4270 results about "Queue" patented technology

The invention relates to the technical field of computing power scheduling, and discloses a heterogeneous resource computing power intelligent scheduling method and system. According to the method, real-time state monitoring is conducted on heterogeneous computing resources, and resource state parameters such as the computing unit utilization rate and the memory occupancy rate are obtained; task attributes and user request parameters of the task queue are collected, historical task data are processed based on the genetic algorithm optimization model to execute task demand prediction, and predicted demand parameters are generated. A dependency graph containing resource unit nodes and communication link roadsides is constructed through a resource topology analysis tool, predicted demand parameters are input into a scheduling priority classifier trained by a graph neural network, and an actual scheduling priority is identified. And executing resource conflict prediction based on the priority, inputting task feature vectors into a conflict resolution module of a fuzzy logic decision maker, outputting actual conflict resolution parameters, and finally integrating to generate a scheduling scheme containing a resource allocation sequence and an execution time table.

The invention provides a resource scheduling control method and system for a big data server, and the method comprises the steps: constructing a multi-dimensional resource portrait module, collecting the CPU, memory, network, storage I / O load and task queue length of each node in real time, and predicting a resource demand trend through a time sequence algorithm; extracting characteristics such as calculation intensity, data dependence, memory requirements, network transmission quantity and the like; adjusting the weight coefficients of the resource utilization rate, the task completion time and the energy consumption efficiency according to the system load and the historical effect; establishing a bipartite graph model by taking a resource trend as a node feature and a task vector as an edge feature, and calculating a matching score through graph convolution and a multi-objective optimization function; the scheduling scheme is synchronized by adopting a consistency algorithm; automatic rollback and reallocation are carried out when resources are detected to be insufficient; and optimizing a weight coefficient and a network parameter through reinforcement learning. Through the method, the system resource utilization rate can be improved, the task execution efficiency is improved, the overall scheduling effect stability is improved, and the system fault recovery time is shortened.

The invention discloses a Direct3D memory model compatible method based on adaptive placeholder resources, which comprises the following steps: establishing resource metadata, accessing a scene library, checking an instruction template and double sandboxes when a DXVK is started, describing related resources by the DXVK through the metadata after a D3D application is started, completing resource mapping and metadata dynamic updating, and allocating the resources to the corresponding sandboxes; the DXVK compiles an application shader code, identifies a resource access instruction, matches an access scene and a check template by combining a pipeline stage and binding slot query metadata, instantiates the check instruction and adds the check instruction to the front of a target instruction; according to the unbound resources, adaptive generation of corresponding occupied resources is carried out, the authority and the state are configured, the unbound resources are bound to a Vulkan descriptor set to replace VKNULLHANDLE, and adaptation of the shader interface and the pipeline state parameters is completed to create PSO (Particle Swarm Optimization); and intercepting an application rendering instruction analysis parameter to construct a command buffer area, submitting the command buffer area to a Vulkan queue to trigger a GPU (Graphic Processing Unit) to execute rendering, and finally completing complete conversion and adaptation from the D3D rendering logic to the Vulkan.

Embodiments include methods, electronic device, storage medium, and computer program to implement a dedicated queue based on a user request. In one embodiment, a method comprises: receiving a first message to optimize one or more data flows based on a quality-of-service request for an end user; enabling a first queue dedicated to the one or more data flows sourced for the end user in the network node based on the first message; and upon a determination of congestion in the network, marking packets in the one or more data flows stored in the first queue as candidates to drop based on a queue size of the first queue, wherein the marking is to set explicit congestion notification bits of the packets.

The invention discloses an energy-efficient task scheduling method for an edge computing system, which comprises the following steps of: acquiring a task state, a computing node resource state, a link state and an energy consumption state according to a unified time slot under a computing power network control domain, and constructing a system state vector; performing priority evaluation on the to-be-scheduled task based on the residual delay budget, the candidate node energy efficiency coefficient and the queue position to obtain a target task set; inputting a system state vector and a target task set into an energy efficiency perception deep reinforcement learning scheduling model, outputting a task-node allocation decision under the constraint of computing node resources and task time delay, and introducing a system-level energy consumption ratio, self-adaptive energy consumption penalty and exploration bias facing high-energy-efficiency nodes into rewards; and scheduling tasks according to the allocation decision, recording state transition and instant rewards, updating a double-commentator and actor network, and performing iterative execution in continuous time slots. According to the method, task success rate, time delay, load balancing and energy-saving performance are considered, and system energy consumption is reduced.

The invention relates to the technical field of robot scheduling, and discloses a robot task management framework system based on asynchronous communication. The system comprises a task asynchronous analysis module for performing time sequence decoupling on an original task instruction stream of a heterogeneous robot terminal by means of a multi-channel buffer queue to separate out independent executable task units; the resource dynamic mapping module is used for establishing a variable granularity resource allocation mapping table according to cluster real-time load and task resource requirements, and recording various resource dynamic binding relationships; the task priority reconstruction module is used for constructing a directed acyclic graph scheduling topology based on task logic dependence and deadline constraint, and generating a weighted execution sequence through topological sorting; the exception isolation processing module captures exceptions during operation and triggers an isolation strategy; and the cross-node cooperation module guarantees the multi-robot cooperation task state consensus through a distributed consistency protocol. The system is suitable for heterogeneous robot cluster complex task scenes.

The invention provides a real-time intelligent agent dynamic scheduling method and system composed of multiple large models, and is applied to the technical field of data processing. The method comprises the following steps: performing dynamic scheduling expansion around a plurality of large-model real-time intelligent agents, firstly receiving task request data of a real-time service scene, performing grouping according to task types, priorities and resource requirements, pre-configuring a scheduling strategy, and generating a pre-scheduling queue and resource pre-planning information; and in combination with the target configuration information, scheduling task information after secondary optimization is obtained through multi-dimensional optimization such as task splitting and priority rearrangement. Dynamically adjusting and generating a target scheduling instruction based on real-time system resources and a model operation state, and cooperatively processing the target scheduling instruction with the large model capability adaptation parameters to form scheduling execution information; and finally, issuing to model nodes according to preset service quality requirements and resource constraints, generating information such as adaptive parameters and priority rules through feature extraction, association processing and the like, and supporting multi-model cooperation to efficiently complete real-time service tasks.

The invention discloses a dynamic scheduling method and system for a satellite-ground cooperative computing task, and the method comprises the following steps: constructing a satellite-ground cooperative computing system comprising a remote sensing satellite cluster, a satellite server node, a ground server / data center and the like, and enabling the satellite cluster to generate an in-orbit task and to describe dependence through DAG; available computing power, task queues and other resource states of a satellite and a ground server are collected in real time, and global observation is formed; on the basis of an MADRL framework, each node is regarded as an independent agent, a task allocation action is generated through a strategy network, and a collaborative decision is modeled through POMDP; optimizing the task execution sequence of the same node by using an EFT algorithm; issuing a task and monitoring execution; task completion time and energy consumption are collected, and a network optimization strategy is updated through a reward function and experience playback; the scheduling strategy is adjusted through loop iteration, satellite offline and link fluctuation are adapted, and the resource utilization rate and the task processing efficiency are improved.

The invention discloses a Direct3D rendering model compatible method based on a dynamic template pool, which comprises the following steps: establishing three types of mapping tables between D3D and Vulkan when compiling DXVK, constructing resource metadata, and creating a core, extended and temporary three-level template pool according to the mapping tables after starting; when the D3D application creates resources, metadata is initialized, parameters are verified, physical memories are allocated and grouped, resource groups are pre-verified, a batch binding command is generated, memory binding is completed, and the metadata is updated; when a resource view is created, a template is matched from a template pool, a handle is generated after instantiation, a resource handle is associated, and a descriptor is bound; when a rendering state is set and a rendering instruction is executed, the rendering state and the rendering instruction are respectively converted into a Vulkan related state and a Vulkan related instruction through a mapping table, a handle is bound after a PSO cache is inquired, a command buffer area is submitted to a GPU queue to execute drawing, and compatible operation of the D3D application on a platform supporting a Vulkan operating system is realized under the condition that the GPU does not support VKKHRmaintence5 and VKKHRmaintence6 extension.

The invention provides a distributed event processing method and electronic equipment, which can be applied to the field of distributed technologies. The distributed event processing method comprises the following steps: receiving an event processing request at a request receiving entrance of a distributed system, and searching a priority and a mutual exclusion execution type corresponding to an event type from a preset event registration library according to an identifier representing the event type in the event processing request; based on the priority and the timestamp of the event processing request, inserting an event object obtained after packaging the event processing request into a priority scheduling queue; after an event object used for execution is extracted from the priority scheduling queue, conflict detection is carried out on a mutual exclusion execution type corresponding to the event type based on a global mapping table maintained in the distributed system; and under the condition that the conflict is not detected, calling a service processor corresponding to the event type to execute the event object, and updating the current execution count of the event type in the global mapping table based on the execution state.

The invention relates to a production abnormality automatic identification and recovery process control method, which comprises the following steps of S1, realizing second-level synchronization of multi-source heterogeneous data, constructing a real-time data flow pipeline and generating a total-factor production situation data flow through a distributed message queue by adopting a Modbus / TCP protocol analysis algorithm based on industrial Internet of Things edge calculation; through the cooperative effect of industrial protocol analysis and distributed message queues, second-level synchronization of multi-source heterogeneous data is realized, a real-time data flow pipeline is constructed, data acquisition delay is effectively eliminated, the timeliness of anomaly detection is ensured, a dynamic weight distribution mechanism of a rule engine and a long and short-term memory network prediction model is adopted, and the real-time performance of the system is improved. By combining sliding window threshold detection, the accuracy and coverage of anomaly recognition are improved, false alarm and missing alarm caused by a single detection mechanism are reduced, and multi-dimensional root cause tracing is performed by combining a fault mode knowledge base through combined application of time sequence correlation analysis and a causal diagram inference engine.

The invention provides a final consistency processing system and method for transaction messages. The method comprises the following steps: acquiring transaction state metadata and a service message body about a transaction message in a distributed system, storing the service message body into a distributed message queue, and storing the transaction state metadata into a memory database; constructing a transaction state index structure; determining a physical offset based on the business message body, creating a transaction state record in the memory database based on the physical offset, and generating a to-be-rechecked transaction task according to the transaction state record; based on the transaction creation time and the system load factor, determining a rechecking interval, and executing the to-be-rechecked transaction task; when it is detected that the transaction state is changed into a transaction confirmation state from a to-be-confirmed state, pulling a service message body according to the physical offset; and performing double confirmation on the pulled service message body. The problems of performance bottleneck and low recovery efficiency caused by strong coupling of the transaction state and the message body are solved.

The invention discloses a PLC scheduling algorithm based on a multi-core soft real-time operating system, and relates to the technical field of voltage control, and the algorithm comprises the following four steps: 1, starting a target industrial control system, completing hardware self-inspection, memory allocation and communication module starting, and sending an initialization ready signal after confirming that an assembly is normal; 2, according to PLC task functions and real-time performance, four types of standardized tasks are divided, priority levels are defined, sub-priorities are configured, a dependency relationship is identified, and a task chain is established; 3, counting the number of tasks and task chains, distributing resources by using multiple binding cores, and constructing a global ready queue to adapt to dynamic scheduling; and 4, monitoring an interrupt event through an interrupt manager, realizing high-priority task preemption by semaphores, and allocating tasks to corresponding CPU cores, so that the problems of soft real-time multi-core task preemption and data conflict are solved, the scheduling real-time performance and the data security are improved, the system failure rate is reduced, and the industrial control efficiency and stability are improved.

Methods, systems, and devices for wireless communications are described. A user equipment (UE) may receive, at a first modem and from a first device, one or more data packets. Based on an explicit congestion notification (ECN) capable transport (ECT) codepoint indication in respective header portions of the data packets, the UE may route the data packets from the first modem to two or more queues associated with a second modem of the UE. Based on receiving buffer and link state information associated with the second modem, the UE may perform a congestion indication procedure that is associated with congestion of a communication link associated with the UE. The UE may thereafter transmit, to a second device, packet data from the two or more queues associated with the second modem of the UE based on a respective priority associated with the two or more queues.

The invention relates to the technical field of remote direct memory access, in particular to an RDMA shared receiving queue optimization method and device, equipment, a medium and a product. A shared receiving queue is divided into multiple containers, and the multiple containers are connected in series through a logic linked list; when the RDMA data of the queue pair arrives, if the container is not bound, distributing a container for the queue pair; monitoring the release state of the shared receiving queue entry; and when the receiving side software monitors that the target container is in the completely idle state, re-linking the target container into the logic linked list for distribution independent of the states of other containers in the plurality of containers. According to the method disclosed by the invention, dynamic reuse of resources can be realized through resource monitoring taking the container as granularity and an out-of-order recovery mechanism independent of other container states. After all the SRQEs in any container are released, the SRQEs can be immediately put into redistribution, so that the problem of waste of tail idle resources caused by occupation of head resources in a traditional linear queue can be thoroughly avoided.

The invention discloses an operation data processing method and system based on a compliance verification library and a risk control plan, and belongs to the technical field of data compliance and risk control, and the method comprises the steps: obtaining an execution plan of a target operation task and risk control strategy information of the execution plan, analyzing operation data items, distributing corresponding compliance verification time windows and levels, and forming an operation data verification scheduling table; according to the verification level and the business domain, a rule set is extracted from a compliance verification library, and rules comprise predefined rules, model output rules and historical violation rules; field-level matching, logic consistency verification and cross-source linkage verification are executed in the verification time window; when the verification result meets the triggering condition, generating an abnormal event, and executing response operations such as delay, recheck or interception; and finally, writing the execution state and the verification result into an abnormal log queue for risk control strategy parameter updating or re-evaluation. According to the method, the dynamic compliance verification and strategy closed-loop optimization of the operation data are realized, and the intelligent compliance control capability of the system is improved.

The invention relates to the technical field of cloud storage, can be applied to business scenes of financial science and technology, medical health and the like, and discloses a load and message queue regulation-based migration method, device, equipment and medium. Monitoring message backlog amount and consumption rate to generate message queue state information, generating migration regulation and control parameters based on the two types of state information, determining concurrency level and issuing rhythm by combining object size classification information, reducing issuing frequency and preferentially scheduling small objects when the message backlog amount is increased and oversize objects exist, and distributing the tasks to the working nodes according to the concurrency level and the issuing rhythm to execute migration, and adjusting migration regulation and control parameters based on execution state updating information. According to the method, concurrency and rhythm are regulated and controlled in real time, object size classification is combined, the problems that traffic is uncontrollable and throughput is affected by large files are solved, and efficient and stable data migration is achieved.

The invention discloses a tamper-resistant end-to-end service data storage method and system, and relates to the technical field of service data storage methods, and the method comprises the steps: initializing a hardware security module and a monotone increasing clock in terminal equipment, building a unique identity identifier of the equipment, and preparing a log environment for recording service data; the method comprises the following steps: collecting current business data and associated meta-information thereof, and constructing input content of current data processing in combination with an equipment identity, a clock value, a data structure template and an abstract value generated by previous business data; hash operation is carried out on the input content to generate an abstract of the business data, the abstract is signed by using a hardware security module to form an equipment side digital signature, and the abstract, the signature and meta-information are written into a log only increasing but not changing and an offline queue; and after the terminal recovers the network connection, generating a unique idempotent key based on the abstract of the service data or the combination of the equipment identity and the clock, and submitting an evidence storage request containing the abstract and the signature to the server.

The invention provides a DAG arrangement and execution method and system based on operator dynamic management, and the method comprises the steps: constructing an initial directed acyclic graph DAG, dynamically loading operator plug-ins registered in the initial DAG, obtaining context information during operation, and storing the context information to a context container; dynamically selecting a next execution path based on the real-time data in the context container, generating an executable operator queue according to the updated DAG structure, and monitoring the execution state of the operator node in real time; and dynamically modifying the DAG structure according to a trigger condition in the execution process, and outputting a final result after the DAG is executed. According to the DAG arrangement and execution method and system based on operator dynamic management, flexible configuration of the task process, dynamic scheduling of the operator level and intelligent decision making during operation are achieved, and therefore the universality, adaptability and execution efficiency of the system are remarkably improved.

The invention relates to a distributed storage access method initiated by a GPU (Graphics Processing Unit) in an AI (Artificial Intelligence) scene. The method comprises the following steps: S100: constructing a software cache module and an I / O (Input / Output) queue module in a GPU memory; s200, the GPU thread queries the target data through the software cache module, reads the target data if the target data is hit, and enters the next step if the target data is not hit; s300, when the I / O request is not hit, the repeated I / O requests initiated by multiple threads are automatically combined to generate a single effective I / O request, and the request is written into an I / O queue module; s400, the user mode distributed storage client monitors the written submission queue, reads the target data and then transmits the target data to the software cache module, and a completion queue of the I / O queue module is updated; and S500, the GPU thread polls the updated completion queue, and updates the reference count of the software cache module after reading the data. The main purpose of the invention is to realize high-throughput and low-delay data transmission without the participation of a CPU (Central Processing Unit).

The invention discloses a priority scheduling method and system for an on-orbit data processing task. The method comprises the following steps: constructing a DAG model, decomposing a macroscopic task into subtasks, defining a subtask dependency relationship, and endowing a subtask node with a priority attribute; setting or dynamically adjusting the priorities of the sub-tasks according to task service types, processing delay requirements or external instructions; maintaining a schedulable queue containing subtasks which have completed precursor tasks and meet scheduling conditions; calculating task scheduling priority indexes based on the queue, sorting, and selecting optimal calculation resource allocation; the priority is dynamically adjusted according to task waiting time or an external instruction during execution, and hunger of low-priority tasks is avoided; the multiple processing units execute the high-priority tasks in parallel according to needs; and after the task is completed, updating a subsequent node dependency state, and adding a new schedulable task cycle. According to the method, high-priority task resource allocation can be guaranteed preferentially, low-priority task hunger is avoided, and the overall throughput rate of the system and the self-adaptive capacity to on-orbit diversified tasks are remarkably improved.

The invention discloses a hybrid real-time migration method for an unmanned aerial vehicle visual model reasoning task, and the method comprises the steps: estimating the migration delay time and the bandwidth needed by migration through continuously monitoring the resource state of unmanned aerial vehicle equipment, and comparing the migration delay time and the bandwidth needed by migration with the set maximum allowable end-to-end time delay, the minimum bandwidth demand and the bandwidth needed by migration; and judging whether task migration is triggered. And when the isomorphic migration is triggered, the source device runs the task being executed to the transferable check point and stops receiving the new task, and the packaged queue and state are transmitted to the target device through hierarchical transmission strategy transmission. When heterogeneous migration is triggered, the source device encapsulates the model state information and the computing context into a lightweight request packet F (r) and transmits the lightweight request packet F (r) to the target device. And after the migration task is completed, switching a system state, and recycling occupied resources. According to the method, limitation of various resource conditions is comprehensively considered, the task can be dynamically migrated to the most suitable node, and non-inductive knowledge and state consistency in the migration process are guaranteed.

The invention discloses a Modbus master station control system for multi-queue task scheduling, and relates to the technical field of automatic control. A multi-queue task scheduling architecture is adopted, firstly, hierarchical classification is performed on Modbus communication tasks according to functional characteristics of the Modbus communication tasks, then different types of tasks are placed in independent task queues with different scheduling strategies, and finally, the tasks are scheduled by a task scheduling manager according to priorities and scheduling strategies of the queues. And in combination with an intelligent optimization mechanism based on historical communication performance data analysis, tasks are efficiently and orderly scheduled and executed, so that the communication efficiency of the Modbus network, the emergency task response speed and the overall reliability of the system are remarkably improved.

According to the automatic driving data transmission scheduling method and device, the equipment and the readable storage medium, dynamic priority weights are generated in a mode of combining reinforcement learning and a rule engine, and bandwidth demand changes, historical transmission performance and queue occupation states of all sensors can be adapted in real time; the problem that a low-priority sensor cannot acquire the bandwidth at a high load is effectively avoided, and meanwhile, efficient utilization of bandwidth resources can be realized through dynamic weight adjustment when a high-priority sensor is idle, so that the bandwidth utilization rate in a vehicle-mounted multi-sensor data transmission process is remarkably improved; the real-time performance and reliability of data transmission required by automatic driving decision making are guaranteed, and then the overall performance of an automatic driving system is improved.

The invention relates to the technical field of network communication, and discloses an AI load-oriented network topology awareness and communication optimization method, an AI load-oriented network topology awareness and communication optimization device and a related system. The method comprises the following steps: acquiring gradient transmission delay data among computing nodes in a distributed AI training system, and constructing a gradient distribution table of gradient tensors and RDMA intelligent network card queue resources; performing aggregation operation on the gradient tensor in the gradient distribution table on a programmable data plane of the RDMA intelligent network card to obtain pre-aggregated gradient data; calculating a transmission efficiency index of the pre-polymerized gradient data and solving to obtain a bandwidth allocation coefficient; and adjusting an MTU frame parameter, a WQE queue depth and a credit window of the RDMA intelligent network card according to the bandwidth allocation coefficient, and generating a hybrid execution scheme. According to the method, intelligent distribution of the gradient tensor between hardware unloading and software processing is achieved, and the performance close to the optimal performance is obtained under the condition that resources are limited.

The invention discloses a machine tool running state collection and evaluation method and system, and relates to the field of intelligent machine tools, and the method comprises the steps: collecting machine tool multi-dimensional data through a multi-channel collection device; performing time alignment on different sampling frequency signals by adopting a dynamic synchronization algorithm of clock drift compensation; executing noise suppression, feature extraction and compressed encoding at the edge computing node to generate local state abstract data; sending the abstract data to a central server through a distributed message queue, and realizing multi-application concurrent access based on load dynamic scheduling; and the central server performs fusion analysis on the abstract data by using a multi-modal fusion deep network model, and outputs a health degree index and an overall operation score of the key parts of the machine tool, thereby realizing intelligent evaluation and health prediction of the operation state of the machine tool. The method realizes global perception and intelligent evaluation of the machine tool operation state, and can be widely applied to operation monitoring and health management of high-end numerical control machine tools, complex machining centers and unmanned production lines.

The invention discloses a large file concurrent uploading method and system based on a dynamic fragmentation strategy and resource optimization, and the method comprises the steps: a client generates a file fingerprint according to the file size of a large file and metadata, and dynamically calculates the fragmentation size of the large file; the client carries out fragment cutting on the large file according to the calculated fragment size, establishes a concurrent uploading queue and uploads a plurality of fragments to a server side in parallel; the server selects an optimal storage node to store each fragment based on a consistent hash ring and a multi-dimensional weight model; the client side and the server side track a fragment uploading state through a fragment serial number and a cyclic redundancy check (CRC) value, and after all fragments are uploaded, integrity check and logic recombination are carried out; and when network interruption or uploading failure is detected, the client side and the server side execute an intelligent retry strategy or a splitting rescheduling strategy of breakpoint resume and failure fragmentation.

The invention provides a task processing method and device, electronic equipment, a storage medium and a program product, and relates to the technical field of computers. The method comprises the steps of adding a to-be-scheduled task to a task queue corresponding to a task type in a plurality of task queues according to the task type of the to-be-scheduled task; wherein the priorities of different task queues are different, and the task scheduling strategies corresponding to different task queues are different; sequentially scanning the plurality of task queues according to the priorities of the task queues, and determining a target queue from the plurality of task queues according to task existence conditions of the scanned task queues; and determining a target task from the target queue according to the task scheduling strategy corresponding to the target queue, and performing task processing based on the target task. According to the method, differentiated scheduling of different types of tasks can be realized.

The present application relates to a data acceleration processing method and apparatus, a chip, a device, and a readable storage medium. The method comprises: receiving a trigger message for a data processing task, the trigger message comprising pointer information of a queue, the pointer information of the queue being address information of the queue in a memory, and the queue being used for storing a task descriptor of the data processing task; updating the pointer information of the queue to a queue manager for scheduling by a queue scheduler; when the queue is selected by the queue scheduler, transmitting the pointer information of the queue to a descriptor cache acquisition unit, so as to notify the descriptor cache acquisition unit to acquire the task descriptor of the data processing task from the memory; and selecting an idle engine from among a plurality of identical data acceleration processing engines as a target engine, and executing the data processing task on the basis of the task descriptor by means of the target engine.

The invention discloses a data transmission method and computer equipment, and relates to the technical field of data transmission. The method comprises the following steps: acquiring a sending work request containing a local sending queue pair number and a work queue element; querying target sending information by taking a local sending queue number as an index; storing the working queue elements and the target sending information as operation contexts into an internal memory of a network card of a request end, and recording corresponding indexes; assembling a target data packet based on the target sending information and the memory index; receiving a feedback data packet of the target data packet and extracting a memory index and a serial number carried by the feedback data packet; if the feedback data packet is a correct confirmation data packet, judging that the sending of the work request is completed and releasing the context; and if the feedback data packet is a confirmation failure data packet, calculating a retransmission data initial address and organizing retransmission. The technical problem that the effective connection number of network cards is limited by hardware resources in the prior art is solved, and the technical effect of supporting large-scale network parallel communication is achieved.