially without delay.

An electronic device includes multiple networking devices arranged in a stack. The networking devices may include configurable ports, where a given configurable port in the configurable ports may be configured as a data port or a stacking port. During operation, a networking device in the stack may be designated as a master in the stack. In response, the networking device may provide one or more probe messages to determine a state of the networking devices, where the state includes one or more connections among the networking devices. Then, the networking device may verify that the one or more connections are correct. When the one or more connections are correct, the networking device may define a subset of the configurable ports in the networking devices as stacking ports.

An information obtaining unit (53) of a monitoring apparatus (40) obtains from each node, topology map information created in each node by an exchange of information and sharing of information between nodes that belong to a ring network. The topology map information is information that indicates a connection relation between the nodes in the ring network. A topology comparison unit (54) of the monitoring apparatus (40) compares the topology map information obtained from each node by the information obtaining unit (53) with topology definition information retained in a memory (42) beforehand, and determines whether or not a configuration of the ring network is according to design. The topology definition information is information that defines the connection relation between the nodes in the ring network.

Embodiments are directed to a method of implementing a packet capture ring. The packet capture ring includes a plurality of appliances, and the plurality of appliances includes a first appliance and a second appliance. The first appliance and the second appliance are both attached to a network tap, and the first appliance works as a master appliance. The master appliance ingests packets from the network tap, encapsulates the packets and forwards encapsulated packets in the packet capture ring. The method includes: detecting, by the second appliance, a failure of the first appliance; working, by the second appliance, as the master appliance; and removing, by the second appliance, the first appliance from a forwarding designation list.

The invention relates to a vehicle having fail-safe internal data transfer. The vehicle comprises a vehicle body and a wired data transfer network provided on the vehicle body. Furthermore, network subscribers are provided on the vehicle body, which network subscribers are connected to each other via respective network nodes of the data transfer network. The data transfer network has a data-transferring ring wiring.

Systems and methods for node isolation detection include, in one or more hub nodes in an Ethernet network having a first ring and a second ring that subtends off the one or more hub nodes on the first ring, operating a first plurality of Operations, Administration, and Maintenance (OAM) sessions with each node in the second ring in a first direction around the second ring; operating a second plurality of OAM sessions with each node in the second ring in a second direction around the second ring; and correlating and detecting faults and node isolation in the second ring based on the first plurality of OAM sessions and the second plurality of OAM sessions.

An example communications device may include communication ports and processing circuitry. The communications device may, when the communication device and peer network nodes are connected in a ring topology to form an access network, detect that a loop exists between the access network and a customer network. In response to detecting the loop, the communication device may automatically block any uplink to the customer network that the communication device may have, and send a message to all of the peer network nodes instructing them to block any uplinks to the customer network they may have.

A method and a device for each network element node in a transoceanic multiplex section shared protection ring to automatically discover a cross-node service topology. According to the present invention, the method for automatically discovering a cross-node service topology comprises the following steps: at the node, generating a message containing service identification information of the node, the service identification information being used for identifying a service flow configured for a protection group; sending the message containing the service identification information through the transoceanic multiplex section shared protection ring; and at at least one of other nodes, determining a cross-node service crossing topology according to the message containing the service identification information. According to the embodiment of the present invention, by adopting a data communication channel to periodically send a broadcast message and interacting with a configuration module and a protection protocol module in real time, when a fault occurs in a ring.

Systems and methods for node isolation detection include, in one or more hub nodes in an Ethernet network having a first ring and a second ring that subtends off the one or more hub nodes on the first ring, operating a first plurality of Operations, Administration, and Maintenance (OAM) sessions with each node in the second ring in a first direction around the second ring; operating a second plurality of OAM sessions with each node in the second ring in a second direction around the second ring; and correlating and detecting faults and node isolation in the second ring based on the first plurality of OAM sessions and the second plurality of OAM sessions.