Abstract: A performance adjusting method and apparatus for a file system architecture. The method includes: determining a plurality of internet bandwidths among a client, an Object Storage Server (OSS), and a storage controller in a file system architecture; determining an overall network architecture of the file system architecture and a maximum transmission bandwidth of the file system architecture; testing read-write performance of a single Object Storage Target (OST) to calculate an Input/Output (IO) processing capability of a storage node; determining whether the maximum transmission bandwidth matches the IO processing capability of the storage node to add or reduce a storage array accordingly to adjust a number of OSTs; executing a performance benchmark test to determine IO read-write performance of the file system architecture; and determining whether the maximum transmission bandwidth matches the IO read-write performance to adjust a number of OSSs or a number of storage arrays accordingly.

Abstract: Disclosed is a method for resource exposure in kubernetes, including: in response to receiving a request for applying for a service resource, judging whether a mapping relationship between a label corresponding to a type of the service resource and virtual IPs exists in a first configuration file of a keepalived service; in response to absence of the mapping relationship, acquiring new virtual IPs and establishing, in the first configuration file, a mapping relationship between the new virtual IPs and the label; and binding the virtual IPs corresponding to the label in the first configuration file to a plurality of pods corresponding to the service resource, so as to expose the plurality of the pods outside a kubernetes cluster by the virtual IP. Further disclosed are a system, a computer device and a readable storage medium.

Abstract: A random write method includes: using a wear-leveling module to scan the number of free blocks and the number of bad blocks in a target super logic unit; using a lookup management module to iteratively update, according to the number of current remaining solid-state disk data frames, the number of historically weighted solid-state disk data frames in a long short-term memory network manner; using dynamic write arbitration to determine an adjustment stage based on the number of historically weighted solid-state disk data frames, and determining the expected number of read and write operations per second based on the adjustment stage; and re-updating the number of historically weighted solid-state disk data frames, and adjusting the actual number of read and write operations per second based on the re-updated number of historically weighted solid-state disk data frames and the expected number of read and write operations per second.

Abstract: Disclosed are a data processing method and apparatus based on an optical neural network, a computer-readable storage medium, and an optical neural network.

Abstract: An nGraph-based graphics processing unit (GPU) backend distributed training method and system, a computer-readable storage medium, and an electronic device. The method includes: receiving a training request, and obtaining corresponding training data; obtaining a Nvidia® Collective multi-GPU Communication Library (NCCL) file by means of a system path of the NCCL file linked to an nGraph framework; invoking an NCCL communication interface configuration according to the training request to obtain a training model, the NCCL communication interface is an NCCL file-based communication operation interface located at a GPU backend of the nGraph framework; and performing GPU backend training on the training data using the training model. The present application can satisfy an urgent need of a user for performing neural network distributed training on the basis of an nGraph GPU backend, thus further improving the performance of deep learning network training.

Abstract: Provided are a bus exception handling method and apparatus, an electronic device and a computer-readable storage medium. The method includes: respectively obtaining, from multiple target buses, multiple pieces of target data corresponding to the multiple target buses, where the multiple target buses include a master bus and one or more candidate buses, and the target data corresponding to the master bus is referred to as first data; determining whether the first data satisfies a bus exception condition, where the bus exception condition is a data bus marker exception condition or a data content exception condition; and in response to determining that the first data satisfies the bus exception condition, selecting a target candidate bus in a healthy state as a new master bus, and updating local bus data.

Abstract: A potential difference early-warning circuit, comprising: a sensing resistor (R2), (PCB GND); a first MOS (M1), a drain of the first MOS being connected to the sensing resistor (R2), and a source of the first MOS (M1) being connected to a safety ground (GND); an operational amplifier (U1A), a positive input end of the operational amplifier being connected to one end of the sensing resistor (R2), and a negative input end of the operational amplifier (U1A) being connected to the other end of the sensing resistor (R2); a second MOS (M2), a gate of the second MOS (M2) being connected to an output end of the operational amplifier (U1A), and a source of the second MOS (M2) being connected to the signal ground; and a controller, a first input end of the controller being connected to the drain of the second MOS (M2), and an output end of the controller being connected to the gate of the second MOS (M2).

Abstract: An inference service management method, apparatus, and system for an inference platform, and a medium, the method comprising: detecting whether there is inference service corresponding to the inference service record in a server according to an inference service record in a database (S110); and if not, then restoring the corresponding inference service according to the inference service record (S120). According to the method, an inference service in a server is detected according to an inference service record in a database to determine whether there is inference service corresponding to the inference service record in the server, if not, it means that the inference service record is inconsistent with a real inference service, and then corresponding inference service may be restored according to the inference service record.

Abstract: The present application provides a bridge supply circuit of a multipath Efuse chip and a server. The supply circuit includes: a first Efuse chip and multiple second Efuse chips, inputs of the first Efuse chip and the multiple second Efuse chips being connected to a power supply, an output of the first Efuse chip being connected to each of the Graphics processing units (GPUs), an output of each of the multiple second Efuse chips being separately connected to each of the GPUs except for a first GPU; damping circuits, each of the damping circuits being separately provided between the output of each of the multiple second Efuse chips and the GPU; and current paths, each of the current paths being separately provided between the output of the first Efuse chip and each of the GPUs except for the first GPU.

Abstract: Provided are a file processing method, apparatus and device, and a readable storage medium. The method includes: judging whether a received file access path exists; if so, judging whether the file access path corresponds to a folder; if so, generating a folder stream, traversing the folder stream, and generating a file object corresponding to the folder stream; storing the file object in a memory, acquiring the size of a corresponding file according to the file object, detecting whether there is a file in the folder, and a size of the file is known and the file has been traversed and used, and if so, deleting, from the memory, the file object corresponding to the file; and acquiring the total size of files in the folder according to the size of each file in the folder.

Abstract: An IIC hang link restoration circuit based on a PCA9511A chip, including: a PCA9511A chip, a negation circuit, and an external MOS tube (Q1). An IC bus inside the PCA9511A chip includes a clock line; the PCA9511A chip has one end connected to a master of the IIC bus and the other end connected to a slave of the IIC bus; the PCA9511A chip is provided with input and output interfaces of the clock line, and a ready signal interface; the negation circuit has one end connected to the ready signal interface and the other end connected to a gate electrode of the external MOS tube (Q1), and a source electrode and a drain electrode of the external MOS tube (Q1) are respectively connected to the input interface and the output interface of the clock line of the PCA9511A chip.

Abstract: A method for managing a shared memory, including: setting a master core, applying, by the master core, to a system for idle hugepage blocks, mapping the idle hugepage blocks to a virtual address space of the master core, and sorting the hugepage blocks in an ascending order of physical address size of the hugepage blocks; dividing, depending on whether the physical addresses are continuous, the hugepage blocks sorted into segments; in response to determining that there are segments satisfying the memory demand of the subsystem, sorting the segments satisfying the memory demand of the subsystem in a descending order of lengths thereof, and remapping the sorted segments to obtain segments having both continuous virtual addresses and continuous real addresses; and in response to determine that there is a segment satisfying the memory demand of the system, releasing hugepage blocks other than the segment satisfying the memory demand of the system.

Abstract: Provided is a phase self-correction circuit, including a trigger signal operation module and a signal phase correction module. The trigger signal operation module and the signal phase correction module are both composed of a plurality of discrete components. The trigger signal operation module is configured to perform a logical operation on an input phase standard reference signal and actual transmission signal, to obtain a target trigger signal for triggering the signal phase correction module; and the signal phase correction module is configured to output, based on trigger modes of the target trigger signal and the actual transmission signal, a self-correction transmission signal with the same waveform as that of the phase standard reference signal, to realize phase self-correction on the actual transmission signal.

Abstract: A memory anomaly processing method and system, an electronic device, and a storage medium. The method includes: reading a memory error quantity of a target memory bank from a memory error register; when the memory error quantity is greater than a preset value, executing a hot-removal operation on the target memory bank; calculating a memory delay parameter, and writing the memory delay parameter into a memory controller, wherein the memory delay parameter is waiting time after the memory controller controls the target memory bank to receive a read/write command; and executing a hot-addition operation on the target memory bank, whereby the memory controller continues to execute a read/write operation on the target memory bank based on the memory delay parameter. It can be seen that, according to the present application, the memory read/write error rate may be reduced.

Abstract: A method and apparatus for improving message processing efficiency of a Flash channel controller are provided. The method includes: S1, after receiving a request message of a functional unit, a Flash interface parses the request message, and constructs a request response message according to a parsing result, wherein the request response message includes a state of the request message; S2, the Flash interface returns the request response message to the functional unit; and S3, the functional unit acquires the state of the request message according to the request response message, and makes, according to whether the state of the request message is normal, a response to the request message before receiving a completion message.

Abstract: A Logical Unit Number (LUN) division method and device includes checking and adjusting a connection manner and numbers of Serial Attached Small Computer System Interface (SCSI) (SAS) connections of storage enclosures and a controller, so as to make a maximum output bandwidth of each storage enclosure consistent (Si); querying controller port identifiers and storage enclosure identifiers (S2); creating a Mdisk array, and adding the corresponding controller port identifier and the corresponding storage enclosure identifier for each Mdisk (S3); logically dividing a storage space in the Mdisk array to create a volume, and dividing the volume into LUNs, whereby Mdisks that form the LUNs are made to come from different storage enclosures and different controller ports (S4); and mapping the LUNs to a host (S5).

Abstract: The present disclosure discloses a network switching method and apparatus, an electronic device, and a computer-read-able storage medium, the method including: creating a smart monitor used for monitoring port changes; if the smart monitor monitors that a smart port for deploying a network is created, performing network deployment of a bare metal server (bms) node by using the smart port, wherein a smart network interface card is installed in the bms node, and the smart network interface card generates a first bare metal port at the bms node and generates a second bare metal port corresponding to the first bare metal port in an operating system of the smart network interface card; if the smart monitor monitors that a neutron port of a neutron network is updated, adding the second bare metal port into a network bridge of the bms node.

Abstract: A method for over current protection includes: when receiving an overcurrent signal that representing an output current of DC-DC conversion module exceeds a OCP value sent by a DC-DC conversion module, controlling a protection module to be disconnected, cutting off an external insertion apparatus, so that the on-board circuit is not powered off; keeping outputting a first enable signal to the DC-DC conversion module within third preset duration longer than first preset duration, and after the first preset duration, determining whether the overcurrent signal sent by the DC-DC conversion module is re-received; if yes, it indicates that overcurrent occurs in the on-board circuit portion, controlling the DC-DC conversion module to be turned off, to stop supplying power for the on-board circuit; if no, keeping controlling the protection module to be disconnected, and continuing to merely supply power to the on-board circuit.

Abstract: An online takeover method for a heterogeneous storage volume, including: executing a service: a host executing upper layer service data access by means of a second volume label of a storage volume of a second storage; generating a volume label: a first storage taking over the storage volume of the second storage and generating a first volume label for the storage volume that has been taken over; flushing data: flushing host side cache data corresponding to the storage volume of the second storage to the storage volume of the second storage; changing a directory: changing directory information of an upper layer service running on the host; and storage migration: when the directory information of the upper layer service has been changed, migrating data of the storage volume to the first storage.

Abstract: A data processing method and system for a cloud platform, and an electronic apparatus and a storage medium. The data processing method includes: an acquisition rate of performance data of each data type in a cloud platform is determined (S101); according to the acquisition rate, a shard corresponding to each data type is determined (S102); a database mapping table according to correspondence relationships between shards and databases is generated (S103); and when a data operation request is received, according to the database mapping table, a target database corresponding to the data operation request is determined, and a corresponding operation in the target database is executed (S104). By using the above method, the load balancing of a cloud platform can be realized, and the data processing efficiency can be improved.