Provided is an image recognition processor. The image recognition processor includes a plurality of nano cores arranged in rows and columns and configured to perform a pattern recognition operation on an input feature using a kernel coefficient in response to each instruction, an instruction memory configured to provide the instruction to each of the plurality of nano cores, a feature memory configured to provide the input feature to each of the plurality of nano cores, a kernel memory configured to provide the kernel coefficients to the plurality of nano cores, and a functional safety processor core configured to receive a result of a pattern recognition operation outputted from the plurality of nano cores to detect the presence of a recognition error, and perform a fault tolerance function on the detected recognition error.

A method of operating a storage device may include determining a fault condition of the storage device, selecting a fault resilient mode based on the fault condition of the storage device, and operating the storage device in the selected fault resilient mode. The selected fault resilient mode may include one of a power cycle mode, a reformat mode, a reduced capacity read-only mode, a reduced capacity mode, a reduced performance mode, a read-only mode, a partial read-only mode, a temporary read-only mode, a temporary partial read-only mode, or a vulnerable mode. The storage device may be configured to perform a namespace capacity management command received from the host. The namespace capacity management command may include a resize subcommand and/or a zero-size namespace subcommand. The storage device may report the selected fault resilient mode to a host.

Techniques are provided for neutralizing replication errors. An operation is executed upon a first storage object and is replicated as a replicated operation for execution upon a second storage object. A first error may be received for the replicated operation. Instead of transitioning to an out of sync state and aborting the operation, a wait is performed until a result of the attempted execution of the operation is received. If the first error is the same as a second error returned for the operation, then the operation and replicated operation are considered successful and a synchronous replication relationship is kept in sync. If the first error and the second error are different errors, then an error response is returned for the operation and the synchronous replication relationship is transitioned to out of sync.

A processing device is provided. The processing device comprises an interface configured to receive information about an operation state of a surrogate processor. Further, the processing device comprises a processing circuitry configured to decide whether an interrupt addressed to the processing circuitry is processed by the processing circuitry or redirected to the surrogate processing circuitry based on an operation state of the processing circuitry and the surrogate processing circuitry.

A method for reliable data synchronization within a network is disclosed. The producer system stories data in a persistent data store and produces one or more data updates. The producer system simultaneously transmits the data updates to a consumer system and initiating storage of the data updates at the producer system. When storage of the data updates at the producer system is complete, the producer system transmits a first acknowledgment to the consumer system. The producer system determines whether a second acknowledgment has been received from the consumer system, wherein the second acknowledgment indicates that the consumer system has successfully stored the data updates at the consumer system. In accordance with a determination that the second acknowledgment has been received from the consumer system, the producer system changes the temporary status of the data updates stored at the producer system to a permanent status.

In part, disclosure relates to a method of regulating checkpointing in an active active fault tolerant system. The method includes receiving a request from a client through a network at a primary computer; copying, by the primary computer, the request from the client to a secondary computer; processing the request from the client, using the primary computer, to generate a primary computer result; processing the copy of the request from the client, using the secondary computer, to generate a secondary computer result; comparing the primary computer result and the secondary computer result to obtain a comparison metric; determining whether a minimum checkpoint interval has been met or exceeded; and if the minimum checkpoint interval has not been met or exceeded, delay initiating a checkpoint process from primary computer to secondary computer.

A method and a system are provided for data management in a transport device, in particular in a train. In a first comparison, a first count value stored in a first control device is compared with a count value stored in a second control device. In a second comparison, a count value selected from the first and second count values on the basis of a result of the first comparison is compared with a control count value stored in a safety device. On the basis of a result of the second comparison, control data stored in the safety device and associated with the control count value are acquired by the first or second control device.

Systems and methods are disclosed for checker cores for fault tolerant processing. For example, an integrated circuit (e.g., a processor) for executing instructions includes a processor core configured to execute instructions of an instruction set; an outer memory system configured to store instructions and data; and a checker core configured to receive committed instruction packets from the processor core and check the committed instruction packets for errors, wherein the checker core is configured to utilize a memory pathway of the processor core to access the outer memory system by receiving instructions and data read from the outer memory system as portions of committed instruction packets from the processor core. For example, data flow from the processor core to the checker core may be limited to committed instruction packets received via dedicated a wire bundle.

The present disclosure describes example service takeover methods, storage devices, and service takeover apparatuses. In one example method, when a communication fault occurs between two storage devices in a storage system, the two storage devices respectively obtain running statuses of the two storage devices. A running status can reflect current usage of one or more system resources of a particular storage device. Then, a delay duration is determined according to the running statuses, where the delay duration is a duration for which the storage device waits before sending an arbitration request to a quorum server. The two storage devices respectively send, after the delay duration, arbitration requests to the quorum server to request to take over a service. The quorum server then can select a storage device in a relatively better running status to take over a host service.

Internet-of-Things (“IoT”) controllers built using hardened industrial technologies which improve functionality and reliability, such as a fixed-loop model in which a loop is repeated with configured time periodicity where sensors are queried, sensor responses are read, configured calculations are performed, and logic rules are evaluated resulting in decisions made and outputs activated. A variety of redundancy techniques are utilized to provide continuous non-stop operation of IoT controllers to compensate for possible hardware and software failures. Robust IoT controller redundancy also allows periodic maintenance, software updates and security patch installation without shutting down the IoT controllers.