Abstract: A method and apparatus for managing buffering of data packets of a network card, a terminal and a storage medium are provided. The method includes: setting ring buffer queues, setting a length of each ring buffer queue, then setting a buffer pool formed by two ring buffer queues, and setting the two ring buffer queues in the buffer pool as a busy queue and an idle queue, respectively; a network card driver receiving data packets from a data link, classifying the data packets, sequentially buffering the classified data packets into the busy queue, and then sequentially mapping addresses of the buffered data packets in the busy queue into the idle queue; acquiring latest addresses of the buffered data packets in the busy queue; and the upper-layer application successively acquiring and processing the buffered data packets, and successively releasing the addresses of the processed buffered data packets in the busy queue.

Abstract: The present application discloses a system, readable storage medium, and method for analyzing a cause of training-task queuing, wherein the method includes the steps of: obtaining a required resource that is required by a training task inputted by a user and a remaining resource of a cluster; in responding to that the remaining resource does not satisfy the required resource, obtaining a plurality of cluster center data that are pre-generated in a cluster model; regarding the required resource and the remaining resource as sample data, and calculating distances between the sample data and each of the cluster center data; and feeding back a cause corresponding to the cluster center datum that has a minimum distance with the sample data.

Abstract: Disclosed in the present disclosure is a hard disk fixing device, including: a seating, which includes a base fixed on a main board and a first pair of elastic buckles perpendicular to the base and arranged at two sides thereof; a first supporting seat, which includes a main body and an extending portion extending from the main body, wherein the extending portion may be snap-fitted between the first pair of elastic buckles, a first guiding slot is provided inside the main body and a through hole is provided in a side surface of the main body, and the first guiding slot is in communication with the through hole; a first adjusting fastener, which includes a sliding rod and a limiting rod connected to the sliding rod.

Abstract: Disclosed are a method, system and device for manufacturing a printed circuit board, and a computer storage medium. The method comprises: acquiring a printed circuit board to be manufactured which includes a via hole; acquiring shape information of the via hole; acquiring connection information of circuit layers in said printed circuit board; assembling preset conducting devices according to the connection information and the shape information, so as to obtain a target conducting device that matches the connection information and the shape information; guiding the target conducting device into the via hole to obtain a conducting printed circuit board; and connecting the conducting printed circuit board to obtain a target printed circuit board, wherein the types of the preset conducting devices comprise a metal conducting device, a non-metal conducting device and a semi-metal conducting device.

Abstract: The present application discloses Graphics Processing Unit (GPU)-shared dispatching and single-machine multi-card methods, systems, and devices, applied to a dispatcher of a central control host. In the present application, a resource occupation marker and an update marker are added to a pod, the resource occupation marker is used to prove that the pod is running, and the update marker is used to indicate that resource usage information in an annotation of the pod is consistent with actual resource usage information, whereby during dispatching, a resource usage situation of each pod may be determined. The resource usage information, i.e., GPU information of an actually used GPU, in the annotation of the pod is updated before dispatching, to ensure that the GPU information in the annotation is consistent with GPU information actually recorded in an environment variable of the pod.

Abstract: Provided are a storage system boot method and apparatus, and a computer-readable storage medium. The method includes: determining a configuration space by a bridge device; during booting a Basic Input Output System, after an initialization of a Peripheral Component Interconnect Express (PCIe) device is detected, determining, by connection information read from the configuration space, whether the preset hardware device is connected; if YES, continuing booting until finished; if NO, sending a reconnection instruction to stabilize a speed of a PCIe hardware link; when a number of the reconnection instruction sent is greater than a first preset value and the preset hardware device fails to be connected, sending a system reboot instruction; and when the number is less than the first preset value, continuing to determine whether the preset hardware device is connected after waiting for a preset duration.

Abstract: A mass file retrieval method and apparatus based on an AI training platform, and a device. The method includes: an AI training platform acquiring a retrieval task issued by a user, the AI training platform generating a retrieval thread flow according to the retrieval task, and controlling a service logic of a retrieval process according to the retrieval thread flow; and the AI training platform, performing sequential encoding on files in a database in units of folders, generating ordered queue folders, extracting retrieval key words from the retrieval task, and performing keyword retrieval on each ordered queue folder by combining a binary search and depth-first traversal.

Abstract: Disclosed are a method for data synchronization between a host side and a Field Programmable Gate Array (FPGA) accelerator, a Bidirectional Memory Synchronize Engine (DMSE), a FPGA accelerator, and a data synchronization system. The method includes: in response to detection of data migration from a host side to a preset memory space, generating second state information according to first state information in a first address space, and writing the second state information to a second address space (S201); and in response to detection of the second state information in the second address space, calling Direct Memory Access (DMA) to migrate data in the preset memory space to a memory space of a FPGA accelerator, and copying the second state information to the first address space, so as to implement synchronization (S202).

Abstract: A method for arbitrating cluster includes: setting heartbeat between every two adjacent nodes in the cluster; in response to disconnection of heartbeat among nodes occurs, dividing the nodes into a plurality of sub-clusters, and determining whether the sub-cluster having a largest number of nodes is unique; in response to the sub-cluster having the largest number is not unique, selecting one node in the sub-clusters having the largest number of nodes, to make the selected nodes send a request of reservation to a logical volume of a third-party storage; in response to success of reservation of node, determining whether data for arbitrating in the logical volume of the third-party storage is valid; and in response to the data is invalid, writing information for arbitrating of the node of success into the logical volume, to make the sub-cluster, to which the node belongs, obtain a control right of the cluster.

Abstract: A B+ tree access method, apparatus, and a medium include: when a B+ tree creation instruction is acquired, determining whether there is an idle underlying tree file mapped to a Dynamic Random Access Memory (DRAM); if YES, storing underlying data of a B+ tree in the underlying tree file; if NO, creating a new target underlying tree file, and mapping the target underlying tree file to the DRAM, so as to store the underlying data of the B+ tree in the target underlying tree file; and when a number of layers of the B+ tree is greater than or equal to a preset threshold, storing, in a preset storage area, data of a layer numbered to be greater than or equal to the preset threshold in the B+ tree. Based on a data structure of the B+ tree, data needs to be accessed from the underlying data whenever read.

Abstract: A metadata volume bitmap data conflict processing method. When the space allocation of a metadata volume encounters conflicts of 0-to-1 and 1-to-0, a conflict bit is set to be 1, so that when a garbage collector periodically performs space collection processing, data loss when data is overwritten due to collection of the bit to space corresponding to the bit in the case that I bit is detected to be 0 at one end may be avoided; when a conflict happens, conflict bits are both set to be 1, so that the conflict bits may be prevented from being allocated again, the space is prevented from being repeatedly allocated and repeatedly written and thus the data loss due to data overwriting is avoided, and after the garbage collection collects the conflict bits, continuous allocation may be performed, thereby implementing the repeated utilization of the conflict bits.

Abstract: Disclosed are a Cluster Trivial Database (CTDB)-based virtual Internet Protocol (IP) address allocation method and apparatus, a distributed storage device, and a computer-readable storage medium. The method is applied to a distributed storage cluster. A known IP Version 4 (IPV4) virtual IP address parameter and available IPV4 virtual IP address parameter corresponding to an IPV4 virtual IP address and a known IP Version 6 (IPV6) virtual IP address parameter and available IPV6 virtual IP address parameter corresponding to an IPV6 virtual IP address may be acquired to obtain an IPV4 allocation result and an IPV6 allocation result by calculation respectively, and then virtual IP addresses are allocated to each normal node by use of the IPV4 allocation result and the IPV6 allocation result.

Abstract: A method for searching for free blocks in bitmap data. For searching for free blocks in the bitmap data, distribution information of free blocks can be directly obtained by performing array value solving operation, with the value of the integer data obtained by dividing the bitmap data into blocks being used as an index, on a template array. Also provided are an apparatus for searching for free blocks in the bitmap data, a device, and a readable storage medium.

Abstract: Disclosed are a FAAS distributed computing method and apparatus. The method includes: decomposing a computation task into multiple steps with correlation and execution order, and constructing multiple mirror images and multiple method groups respectively; creating multiple dockers in a process and allocating hardware resources according to the mirror image corresponding to a specific step; transferring, according to a dynamic pointer of the corresponding method group, a data processing result to the method group corresponding to the next step, and simultaneously storing a generated intermediate result in a distributed memory file system; modifying dynamic pointers of the multiple method groups in real-time to dynamically adjust the correlation and execution order of the multiple steps; monitoring a running state of each method in the multiple method groups, restarting the docker of the method or loading latest docker snapshot of the method and expanding or reducing a capacity of the method groups.

Abstract: A hardware environment-based data quantization method includes: parsing a model file under a current deep learning framework to obtain intermediate computational graph data and weight data that are independent of a hardware environment; performing calculation on image data in an input data set through a process indicated by an intermediate computational graph to obtain feature map data; separately performing uniform quantization on the weight data and the feature map data of each layer according to a preset linear quantization method, and calculating a weight quantization factor and a feature map quantization factor (S103); combining the weight quantization factor and the feature map quantization factor to obtain a quantization parameter that makes hardware use shift instead of division; and finally, writing the quantization parameter and the quantized weight data to a bin file according to a hardware requirement so as to generate quantized file data (S105).

Abstract: A method includes: setting, in a main training network, an auxiliary training network having the same architecture as the main training network, performing data enhancement on an original image to obtain an enhanced image; inputting the original image into the main training network, inputting the enhanced image into the auxiliary training network; determining whether an intersection-over-union value of a second prediction frame generated by the auxiliary training network and a target frame is greater than an intersection-over-union value of a first prediction frame generated by the main training network and the target frame; and in response to the intersection-over-union value of the second prediction frame and the target frame being greater than the intersection-over-union value of the first prediction frame and the target frame, replacing the intersection-over-union value of the first prediction frame and the target frame with the intersection-over-union value of the second prediction frame and the target f

Abstract: A method and apparatus for processing cluster node failure, a computer device and a readable storage medium. The method includes: circularly acquiring state information of multiple nodes in a cluster, and on the basis of the state information, determining whether a corresponding node fails; in response to failure of the node, sending failure information to multiple OSDs under the node; in response to the multiple OSDs receiving the failure information, according to the failure information, selecting a Monitor to send down information, and setting states of the multiple OSDs to be down; and in response to the Monitor receiving the down information, updating an OSDMap on the basis of the down information, and sending the updated OSDMap to OSDs under other nodes.

Abstract: Provided are a turbulence field update method, apparatus, and device, and a computer-readable storage medium. The method includes: obtaining sample turbulence data; performing model training by use of the sample turbulence data to obtain a reinforcement learning turbulence model; calculating initial turbulence data of a turbulence field by use of a Reynolds Averaged Navior-Stokes (RANS) equation; processing the initial turbulence data by use of the reinforcement learning turbulence model to obtain a predicted Reynolds stress; and performing calculation on the predicted Reynolds stress by use of the RANS equation to obtain updated turbulence data.

Abstract: A method and apparatus for online migration of a multi-disk virtual machine into different storage pools are provided. The method includes: constructing disk trees by using each active disk to be migrated of a virtual machine and a plurality of snapshot disks that are correspondingly generated based on each active disk to be migrated, so as to record a generation time sequence of a plurality of disks in the disk tree; migrating a memory snapshot file; and determining a target storage pool corresponding to each active disk to be migrated, and migrating a plurality of disk trees into corresponding target storage pools respectively according to the generation time sequence of the disk to be migrated in the disk tree.

Abstract: A single-level single-line full-duplex bus communication method and system are disclosed. The method includes: transmitting, by a first signal transceiver, data according to a first internal transmitter clock F1, simultaneously monitoring a level change on a bus, and parsing received data; transmitting, by a second signal transceiver, data according to a second internal transmitter clock F2, simultaneously monitoring the level change on the bus, and parsing received data; and communicating between the first and second signal transceivers by means of a single line, wherein the first and second transmitter clocks satisfy a relationship: F1>F2*(length of data unit+2). The system achieves single-level single-line full-duplex communication by using different coding formats and different internal transmitter clocks, whereby the number of signal lines can be reduced, single-level communication can be achieved by using universal digital levels, i.e., 0, 1, and the hardware implementation difficulty can be reduced.