Dynamic adaptive data stream acceleration transmission method and device for signal creation environment

Abstract

The application provides a dynamic adaptive data stream acceleration transmission method and device for a signal creation environment. The method comprises the following steps: detecting the capabilities of a central processing unit, an operating system, a network interface and a cryptographic acceleration unit in a signal creation computing node to obtain environment capability information, and organizing the environment capability information into a signal creation environment capability profile; analyzing the service type, data size, timeliness level and security level of a data stream, and combining a preset classification rule to generate a data stream profile corresponding to the data stream; collecting running state information across layers and organizing the running state information into a running state vector; based on the signal creation environment capability profile, the data stream profile and the running state vector, executing a preset strategy decision process to determine a transmission strategy parameter set for the data stream; and sending data fragments according to the transmission strategy parameter set. The application can improve the utilization rate of signal creation environment resources, improve the transmission efficiency of multiple types of data streams and reduce the manual optimization overhead of transmission parameters.

Description

Technical Field

[0001] This application relates to the field of data transmission technology, and in particular to a dynamic adaptive data stream acceleration transmission method and apparatus for the domestic IT innovation environment. Background Technology

[0002] In the process of upgrading information systems in key industries such as finance and power, the adoption of domestically produced central processing units, operating systems, and cryptographic algorithms in domestically developed information technology environments has become the mainstream construction direction. In such environments, multiple services such as large-scale file distribution, database synchronization, log aggregation, and real-time business messaging run in parallel, placing high concurrency, high reliability, and controllable latency requirements on cross-node and cross-regional data flow transmission.

[0003] In existing technologies, some domestically developed large file transfer systems and cross-gateway transfer products typically define custom transmission protocols at the application layer. Through methods such as multi-connection concurrency, data fragmentation, and breakpoint resumption, they improve data transmission performance to some extent in complex network environments. Simultaneously, there are also adaptive adjustment schemes based on bandwidth estimation and packet loss detection, used to adjust sending rates and retransmission strategies for high-latency, high-packet-loss links.

[0004] However, existing solutions mostly focus on parameter optimization at the protocol and link layers, and generally suffer from the following problems: First, they lack fine-grained modeling and utilization of the hardware and software capabilities of the domestically developed computing nodes, and fail to dynamically adjust transmission strategies based on the differences in capabilities of domestic processors, network interfaces, and cryptographic acceleration units; second, they lack a unified profile and classification of application layer attributes such as service type, timeliness level, and security level, and multiple service data streams often use the same set of fixed transmission parameters, making it difficult to take into account the differentiated requirements of different services for latency and throughput; in addition, cross-layer operational status information cannot be used in a closed loop in policy decision-making, and transmission strategy updates rely more on manual experience configuration, lacking a systematic adaptive adjustment mechanism, making it difficult to respond in a timely manner to the dynamic changes brought about by complex loads and security policies in the domestically developed environment. Summary of the Invention

[0005] In view of this, embodiments of this application provide a dynamic adaptive data stream acceleration transmission method and apparatus for the domestic IT innovation environment, in order to solve the problems of lack of domestic IT innovation environment capability awareness, lack of data stream profile management, and lack of cross-layer state-driven adaptive strategy optimization in the prior art.

[0006] A first aspect of this application provides a dynamic adaptive data stream acceleration transmission method for a domestically developed information technology (IT) environment, comprising: performing capability detection on a central processing unit, operating system, network interface, and cryptographic acceleration unit in an IT computing node to obtain environmental capability information characterizing computing capabilities, network transceiver capabilities, and cryptographic processing capabilities, and organizing the environmental capability information into an IT environment capability profile; upon receiving a data stream transmission request, parsing the service type, data size, timeliness level, and security level of the data stream, generating a data stream profile corresponding to the data stream based on preset classification rules, and assigning a stream identifier to the data stream to identify the data stream profile; during data stream transmission, collecting cross-layer operational status information corresponding to the IT environment capability profile and the data stream profile, and transmitting the operational status information... Information is organized into a runtime state vector. Based on the capability profile, data flow profile, and runtime state vector of the information technology innovation environment, a preset strategy decision-making process is executed to determine the set of transmission strategy parameters for the data flow. According to the set of transmission strategy parameters, the data flow corresponding to the flow identifier is processed by fragmentation, connection and queue configuration, cryptographic processing configuration, and transmission scheduling. Data fragments are sent according to the fragmentation configuration parameters and concurrent transmission configuration parameters in the set of transmission strategy parameters. During the transmission process, transmission performance indicators corresponding to the set of transmission strategy parameters are collected, the transmission performance indicators are associated with the set of transmission strategy parameters, and the set of transmission strategy parameters is updated based on the transmission performance indicators. The updated set of transmission strategy parameters is used for accelerated transmission control of subsequent data flows with the same or similar data flow profiles.

[0007] A second aspect of this application provides a dynamic adaptive data stream acceleration transmission device for a domestic IT innovation environment, comprising: a detection module, configured to detect the capabilities of a central processing unit, operating system, network interface, and cryptographic acceleration unit in a domestic IT innovation computing node, obtain environmental capability information characterizing computing capabilities, network transceiver capabilities, and cryptographic processing capabilities, and organize the environmental capability information into a domestic IT innovation environment capability profile; a parsing module, configured to, upon receiving a data stream transmission request, parse the service type, data size, timeliness level, and security level of the data stream, generate a data stream profile corresponding to the data stream based on preset classification rules, and assign a stream identifier to the data stream to identify the data stream profile; and a collection module, configured to, during data stream transmission, collect cross-layer operational status information corresponding to the domestic IT innovation environment capability profile and the data stream profile, and convert the operational status information into a data stream profile. The system is organized into a runtime state vector. A determination module, based on the capability profile of the information technology innovation environment, the data flow profile, and the runtime state vector, executes a preset strategy decision-making process to determine the set of transmission strategy parameters for the data flow. A sending module, based on the transmission strategy parameter set, performs fragmentation processing, connection and queue configuration, password processing configuration, and sending scheduling on the data flow corresponding to the flow identifier. It then sends the data fragments according to the fragmentation configuration parameters and concurrent transmission configuration parameters in the transmission strategy parameter set. An update module, during the sending process, collects transmission performance indicators corresponding to the transmission strategy parameter set, associates these indicators with the transmission strategy parameter set, updates the transmission strategy parameter set based on the indicators, and uses the updated transmission strategy parameter set for accelerated transmission control of subsequent data flows with the same or similar data flow profiles.

[0008] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.

[0009] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.

[0010] The above-described technical solutions adopted in the embodiments of this application can achieve the following beneficial effects:

[0011] By probing the central processing unit, operating system, network interface, and cryptographic acceleration unit in the domestic IT innovation computing nodes, environmental capability information representing computing power, network transceiver capabilities, and cryptographic processing capabilities is obtained, and this environmental capability information is organized into a domestic IT innovation environment capability profile. Upon receiving a data stream transmission request, the service type, data size, timeliness level, and security level of the data stream are parsed, and a data stream profile corresponding to the data stream is generated based on preset classification rules. A stream identifier is assigned to the data stream to identify the data stream profile. During data stream transmission, cross-layer collection of operational status information corresponding to the domestic IT innovation environment capability profile and the data stream profile is performed, and this operational status information is organized into an operational status vector. Based on the domestic IT innovation environment capability profile... This application uses image and data stream profiles and runtime status vectors to execute a preset strategy decision-making process, determining the set of transmission strategy parameters for the data stream. Based on the transmission strategy parameter set, it performs fragmentation processing, connection and queue configuration, cryptographic processing configuration, and transmission scheduling on the data stream corresponding to the stream identifier. Data fragments are then transmitted according to the fragmentation configuration parameters and concurrent transmission configuration parameters in the transmission strategy parameter set. During transmission, transmission performance indicators corresponding to the transmission strategy parameter set are collected, correlated with the transmission strategy parameter set, and updated based on the transmission performance indicators. The updated transmission strategy parameter set is then used for accelerated transmission control of subsequent data streams with the same or similar data stream profiles. This application can improve resource utilization in the domestic IT innovation environment, increase the transmission efficiency of multiple types of data streams, and reduce the overhead of manual tuning of transmission parameters. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart illustrating the dynamic adaptive data stream acceleration transmission method for the information technology innovation environment provided in this application embodiment;

[0014] Figure 2 This is a schematic diagram of the structure of the dynamic adaptive data stream acceleration transmission device for the information technology innovation environment provided in the embodiments of this application;

[0015] Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0016] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0017] In existing technologies, various domestically developed large file transfer systems and cross-network data transmission products have emerged for the information technology innovation environment of key industries such as government affairs, finance, and power. These systems typically define custom transmission protocols at the application layer and optimize the transmission performance of high-latency and high-packet-loss links through methods such as multi-connection concurrency, data fragmentation, and breakpoint resumption. Meanwhile, some solutions also use network-side parameters such as bandwidth estimation and packet loss detection to adaptively adjust the sending rate and retransmission strategy to improve data transmission performance in complex network environments.

[0018] However, these existing technologies generally suffer from the following problems: First, the transmission strategies mainly involve fixed configurations or coarse-grained adjustments based on link layer and protocol layer parameters, lacking fine-grained modeling and awareness of the hardware and software capabilities of the central processing unit, operating system, network interface, and cryptographic acceleration unit within the domestic computing node. This makes it difficult to adjust the transmission strategies in a timely manner based on the differences in capabilities of domestic processors and domestic cryptographic units. Second, there is a lack of unified profile management for application layer attributes such as service type, data scale, timeliness level, and security level of data flows. Different service data flows often share a set of fixed transmission parameters, making it difficult to simultaneously meet the differentiated needs of high real-time services and high-throughput services. Third, information such as network status, system load, and cryptographic operation load collected across layers during operation fails to form a closed-loop feedback mechanism. The adjustment of transmission parameters mainly relies on manual experience and lacks a systematic and self-evolving strategy update mechanism.

[0019] In view of this, this application proposes a dynamic adaptive data stream acceleration transmission method for the domestic IT innovation environment: First, the capabilities of the central processing unit, operating system, network interface, and cryptographic acceleration unit in the domestic IT innovation computing node are probed, and multi-dimensional data representing computing capabilities, network transmission and reception capabilities, and cryptographic processing capabilities are organized into a capability profile for the domestic IT innovation environment; Second, when a data stream transmission request is received, the service type, data scale, timeliness level, and security level of the data stream are analyzed, and a data stream profile is generated in combination with preset classification rules. During the data stream transmission process, the operational status information of the network layer, system layer, application layer, and cryptographic processing layer is collected across layers to construct an operational status vector; Third, based on... The system generates capability profiles, data flow profiles, and operational state vectors for the domestic IT innovation environment. It then executes a policy decision-making process that includes rule mapping and learning correction to generate a set of transmission policy parameters that match the current environment and data flow characteristics. Based on these parameters, it accelerates data flow transmission in stages such as fragmentation processing, connection and queue configuration, cryptographic processing pipeline, and transmission scheduling. Finally, during transmission, it collects transmission performance indicators corresponding to the transmission policy parameter set, constructs a policy performance sample library, and continuously updates the transmission policy parameter set through a policy update model. The updated policy is then used to control the transmission of subsequent data flows with the same or similar data flow profiles, forming a closed-loop adaptive acceleration mechanism for the domestic IT innovation environment.

[0020] Through the above technical solutions, this application can achieve dynamic adaptive configuration and iterative optimization of transmission strategies by utilizing the capability profiles, data flow profiles, and cross-layer operation state vectors of domestically developed computing nodes without changing the existing network infrastructure. This will improve the utilization rate of hardware and software resources in the domestically developed environment, increase the transmission efficiency of multi-type business data flows, and reduce the workload of manual tuning of transmission parameters.

[0021] The technical solution of this application will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0022] Figure 1 This is a flowchart illustrating a dynamic adaptive data stream acceleration transmission method for the domestic IT innovation environment provided in this application embodiment. Figure 1 As shown, this dynamic adaptive data stream acceleration transmission method for the information technology innovation environment can specifically include:

[0023] S101 performs capability detection on the central processing unit, operating system, network interface and cryptographic acceleration unit in the domestic innovation computing node to obtain environmental capability information that characterizes computing capability, network transceiver capability and cryptographic processing capability, and organizes the environmental capability information into a domestic innovation environment capability profile.

[0024] S102, upon receiving a data stream transmission request, the service type, data size, timeliness level, and security level of the data stream are parsed, and a data stream profile corresponding to the data stream is generated by combining preset classification rules, and a stream identifier is assigned to the data stream to identify the data stream profile.

[0025] S103, during data flow transmission, cross-layer acquisition of operational status information corresponding to the capability profile and data flow profile of the information technology innovation environment, and organization of operational status information into operational status vectors;

[0026] S104, based on the capability profile, data flow profile and operation status vector of the information technology innovation environment, execute the preset strategy decision process to determine the set of transmission strategy parameters for data flow;

[0027] S105, according to the transmission strategy parameter set, the data stream corresponding to the stream identifier is processed by fragmentation, connection and queue configuration, password processing configuration and transmission scheduling, and the data fragments are sent according to the fragmentation configuration parameters and concurrent transmission configuration parameters in the transmission strategy parameter set;

[0028] S106: During the transmission process, the transmission performance indicators corresponding to the transmission strategy parameter set are collected, the transmission performance indicators are associated with the transmission strategy parameter set, and the transmission strategy parameter set is updated based on the transmission performance indicators. The updated transmission strategy parameter set is then used for accelerated transmission control of subsequent data streams with the same or similar data stream profiles.

[0029] In some embodiments, capability probes are performed on the central processing unit, operating system, network interface, and cryptographic acceleration unit in the domestically developed computing node to obtain environmental capability information characterizing computing capabilities, network transceiver capabilities, and cryptographic processing capabilities, including:

[0030] Obtain hardware description information and operating system kernel status information of the domestically developed computing node, determine the number of cores, clock frequency, cache configuration and operating system scheduling policy parameters of the central processing unit, and obtain configuration information to characterize the basic characteristics of computing capabilities;

[0031] Pre-defined computational probing tasks and cryptographic operation probing tasks are executed on the central processing unit and cryptographic acceleration unit. The counts of instruction execution, storage access, and cryptographic operation completion are collected. Combined with configuration information, multi-dimensional capability measurement data representing computational and cryptographic processing capabilities are calculated.

[0032] Preset network transmit / receive probe data packets are sent to the network interface. Under different message lengths and different queue configurations, transmit / receive latency, packet drop counts, and queue length changes are collected. Multidimensional capability measurement data characterizing network transmit / receive capabilities are then calculated.

[0033] The multidimensional capability measurement data representing computing power, network transmission and reception capabilities, and cryptographic processing capabilities are normalized and encoded to generate environmental capability information.

[0034] Specifically, in this embodiment, the capability detection module first obtains the hardware description information and operating system kernel status information of the domestically developed computing node. The hardware description information includes the number of cores in the central processing unit, its clock speed parameters, multi-level cache capacity and hierarchy, total memory capacity, etc.; the operating system kernel status information includes the currently used scheduling policy type, scheduling time slice configuration, interrupt handling policy, and network protocol stack configuration parameters, etc. The capability detection module extracts the number of cores, clock speed, cache configuration, and operating system scheduling policy parameters of the central processing unit by reading the hardware description interface and kernel status interface provided by the system, and combines them to form configuration information used to characterize the basic characteristics of computing capabilities, such as recorded as structured fields such as "number of cores = 16, clock speed = 2.6GHz, total capacity of L3 cache = 20MB, scheduling policy is a hybrid of time-slice round-robin and completely fair".

[0035] After acquiring basic configuration information, the capability detection module executes preset computational detection tasks and cryptographic operation detection tasks on the central processing unit and cryptographic acceleration unit, respectively. The computational detection tasks may include arithmetic operation tasks mainly involving integer operations, matrix approximation calculation tasks mainly involving floating-point operations, and sequential read / write and random read / write tasks mainly involving memory access. By running the above tasks within a limited time window, the module collects the instruction execution count, cache access count, and memory read / write access count of the central processing unit within that time window.

[0036] The cryptographic operation probing task invokes a preset domestic cryptographic algorithm on the cryptographic acceleration unit or central processing unit, such as performing multiple rounds of encryption and decryption operations on a fixed-length data block, and counts the number of cryptographic operation batches completed and the number of failures per unit time. The capability probing module combines the above-mentioned count data with the aforementioned configuration information to calculate multi-dimensional capability measurement data characterizing computing power and cryptographic processing capabilities, which may include multiple dimensions such as the number of valid instruction executions per unit time, the number of cache accesses per unit time, the number of cryptographic data block processing times per unit time, and the average length of the cryptographic operation queue.

[0037] Subsequently, the capability detection module sends preset network transmit / receive probe data packets to the network interface of the domestically developed computing node to evaluate its network transmit / receive capabilities. In this embodiment, multiple sets of probe packets with different message lengths are set, such as probe packets with lengths of 512 bytes, 4KB, and 64KB, and transmission and reception tests are performed under different queue configurations. These different queue configurations include single-queue transmit / receive mode, multi-queue parallel transmit / receive mode, and queue configuration mode bound to different processing kernels.

[0038] During the test, the capability detection module records the transmission and reception latency, round-trip latency, number of dropped packets, and length changes of each receiving and sending queue for various detection messages. By statistically analyzing and calculating these raw data, multi-dimensional capability metrics representing network transmission and reception capabilities are obtained, such as average latency under different message lengths, jitter range, peak throughput under different queue modes, and stable queue length range.

[0039] Furthermore, after obtaining multi-dimensional capability measurement data representing computing power, network transmission and reception capabilities, and cryptographic processing capabilities, the capability detection module performs normalization and encoding aggregation on the data of each dimension. Normalization can map capability measurement data of different dimensions to a unified numerical range based on preset upper and lower limits of capability indicators. For example, the number of instruction executions per unit time, the number of cryptographic operation batches, and network throughput are converted into dimensionless values ​​between 0 and 1 according to their respective reference ranges to eliminate dimensional differences between different indicators.

[0040] During the encoding and aggregation process, the capability detection module combines normalized computing capability indicators, network capability indicators, and cryptographic processing capability indicators into a structured capability vector according to preset capability dimensions. Based on this capability vector, it further generates hierarchical labels or capability level identifiers. For example, the computing capability vector is mapped to an identifier indicating "high computing capability level," the network transceiver capability vector is mapped to an identifier indicating "medium network throughput and high latency stability," and the cryptographic processing capability vector is mapped to an identifier indicating "available and high-performance cryptographic hardware acceleration capability." Finally, the capability detection module encodes the aforementioned capability vectors and corresponding capability level identifiers into environmental capability information and associates this information with the unique identifier of the domestic IT computing node, thus forming a domestic IT environment capability profile that can be directly accessed by the subsequent policy decision-making module.

[0041] Through the implementation of this embodiment, the system can automatically complete the capability detection of the central processing unit, operating system, network interface and cryptographic acceleration unit during the initialization stage of the domestic IT computing node. It transforms the originally discrete performance count data into structured environmental capability information, which helps to improve the accuracy and comparability of capability characterization. This provides a reliable basis for subsequent data flow transmission strategy decisions based on the domestic IT environment capability profile, thereby reducing the workload of manual configuration and laying the foundation for dynamic adaptive acceleration strategies.

[0042] In some embodiments, the business type, data scale, timeliness level, and security level of the data stream are parsed, and a data stream profile corresponding to the data stream is generated by combining preset classification rules, including:

[0043] Extract business attribute parameters and security attribute parameters from the business identifier field, data length field, time constraint field, and security policy identifier field in the data stream transmission request, and construct the original feature set by combining them with the historical transmission records corresponding to the data stream;

[0044] The original feature set is standardized and encoded, and a first feature vector representing the business type, data scale, timeliness level, and security level is generated using a pre-defined multi-dimensional feature representation model.

[0045] The first feature vector is compared with multiple reference feature vectors in the pre-built data flow profile reference library for similarity measurement. Based on the similarity measurement results, the classification label and hierarchical mark of the data flow are determined according to the preset classification rules.

[0046] The classification label, hierarchical marker, and first feature vector are combined and encoded to generate a data stream profile.

[0047] Specifically, when a business system initiates a data stream transmission request, the transmission request message includes a business identifier field, a data length field, a time constraint field, and a security policy identifier field. For example, the business identifier of a request can indicate a type identifier such as "real-time approval message," "batch document distribution," "log collection," or "database incremental synchronization." The data length field indicates the estimated total amount of data to be transmitted, the time constraint field indicates the maximum allowable end-to-end latency or completion time limit, and the security policy identifier field indicates security requirements such as whether end-to-end encryption with national cryptographic standards is mandatory, whether link-level encryption is allowed, and whether integrity verification is required.

[0048] The data flow profile generation module first extracts business attribute parameters and security attribute parameters from the above fields. Based on this, it further queries the historical transmission records of the business identifier and the target node recorded in the system, such as historical average file size, historical average transmission rate, historical number of failed retransmissions, and historical encryption methods. The business attribute parameters, security attribute parameters and historical statistical parameters of this request are combined to form the original feature set.

[0049] After constructing the initial feature set, the data flow profiling generation module standardizes and encodes features of different dimensions. For continuous numerical features such as data length, maximum allowable latency, and historical average transmission rate, interval scaling or logarithmic transformation can be used to map them to a unified numerical range. For discrete categorical features such as business identifiers, security policy identifiers, and encryption levels, single-class encoding or embedded vector encoding can be used to map different business types and security policy types to fixed-length real-valued vectors.

[0050] Furthermore, after completing numerical standardization and encoding, the module inputs the obtained features into a pre-defined multi-dimensional feature representation model. This model can employ a multi-layer fully connected network or a lightweight autoencoder structure to perform non-linear combination and dimensionality reduction on the input features, outputting a first feature vector. This first feature vector simultaneously characterizes the comprehensive features of the data stream, such as its business type, data scale, timeliness level, and security level. In actual deployment, this first feature vector can be selected as a vector of several fixed dimensions, such as 32-dimensional or 64-dimensional, to balance expressive power and computational overhead.

[0051] To map the first feature vector of the current data flow to predefined business categories and levels within the system, the data flow profile generation module maintains a data flow profile reference library. This library pre-stores multiple reference feature vectors and their corresponding standard business labels and level markers. The reference feature vectors can be obtained before system deployment through sample collection and offline training on typical business scenarios. For example, a set of representative feature vectors can be extracted for each business type, such as "real-time approval message type," "batch document transmission type," "continuous log collection type," and "periodic database synchronization type," and each type can be assigned a timeliness level, high / low priority, and security level marker.

[0052] After generating the first feature vector, the data flow profiling module measures the similarity between this vector and each reference feature vector in the reference library. Similarity measurement can use cosine similarity, inverse Euclidean distance, or other similarity functions based on vector distance. The module selects one or more reference feature vectors with the highest similarity or falling within a certain similarity threshold range according to preset classification rules. Based on the standard business labels and level markers corresponding to these reference feature vectors, it determines the classification label and level marker for the current data flow, such as "Real-time Control - High Priority - High Security Level" or "Batch Files - Medium Priority - High Security Level," etc.

[0053] After determining the category label and hierarchical marker, the data flow profile generation module combines and encodes the category label, hierarchical marker, and first feature vector to generate the final data flow profile. In one implementation, the module can encode the category label and hierarchical marker into fixed-length discrete marker vectors, concatenate them with the first feature vector to form an extended feature vector, and use this extended feature vector as the core content of the data flow profile, associating it with the flow identifier of the data flow for storage.

[0054] In this way, for multiple transmission requests of the same business type and similar data scale, since the similarity between their first feature vector and the reference feature vector is close, the corresponding data flow profiles also have high similarity. Thus, in the subsequent strategy decision module, the set of transmission strategy parameters that have performed well in the past can be quickly selected by recognizing the data flow profiles.

[0055] For example, in a real-world application scenario, the headquarters approval system sends real-time approval messages to branches across the country. Each message has a small data size, strict end-to-end latency requirements, and requires end-to-end encryption. For this type of request, the data flow profile generation module extracts the business identifier "real-time approval message" from the request message. The data length field reflects the small amount of data transmitted at once, the time constraint field indicates a short maximum allowable delay, and the security policy field indicates that encryption using domestic cryptographic algorithms is mandatory. Based on historical transmission records, the module finds that this type of business typically occurs frequently during working hours, with few retransmissions in historical transmissions. After standardization and encoding, the first feature vector output by the multi-dimensional feature representation model has a high similarity to the "real-time control class - high priority - high security level" reference vector in the reference library. The system labels this data flow as real-time control according to classification rules, with a high timeliness level and a high security level. The data flow profile generated after combined encoding will be identified by the subsequent policy decision module as a high-priority data flow requiring priority latency protection, limited fragment length, and dedicated queue scheduling.

[0056] In some embodiments, cross-layer acquisition of operational status information corresponding to the capability profile and data flow profile of the information technology innovation environment is performed, and the operational status information is organized into an operational status vector, including:

[0057] Based on the capability profile and data flow profile of the information technology innovation environment, network layer monitoring indicators, system layer monitoring indicators, application layer monitoring indicators and cryptographic processing monitoring indicators corresponding to the target data flow are selected from the preset monitoring indicator library to generate cross-layer monitoring configuration;

[0058] According to the cross-layer monitoring configuration, the operating status information of the target data stream in the current transmission stage is obtained by combining time window sampling and event-triggered sampling, forming multi-dimensional operating status observation data;

[0059] The multidimensional operational status observation data is aggregated and normalized over time, and the network state sub-vector, system state sub-vector, application state vector and cryptographic processing state vector are generated by using a preset multidimensional time series feature extraction model.

[0060] A feature fusion algorithm based on attention weighting mechanism is used to fuse and encode the network state sub-vector, system state sub-vector, application state sub-vector, and cryptographic processing state sub-vector to obtain the running state vector.

[0061] Specifically, the system first pre-builds a monitoring indicator library, which stores various observable indicators categorized by network layer, system layer, application layer, and cryptographic processing layer. For example, network layer monitoring indicators include network round-trip latency, one-way latency, packet drop count, retransmission count, length of each sending and receiving queue, and the amount of data successfully sent and received per unit time; system layer monitoring indicators include overall utilization of the central processing unit, utilization of processing kernels allocated to transmission threads, memory usage ratio, and number of context switches; application layer monitoring indicators include the length of the sending buffer queue corresponding to a certain flow identifier, the length of the receiving buffer queue, the number of messages to be processed at the application end, and the latency of application-level acknowledgment feedback; cryptographic processing monitoring indicators include the queue length of cryptographic computation tasks, the number of encrypted data blocks processed per unit time, the average single-block encryption latency, and the utilization rate of cryptographic acceleration units.

[0062] When a target data stream begins transmission, the cross-layer operational status acquisition and fusion module takes the data stream's information technology innovation environment capability profile and data stream profile as input. Based on the data stream's business type, timeliness level, and security level, it selects the most relevant set of indicators from the monitoring indicator library to generate a cross-layer monitoring configuration. For example, for the aforementioned high-timeliness data stream of "real-time approval messages," the monitoring configuration will focus on network latency, sending queue length, transmission thread load, and cryptographic operation latency; while for the high-volume data stream of "batch document distribution," the monitoring configuration will emphasize throughput indicators and queue length stability indicators.

[0063] After generating the cross-layer monitoring configuration, this embodiment uses a combination of time window sampling and event-triggered sampling to collect the operational status information of the target data stream during the current transmission phase. Time window sampling refers to periodically collecting the instantaneous values ​​or statistical values ​​within a window of the selected monitoring indicators at preset time intervals, such as collecting the average network round-trip latency, average queue length, and central processing unit utilization every 100 milliseconds. Event-triggered sampling, on the other hand, immediately triggers an additional sampling when a sudden change in a key indicator or exceeding a preset threshold is detected. For example, if a sending queue length suddenly exceeds a set upper limit, the network round-trip latency continues to increase in a short period of time, or the cryptographic operation queue length shows a significant increase, a fine-grained indicator snapshot is immediately recorded.

[0064] Through the two methods described above, the cross-layer operation status acquisition and fusion module generates multi-dimensional operation status observation data over a time series during a transmission phase of the target data stream. This observation data includes both stable changes recorded at fixed time steps and encrypted records of abnormal fluctuations.

[0065] Subsequently, this embodiment performs time series aggregation and normalization processing on the generated multidimensional operational status observation data. Time series aggregation can calculate the statistical characteristics of each monitoring indicator within each sliding time window, such as the average, maximum, and minimum values ​​within the window, as well as the rate of change compared to the previous window, thereby compressing the original high-frequency sampling data into a sequence feature that is representative in the time dimension.

[0066] Meanwhile, in order to eliminate the inconsistency in numerical scale caused by differences in the dimensions of different indicators and the differences in capabilities between different domestic IT innovation computing nodes, the system maps the aggregated results of each monitoring indicator to a unified numerical range based on the capability range recorded in the domestic IT innovation environment capability profile and the historical statistical baseline. For example, by using linear normalization or nonlinear scaling, indicators such as network latency, queue length, and utilization rate are uniformly converted into dimensionless values ​​between 0 and 1.

[0067] After aggregation and normalization, the cross-layer operation status acquisition and fusion module inputs different categories of monitoring data into a preset multi-dimensional time series feature extraction model. The model can use a lightweight neural network structure designed for time series or other algorithm structures with time series pattern extraction capabilities, and outputs four sub-vectors to represent the network state, system state, application state, and cryptographic processing state, namely the network state sub-vector, system state sub-vector, application state sub-vector, and cryptographic processing state sub-vector.

[0068] For example, in real-world testing scenarios, network state sub-vectors can highlight the current latency fluctuation pattern and packet loss trend, system state sub-vectors can reflect whether the processing kernel is close to saturation, application state sub-vectors can reflect whether the sending and receiving buffers are piling up, and cryptographic processing state sub-vectors can reflect whether there is significant queuing in the cryptographic operation channel.

[0069] After obtaining the above four types of state sub-vectors, this embodiment uses a feature fusion algorithm based on an attention weighting mechanism to fuse and encode them to obtain a single running state vector. Specifically, the cross-layer running state acquisition and fusion module assigns a set of learnable attention weight coefficients to each type of state sub-vector. The calculation of attention weights can be based on the content features of the state sub-vector itself, or it can be combined with the business type and timeliness level information in the corresponding data flow profile.

[0070] For example, for real-time approval message data streams, the attention mechanism will automatically increase the weights of the network state sub-vector and the application state sub-vector, while giving appropriate but relatively low weights to the system state sub-vector and the cryptographic processing state sub-vector; for batch document distribution data streams, the weights of the system state sub-vector and the cryptographic processing state sub-vector may be emphasized more.

[0071] After calculating the attention weights of each sub-vector, the system performs weighted summation or concatenation transformation on each sub-vector according to the weights to generate an operational state vector that uniformly represents the overall operational state of the current transmission stage. This operational state vector has a fixed length and contains key temporal feature information of the network layer, system layer, application layer, and cryptographic processing layer. It is associated with the flow identifier of the data stream and is used as input by the policy decision module to dynamically adjust the set of transmission policy parameters.

[0072] For example, in a sample scenario, when a domestically developed computing node simultaneously processes real-time approval message data streams and batch document distribution data streams, for the real-time approval message data stream, the cross-layer operation status acquisition and fusion module frequently collects network latency and application queue status within a short time window. When network latency increases or application queues accumulate, it significantly enhances the influence of network state sub-vectors and application state sub-vectors in the fusion process through an attention mechanism, making the obtained operation status vector more reflective of latency risks. For the batch document distribution data stream, the module focuses on system-level CPU utilization and password processing queue length, increasing the attention weight of system state sub-vectors and password processing state sub-vectors, making the operation status vector more sensitive to overall resource utilization.

[0073] In some embodiments, based on the capability profile of the domestic IT innovation environment, the data flow profile, and the operational state vector, a preset strategy decision-making process is executed to determine the set of transmission strategy parameters for the data flow, including:

[0074] Based on the capability profile and data flow profile of the information technology innovation environment, candidate strategy templates that match the current capability characteristics and data flow profile characteristics of the information technology innovation environment are selected from the preset strategy template library, and an initial strategy parameter set is generated based on the candidate strategy templates.

[0075] The running state vector and the initial policy parameter set are used as joint inputs. The initial policy parameter set is adjusted and calculated using a pre-defined multi-layer parameter mapping model to obtain an intermediate policy parameter set that represents the target policy value in the current transmission stage. The multi-layer parameter mapping model includes a rule-based mapping sub-model and a learning-based parameter correction sub-model.

[0076] Based on preset policy constraint rules and security compliance rules, the intermediate policy parameter set is constrained, verified, and corrected. Policy values ​​that do not match the capability profile of the information technology innovation environment are removed, and policy values ​​that violate security compliance rules are adjusted to generate a transmission policy parameter set that meets the constraint conditions.

[0077] Specifically, a policy template management module is pre-built during system initialization, and a preset policy template library is maintained in the policy template management module. The policy template library stores a variety of pre-configured policy templates for typical information technology innovation environment capability combinations and typical data flow profile types. Each candidate policy template contains parameter baselines for multiple aspects such as fragmentation configuration, concurrent transmission, retransmission and redundancy control, queue scheduling, and cryptographic processing path selection.

[0078] For example, one type of template can be configured for the capability profile of the domestic IT innovation environment ("high computing power, medium network capability, and strong cryptographic processing capability") and the data flow profile ("batch file type, medium priority, high security level"). Another type of template can be configured for real-time messaging services ("network latency sensitive, small message size"). When the current target data flow arrives, the strategy decision module first reads the capability profile of the domestic IT innovation computing node and obtains the profile information of the data flow from the data flow profile generation module. Then, the two are used as search conditions to match in the preset strategy template library. Based on the preset similarity measurement rules, one or more candidate strategy templates that match the current domestic IT innovation environment capability characteristics and data flow profile characteristics are selected.

[0079] During the matching process, the module can prioritize key fields such as business type, timeliness level, and security level, and combine them with the computing power level, network transmission and reception capability level, and cryptographic processing capability level in the domestic IT environment capability profile to perform a comprehensive score, selecting the candidate policy template with the highest score as the template baseline for this decision. Subsequently, the policy decision module generates an initial policy parameter set based on the pre-configured parameters in the selected candidate policy template. This initial policy parameter set records the initial values ​​of various transmission parameters in a structured form, such as initial fragment length, initial concurrent connection count, initial retransmission window size, whether to enable forward error correction, and whether cryptographic operations prioritize hardware or software paths.

[0080] Furthermore, after generating the initial policy parameter set, this embodiment uses the running state vector and the initial policy parameter set as joint inputs, which are then processed by a multi-layer parameter mapping model for adjustment calculations. This multi-layer parameter mapping model consists of two parts: a rule-based mapping sub-model and a learning-based parameter correction sub-model. The rule-based mapping sub-model is used to establish a deterministic mapping relationship between the running state vector and the policy parameter set, while the learning-based parameter correction sub-model is used to perform fine-grained corrections on the rule mapping results based on historical transmission performance and policy performance samples.

[0081] For example, in some specific implementations, the rule-based mapping sub-model adjusts the initial policy parameters according to the interval positions of some key indicators in the running state vector. When the running state vector reflects that the current network round-trip latency is continuously increasing and the packet loss rate is close to the preset upper limit, the sub-model will reduce the number of concurrent connections in the policy parameter set, reduce the sending window, and appropriately increase the retransmission waiting time. When the running state vector indicates that the current central processing unit utilization is close to saturation and the cryptographic processing queue length is long, the sub-model will reduce the compression level and reduce the depth of the additional encryption pipeline to avoid further occupying processing resources.

[0082] The learning-based parameter correction sub-model learns the correspondence between various strategy parameter combinations and transmission performance feature vectors under similar environments and data flow profiles from a historical strategy performance sample library. Based on the rule-based mapping results, it slightly adjusts the specific values ​​of each parameter to approximate the parameter combinations that performed better in history.

[0083] For example, for real-time approval message data streams, the correction sub-model can combine historical samples to further fine-tune the fragment length from the medium value suggested by the rules to a smaller value, in order to achieve lower latency jitter. After the above two-level processing, the multi-layer parameter mapping model outputs an intermediate policy parameter set, which has comprehensively considered the current domestic IT innovation environment capabilities, data stream characteristics, and the current operational status of the transmission stage.

[0084] Furthermore, to ensure that the intermediate policy parameter set is executable and meets security and compliance requirements in a real-world IT innovation environment, this embodiment also includes a policy constraint and compliance verification unit in the policy decision module. The policy constraint and compliance verification unit verifies each item in the aforementioned intermediate policy parameter set according to preset policy constraint rules and security compliance rules. Policy constraint rules include constraints consistent with the capability profile of the IT innovation environment, such as prohibiting the use of zero-copy transmission policies on operating systems that do not support zero-copy, prohibiting the configuration of multi-queue binding policies on network interfaces that do not have multi-queue capabilities, and prohibiting the configuration of parameter combinations that must rely on hardware cryptographic paths on nodes where the cryptographic acceleration unit is unavailable.

[0085] Security compliance rules are related to business security levels and domestic IT innovation security specifications. For example, under a high-security data flow profile, it is not allowed to disable end-to-end encryption, reduce the strength of cryptographic algorithms, or use uncertified compression algorithm paths. The policy constraint and compliance verification unit checks the value of each parameter in the intermediate policy parameter set. Policy values ​​that do not conform to the domestic IT innovation environment capability profile are directly removed or reverted to the default security value. Policy values ​​that violate security compliance rules are adjusted or replaced with compliant alternative parameters.

[0086] During the verification process, if a combination conflict is found between multiple parameters, such as simultaneously requiring a high number of concurrent connections and extremely low central processing unit utilization, the module can also adjust the relevant parameters in a coordinated manner through preset conflict resolution rules to ensure that the overall policy configuration remains internally consistent under performance and security constraints. After the above constraint verification and correction, the policy decision module finally generates a set of transmission policy parameters that meet the constraints. This set of transmission policy parameters serves as the direct input for subsequent fragmentation processing, connection and queue configuration, cryptographic processing configuration, and transmission scheduling.

[0087] In some specific examples, when the domestic computing node processes the aforementioned real-time approval message data stream and batch document distribution data stream simultaneously, for the real-time approval message data stream, the strategy decision module selects a real-time control template that prioritizes low latency from the strategy template library as a candidate strategy template based on its data stream profile and current running state vector. The generated initial strategy parameter set has a small fragment length, a moderate number of concurrent connections, and prioritizes the configuration of independent transmission threads and queues.

[0088] When the running state vector reflects a high current network load, the multi-layer parameter mapping model reduces the number of concurrent connections and slightly narrows the sending window through a rule-based mapping sub-model, avoiding further amplification of queuing latency on congested links. Simultaneously, the learning-based parameter correction sub-model fine-tunes the cryptographic processing pipeline depth to a smaller value by referencing historical samples, reducing latency fluctuations. The policy constraint and compliance verification unit ensures that the end-to-end encryption policy is always enabled and uses compliant domestic cryptographic algorithms during the verification process, ultimately generating a set of transmission policy parameters that balances low latency and security.

[0089] For batch document distribution data streams, the strategy decision module selects a batch file template that prioritizes high throughput. When the running status vector indicates that the system still has idle resources and the network packet loss rate is low, it increases the number of concurrent connections through a multi-layer parameter mapping model and enables appropriate forward error correction coding. At the same time, it makes full use of the cryptographic acceleration unit in the cryptographic processing path selection. After constraint and compliance verification, the resulting set of transmission strategy parameters is more inclined to improve overall throughput and link utilization.

[0090] Through the implementation of this embodiment, the strategy decision-making module can quickly select suitable candidate strategy templates from the strategy template library under the comprehensive constraints of the capability profile, data flow profile, and operational status vector of the domestic IT innovation environment. It then generates a set of transmission strategy parameters tailored to the current environment and business characteristics through a multi-layer parameter mapping model and constraint verification mechanism. On one hand, this embodiment tightly couples static capability profiles, business profiles, and dynamic operational status into a unified decision-making process, avoiding the limitations of relying solely on fixed parameter configurations or single-level indicators for coarse optimization. On the other hand, by introducing a multi-layer decision-making structure combining template-driven, rule mapping, and learning correction, and superimposing verification of strategy constraint rules and security compliance rules, this embodiment can automatically complete the adjustment and optimization of transmission strategy parameters while meeting the security requirements of domestic IT innovation. This reduces manual configuration workload, improves the timeliness and rationality of strategy updates, and makes the accelerated transmission control of multiple types of data flows in the domestic IT innovation environment more refined, stable, and controllable.

[0091] In some embodiments, based on a set of transmission policy parameters, the data stream corresponding to the stream identifier is fragmented, configured for connections and queues, configured for cryptographic processing, and scheduled for transmission, including:

[0092] Based on the fragmentation configuration parameters in the transmission strategy parameter set, the original data stream corresponding to the stream identifier is segmented, and a stream identifier, sequence number mark and security attribute mark are added to each data fragment to generate a data fragment sequence to be sent.

[0093] Based on the concurrent transmission configuration parameters and connection configuration parameters in the transmission strategy parameter set, multiple transmission channels are established on the domestic computing node, the processing threads and network interface queues corresponding to each transmission channel are determined, the data fragment sequence to be sent is divided into multiple sub-sequences according to the preset mapping rules, and each sub-sequence is assigned to different transmission channels and network interface queues respectively.

[0094] Based on the cryptographic processing path configuration parameters in the transmission strategy parameter set, a hierarchical cryptographic processing pipeline is constructed. Based on the capability profile of the information technology innovation environment, the corresponding cryptographic operation is selected to be executed in the central processing unit or the cryptographic acceleration unit. The data fragments to be sent are queued and dequeued according to the depth and parallelism of the hierarchical cryptographic processing pipeline.

[0095] Based on the retransmission and redundancy control parameters and priority scheduling parameters in the transmission strategy parameter set, a transmission scheduling sequence is generated for multiple transmission channels and network interface queues. The queue scheduling algorithm is used to select the target data fragment from the ready data fragments to be sent. According to the transmission scheduling sequence, the transmission interface is called on the corresponding transmission channel to send the target data fragment to the target node.

[0096] Specifically, after the policy decision module outputs a set of transmission policy parameters for a specific target data stream, the fragmentation and marking module first reads the fragmentation configuration parameters from the transmission policy parameter set. These parameters include the target fragment length, alignment constraint rules, maximum batch fragment size, and security marking configuration method. In some examples, for batch document data streams, the target fragment length given in the policy parameters can be 64KB, and the alignment constraint requires that fragment boundaries be aligned as close as possible to the underlying storage page boundaries or file system block boundaries to reduce additional copy operations.

[0097] The fragmentation and tagging module sequentially segments the original data stream corresponding to the stream identifier according to the target fragment length. For the last data block that is shorter than the target fragment length, alignment padding can be performed based on whether padding rules are enabled in the policy parameters. For each generated data fragment, the fragmentation and tagging module appends the stream identifier corresponding to the data stream, a monotonically increasing sequence number, and a security attribute tag generated based on the data stream profile to the fragment header. The security attribute tag indicates whether the fragment needs encryption, whether integrity verification information needs to be appended, and the corresponding security level classification. Through the above processing, a structured sequence of data fragments to be sent is formed, and each fragment can be uniquely identified and tracked by subsequent modules.

[0098] Furthermore, after constructing the sequence of data fragments to be sent, this embodiment establishes multiple transmission channels on the domestically developed computing node based on the concurrent transmission configuration parameters and connection configuration parameters in the transmission strategy parameter set. Connection configuration parameters may include the transmission protocol type used for each transmission channel, whether a long-connection mechanism is enabled, and the congestion control mode corresponding to each channel; concurrent transmission configuration parameters may include the number of concurrent channels and the maximum number of unacknowledged fragments allowed per channel. The connection management module creates several transmission channels in the operating system or user-space protocol stack based on these parameters, and, combined with the network interface multi-queue capabilities and the number of central processing unit cores recorded in the domestically developed environment capability profile, maps each transmission channel to different processing threads and network interface queues.

[0099] For example, in a domestically developed node equipped with multi-queue network interface cards (NICs) and multi-core processors, a dedicated low-latency channel can be allocated to high-priority real-time message data streams and bound to a separate network interface queue and processing kernel. Three high-throughput channels can be allocated to batch document data streams and bound to the remaining network interface queues. Subsequently, the connection management module divides the data fragment sequence to be sent into multiple sub-sequences according to preset mapping rules. These preset mapping rules can employ round-robin allocation, hash allocation based on sequence number, or dynamic allocation based on priority and queue load. Taking round-robin allocation as an example, the first fragment is allocated to channel 1, the second fragment to channel 2, and so on, ensuring a roughly balanced load across channels. Taking priority mapping rules as an example, fragments marked as high priority can be preferentially allocated to lower-latency channels, while fragments marked as low priority or those with delay capabilities can be allocated to channels with high throughput.

[0100] Regarding cryptographic processing configuration, this embodiment constructs a hierarchical cryptographic processing pipeline based on cryptographic processing path configuration parameters in the transmission strategy parameter set. The cryptographic processing path configuration parameters indicate whether, given the current capability profile of the domestic IT innovation environment, software cryptographic operations should be prioritized on the central processing unit or hardware cryptographic operations should be prioritized by calling the cryptographic acceleration unit, and provide the pipeline's hierarchical structure and parallelism configuration.

[0101] For example, in a domestic IT innovation node equipped with a high-performance cryptographic acceleration unit, for batch file data streams with high security levels, the policy parameters can specify a three-layer cryptographic processing pipeline: In the preprocessing stage, fragments are padded and grouped, merging several small fragments into data blocks suitable for hardware computation; in the cryptographic computation stage, the data blocks are encrypted using domestic cryptographic algorithms and integrity tags are generated within the cryptographic acceleration unit; in the post-processing stage, the encrypted results and security tags are assembled back into the fragment structure. The pipeline scheduling module, based on the pipeline depth and parallelism configuration given in the transmission policy parameter set, establishes corresponding task queues for each layer and allocates an appropriate number of execution threads or hardware queues to each layer according to the capability profile of the central processing unit and cryptographic acceleration unit in the domestic IT innovation environment capability profile.

[0102] In practice, fragments first enter a preprocessing queue before entering the transmission queue. After preprocessing, they are submitted in batches to the computation queue corresponding to the cryptographic acceleration unit. After encryption and integrity processing, they enter the post-processing queue. Finally, fragments that have been processed and have security attribute tags and integrity labels are put into the transmission queue. Through queuing and dequeueing control of each queue layer, the pipeline scheduling module can make full use of cryptographic acceleration resources without blocking the overall transmission process.

[0103] Furthermore, during the transmission scheduling phase, this embodiment generates transmission scheduling sequences for multiple transmission channels and network interface queues based on retransmission and redundancy control parameters and priority scheduling parameters in the transmission strategy parameter set. The retransmission and redundancy control parameters may include a time threshold for fragmentation timeout retransmission, a maximum number of retransmissions, whether forward error correction redundancy fragmentation is enabled, and the redundancy ratio, etc.; the priority scheduling parameters may include the priority weights of data fragments of different service levels, whether strict priority scheduling or weighted fair scheduling is adopted, etc.

[0104] The sending scheduling module combines these parameters to construct a local scheduling sequence for each transmission channel, while maintaining a cross-channel scheduling strategy globally. For example, for real-time approval message data streams, the sending scheduling module can use a strict priority queue, placing its fragments to be sent in a high-priority ready queue. As long as there are ready fragments in the high-priority queue, the fragments in that queue are selected for sending first. For batch document distribution data streams, a weighted fair queue strategy can be used to ensure continuous transmission at a high throughput within the remaining bandwidth.

[0105] During actual transmission, the transmission scheduling module periodically checks the status of each network interface queue, selects the target data fragment from the ready data fragments according to the scheduling sequence, and calls the underlying transmission interface on the corresponding transmission channel to send the target data fragment. At the same time, it registers the transmission time and sequence number of the fragment in the retransmission management structure. If no acknowledgment is received or loss is detected within a preset time, the corresponding fragment is re-queued into the transmission queue according to the retransmission strategy.

[0106] In some specific example scenarios, when a domestically developed computing node simultaneously processes a real-time approval message data stream and a batch document distribution data stream, for the real-time approval message data stream, the strategy parameter set can be set to a smaller fragment length, a lower number of concurrent channels, and a high-priority scheduling flag. The fragmentation module splits the short message into several small fragments and attaches a high-priority security tag. The connection and queue configuration module creates a dedicated channel for it and binds it to a low-latency network interface queue. The password processing pipeline chooses to perform fast encryption at a shallow pipeline depth on the central processing unit. The sending scheduling module adopts a strict priority queue algorithm to ensure that whenever there is a ready fragment for the data stream, it is sent first.

[0107] For batch document distribution data streams, the strategy parameter set can be set to a larger fragment length, a higher number of concurrent channels, and enable forward error correction redundant fragmentation. Thus, the fragmentation module splits large files into multiple larger fragments, the connection and queue configuration module allocates multiple high-throughput channels for them and distributes the load across multiple network interface queues, the password processing pipeline prioritizes performing large-scale encryption processing in the password acceleration unit, and the sending scheduling module adopts a weighted fair queue algorithm to fully utilize the remaining bandwidth to transmit document fragments while ensuring the low latency requirements of real-time approval message data streams.

[0108] In some embodiments, during transmission, transmission performance metrics corresponding to the transmission strategy parameter set are collected, and the transmission performance metrics are associated with the transmission strategy parameter set, including:

[0109] Select throughput, latency, retransmission and redundancy usage, processing thread load, and cryptographic operation usage from the preset performance monitoring indicator library to generate a performance monitoring configuration corresponding to the transmission strategy parameter set.

[0110] According to the performance monitoring configuration, during the execution of the data stream fragmentation sending, confirmation feedback receiving and password processing pipeline, multi-dimensional transmission performance observation data corresponding to the stream identifier is collected in a sliding time window manner.

[0111] The multidimensional transmission performance observation data is normalized and outlier filtered out. The processed multidimensional transmission performance observation data is then converted into a transmission performance feature vector that characterizes the performance characteristics of the current transmission stage using a pre-defined multidimensional feature mapping model.

[0112] The transmission performance feature vector is combined and encoded with the corresponding transmission strategy parameter set, the information technology innovation environment capability profile, and the data flow profile to generate a strategy performance association record, which is then stored in the strategy performance sample library.

[0113] Specifically, when a target data stream enters the transmission phase, the performance monitoring module first selects applicable monitoring indicators from a pre-defined performance monitoring indicator library based on the data stream's profile information and the current transmission strategy parameter set, generating a performance monitoring configuration corresponding to the transmission strategy parameter set. The performance monitoring indicator library predefines various types of performance indicators, including throughput indicators such as the amount of data successfully transmitted per unit time, the amount of data acknowledged per unit time, and bandwidth utilization; latency indicators such as round-trip latency statistics for fragment transmission to acknowledgment, latency jitter range, and end-to-end completion time; retransmission and redundancy utilization indicators such as the number of fragment retransmissions, retransmission ratio, and redundant fragment ratio; processing thread load indicators such as the proportion of processing time occupied by the transmission threads allocated to the data stream and context switching frequency; and cryptographic operation utilization indicators such as the queue length of each layer of the cryptographic processing pipeline, the number of data blocks processed per unit time, and the utilization rate of cryptographic acceleration units.

[0114] After generating the performance monitoring configuration, this embodiment, according to the performance monitoring configuration, collects multi-dimensional transmission performance observation data corresponding to the data stream identifier using a sliding time window during the data stream fragmentation transmission, acknowledgment reception, and cryptographic processing pipeline execution. The sliding time window can be set, for example, every 200 milliseconds, and the window moves continuously on the time axis, with partial overlap between adjacent windows. Within each window, the performance monitoring module collects various performance observations from the transmission scheduling module, acknowledgment reception processing module, and cryptographic processing pipeline scheduling module, such as the number of successfully transmitted fragments, the number of received acknowledgments, measured round-trip latency samples, the number of retransmission events within the window, the average and maximum length of the cryptographic pipeline queue, and the average utilization of the transmission threads allocated to the data stream within that window.

[0115] For real-time approval message data streams, the module focuses on recording the maximum round-trip latency and latency jitter in each window. For batch document distribution data streams, it focuses on recording the effective throughput and retransmission ratio in each window. By continuously collecting data through sliding time windows, a multi-dimensional transmission performance observation data sequence covering the entire transmission phase is formed.

[0116] Furthermore, after collecting multi-dimensional transmission performance observation data, the performance monitoring module performs normalization and outlier filtering on this data. Normalization maps the original observations to a unified dimensionless numerical range based on the capability profile of the domestic IT innovation environment and the various capability ranges and typical performance intervals recorded in historical sample statistics. For example, throughput is scaled proportionally to the maximum achievable bandwidth of the node, latency is normalized to the historical baseline latency, and thread load is standardized to the maximum processing capacity, thereby eliminating direct comparability differences in indicator values ​​between different nodes and different services.

[0117] Outlier filtering is primarily used to remove isolated extreme samples caused by transient jitter, abnormal interference, or brief faults. For example, it removes abnormally large latency values ​​or abnormally high retransmission ratios within a single window to prevent these extreme samples from unreasonably impacting the overall performance characteristic modeling. After the above preprocessing, this embodiment inputs the multidimensional normalized observation data from each time window into a preset multidimensional feature mapping model for feature extraction. This feature mapping model can be a lightweight deep feature mapping structure or other model structures with nonlinear mapping capabilities, used to compress high-dimensional, redundant window-level observation data into a lower-dimensional performance representation space.

[0118] By aggregating and transforming the features of multiple windows on the time axis, the model finally outputs a transmission performance feature vector, which is used to characterize the overall performance characteristics of the current transmission stage under the selected set of policy parameters. For example, it comprehensively reflects the balance achieved by the set of policy parameters between throughput, latency, retransmission and resource consumption.

[0119] Furthermore, after obtaining the transmission performance feature vector, this embodiment combines and encodes the transmission performance feature vector with the corresponding transmission strategy parameter set, the information technology innovation environment capability profile, and the data flow profile to generate a strategy performance association record. For example, the strategy performance association record contains four types of information: first, the information technology innovation environment capability profile representing the current node's hardware and software capabilities; second, the data flow profile representing the current business characteristics and security level; third, the specific parameter values ​​of the transmission strategy parameter set used in this transmission; and fourth, the transmission performance feature vector extracted by the aforementioned multi-dimensional feature mapping model.

[0120] Combinatorial encoding can be accomplished by dividing each type of information into fixed fields or embedding fixed-length sub-vectors. The capability profile and data flow profile of the information technology innovation environment are encoded into capability vectors and profile vectors, respectively. These vectors are then concatenated with the policy parameter vector obtained by encoding the policy parameter set and the transmission performance feature vector to form a complete policy performance association record.

[0121] The performance monitoring module stores the policy performance correlation record in the policy performance sample library, and establishes indexes according to data flow profile type, information technology innovation environment capability profile identifier, and policy parameter set identifier. This allows the subsequent policy update model to quickly retrieve relevant samples based on these indexes when looking for similar environments and similar business scenarios. For example, for a batch document distribution task executed multiple times, a policy performance correlation record corresponding to the policy parameter set used in that task will be generated after each task is completed. The sample library gradually accumulates the performance of different policy parameter combinations under different loads and network conditions.

[0122] In some embodiments, the transmission strategy parameter set is updated based on transmission performance metrics, and the updated transmission strategy parameter set is used for accelerated transmission control of subsequent data streams with the same or similar data stream profiles, including:

[0123] Select a set of target samples from the strategy performance sample library that match the information technology innovation environment capability profile, data flow profile and transmission strategy parameter set corresponding to the target data flow. Each sample in the target sample set includes the transmission performance feature vector and the corresponding transmission strategy parameter set stored in the strategy performance association record.

[0124] The transmission performance feature vectors in the target sample set are aggregated and statistically analyzed and multidimensional similarity measures are performed. Combined with the preset performance evaluation criteria, the candidate strategy parameter set that meets the preset performance threshold is determined. The candidate strategy parameter set and the current transmission strategy parameter set are used together as the input for strategy update.

[0125] The policy update input is used as the input to the preset policy update model, which includes a rule-based parameter adjustment sub-model and a reinforcement learning-based parameter search sub-model. The parameter adjustment sub-model corrects policy values ​​that do not conform to the constraint rules, and the parameter search sub-model iteratively searches the transmission policy parameter set in the parameter space that satisfies the constraint rules to obtain the updated transmission policy parameter set.

[0126] The updated transmission strategy parameter set is indexed with the corresponding information technology innovation environment capability profile and data flow profile. A strategy index structure for retrieval by data flow profile is constructed. When a transmission request for a subsequent data flow with the same or similar data flow profile is received, the updated transmission strategy parameter set matching the data flow profile is retrieved according to the strategy index structure. The retrieved updated transmission strategy parameter set is used as the initial configuration of the preset strategy decision process for the accelerated transmission control of subsequent data flows.

[0127] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.

[0128] Figure 2 This is a schematic diagram of the structure of a dynamic adaptive data stream acceleration transmission device for the information technology innovation environment provided in this application embodiment. Figure 2 As shown, this dynamic adaptive data stream acceleration transmission device for the information technology innovation environment includes:

[0129] The detection module 201 is used to detect the capabilities of the central processing unit, operating system, network interface and cryptographic acceleration unit in the domestic innovation computing node, obtain environmental capability information that characterizes computing capabilities, network transceiver capabilities and cryptographic processing capabilities, and organize the environmental capability information into a domestic innovation environment capability profile.

[0130] The parsing module 202 is used to parse the business type, data size, timeliness level and security level of the data stream when a data stream transmission request is received, generate a data stream profile corresponding to the data stream in combination with preset classification rules, and assign a stream identifier to the data stream to identify the data stream profile.

[0131] The acquisition module 203 is used to collect cross-layer operational status information corresponding to the capability profile and data flow profile of the information technology innovation environment during data flow transmission, and organize the operational status information into an operational status vector.

[0132] The determination module 204 is used to determine the set of transmission strategy parameters for data flow based on the capability profile, data flow profile and running status vector of the information technology innovation environment, by executing a preset strategy decision process.

[0133] The sending module 205 is used to perform fragmentation processing, connection and queue configuration, password processing configuration and sending scheduling on the data stream corresponding to the stream identifier according to the transmission strategy parameter set, and send the data fragments according to the fragmentation configuration parameters and concurrent transmission configuration parameters in the transmission strategy parameter set.

[0134] The update module 206 is used to collect transmission performance indicators corresponding to the transmission strategy parameter set during the transmission process, associate the transmission performance indicators with the transmission strategy parameter set, update the transmission strategy parameter set based on the transmission performance indicators, and use the updated transmission strategy parameter set for accelerated transmission control of subsequent data streams with the same or similar data stream profiles.

[0135] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0136] Figure 3 This is a schematic diagram of the electronic device 3 provided in an embodiment of this application. Figure 3 As shown, the electronic device 3 of this embodiment includes: a processor 301, a memory 302, and a computer program 303 stored in the memory 302 and executable on the processor 301. When the processor 301 executes the computer program 303, it implements the steps in the various method embodiments described above. Alternatively, when the processor 301 executes the computer program 303, it implements the functions of each module / unit in the various device embodiments described above.

[0137] Electronic device 3 can be a desktop computer, laptop, handheld computer, cloud server, or other electronic device. Electronic device 3 may include, but is not limited to, processor 301 and memory 302. Those skilled in the art will understand that... Figure 3 This is merely an example of electronic device 3 and does not constitute a limitation on electronic device 3. It may include more or fewer components than shown, or different components.

[0138] The processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0139] The memory 302 can be an internal storage unit of the electronic device 3, such as a hard disk or memory of the electronic device 3. The memory 302 can also be an external storage device of the electronic device 3, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 3. The memory 302 can also include both internal and external storage units of the electronic device 3. The memory 302 is used to store computer programs and other programs and data required by the electronic device.

[0140] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0141] If integrated modules / units are implemented as software functional units and sold or used as independent products, they can be stored in a readable storage medium (e.g., a computer-readable storage medium). Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program may include computer program code, which may be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable storage medium may include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0142] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A dynamic adaptive data stream acceleration method for the information technology innovation environment, characterized in that, include: In the domestic IT innovation computing node, the capabilities of the central processing unit, operating system, network interface and cryptographic acceleration unit are detected to obtain environmental capability information that characterizes computing capabilities, network transceiver capabilities and cryptographic processing capabilities, and the environmental capability information is organized into a domestic IT innovation environment capability profile. Upon receiving a data stream transmission request, the service type, data size, timeliness level, and security level of the data stream are parsed. A data stream profile corresponding to the data stream is generated by combining preset classification rules, and a stream identifier is assigned to the data stream to identify the data stream profile. During data stream transmission, cross-layer acquisition of operational status information corresponding to the information technology innovation environment capability profile and the data stream profile is performed, and the operational status information is organized into an operational status vector. Based on the capability profile, data flow profile, and operational status vector of the information technology innovation environment, a preset strategy decision-making process is executed to determine the set of transmission strategy parameters for the data flow. According to the transmission strategy parameter set, the data stream corresponding to the stream identifier is processed by fragmentation, connection and queue configuration, password processing configuration and transmission scheduling, and the data fragments are sent according to the fragmentation configuration parameters and concurrent transmission configuration parameters in the transmission strategy parameter set; During the transmission process, transmission performance indicators corresponding to the transmission strategy parameter set are collected, the transmission performance indicators are associated with the transmission strategy parameter set, and the transmission strategy parameter set is updated based on the transmission performance indicators. The updated transmission strategy parameter set is then used for accelerated transmission control of subsequent data streams with the same or similar data stream profiles. This includes updating the transmission strategy parameter set based on transmission performance indicators, and using the updated transmission strategy parameter set for accelerated transmission control of subsequent data flows with the same or similar data flow profiles, including: Select a set of target samples from the strategy performance sample library that match the information technology innovation environment capability profile, data flow profile and transmission strategy parameter set corresponding to the target data flow. Each sample in the target sample set includes the transmission performance feature vector and the corresponding transmission strategy parameter set stored in the strategy performance association record. The transmission performance feature vectors in the target sample set are aggregated and statistically analyzed and multidimensional similarity measures are performed. Combined with the preset performance evaluation criteria, the candidate strategy parameter set that meets the preset performance threshold is determined. The candidate strategy parameter set and the current transmission strategy parameter set are used together as the input for strategy update. The policy update input is used as the input to the preset policy update model, which includes a rule-based parameter adjustment sub-model and a reinforcement learning-based parameter search sub-model. The parameter adjustment sub-model corrects policy values ​​that do not conform to the constraint rules, and the parameter search sub-model iteratively searches the transmission policy parameter set in the parameter space that satisfies the constraint rules to obtain the updated transmission policy parameter set. The updated transmission strategy parameter set is indexed with the corresponding information technology innovation environment capability profile and data flow profile. A strategy index structure for retrieval by data flow profile is constructed. When a transmission request for a subsequent data flow with the same or similar data flow profile is received, the updated transmission strategy parameter set matching the data flow profile is retrieved according to the strategy index structure. The retrieved updated transmission strategy parameter set is used as the initial configuration of the preset strategy decision process for the accelerated transmission control of subsequent data flows.

2. The method according to claim 1, characterized in that, The process of probing the central processing unit, operating system, network interface, and cryptographic acceleration unit within the domestically developed computing node to obtain environmental capability information characterizing computing capabilities, network transceiver capabilities, and cryptographic processing capabilities includes: The hardware description information and operating system kernel status information of the domestically developed computing node are obtained, and the number of cores, clock frequency, cache configuration and operating system scheduling policy parameters of the central processing unit are determined to obtain configuration information used to characterize the basic characteristics of computing capabilities. Preset computational probing tasks and cryptographic operation probing tasks are executed on the central processing unit and cryptographic acceleration unit. Instruction execution count, storage access count and cryptographic operation completion count are collected. Combined with the configuration information, multi-dimensional capability measurement data characterizing computational capability and cryptographic processing capability are calculated. Preset network transmit and receive probe data packets are sent to the network interface. Under different message lengths and different queue configurations, transmit and receive latency, packet drop counts and queue length change information are collected to calculate multi-dimensional capability measurement data characterizing network transmit and receive capabilities. The multidimensional capability measurement data representing computing power, network transceiver capability, and cryptographic processing capability are normalized and encoded to generate the environmental capability information.

3. The method according to claim 1, characterized in that, The process involves analyzing the business type, data scale, timeliness level, and security level of the data stream, and combining this with preset classification rules to generate a data stream profile corresponding to the data stream, including: Extract business attribute parameters and security attribute parameters from the business identifier field, data length field, time constraint field and security policy identifier field in the data stream transmission request, and construct an original feature set by combining it with the historical transmission records corresponding to the data stream; The original feature set is standardized and encoded, and a first feature vector representing the business type, data scale, timeliness level, and security level is generated using a preset multidimensional feature representation model. The first feature vector is compared with multiple reference feature vectors in a pre-built data flow profile reference library for similarity measurement. Based on the similarity measurement results, the classification label and hierarchical mark of the data flow are determined according to a preset classification rule. The classification label, hierarchical marker, and the first feature vector are combined and encoded to generate the data stream profile.

4. The method according to claim 1, characterized in that, The cross-layer acquisition corresponds to the operational status information of the information technology innovation environment capability profile and the data flow profile, and the operational status information is organized into an operational status vector, including: Based on the capability profile of the information technology innovation environment and the data flow profile, network layer monitoring indicators, system layer monitoring indicators, application layer monitoring indicators and cryptographic processing monitoring indicators corresponding to the target data flow are selected from the preset monitoring indicator library to generate cross-layer monitoring configuration; According to the cross-layer monitoring configuration, the operating status information of the target data stream in the current transmission stage is obtained by combining time window sampling and event-triggered sampling, forming multi-dimensional operating status observation data; The multidimensional operational status observation data is subjected to time series aggregation and normalization processing, and network state sub-vectors, system state sub-vectors, application state vectors and cryptographic processing state sub-vectors are generated by using a preset multidimensional time series feature extraction model. A feature fusion algorithm based on an attention weighting mechanism is used to fuse and encode the network state sub-vector, system state sub-vector, application state sub-vector, and cryptographic processing state sub-vector to obtain the running state vector.

5. The method according to claim 1, characterized in that, Based on the capability profile, data flow profile, and operational status vector of the domestic IT innovation environment, a preset strategy decision-making process is executed to determine the set of transmission strategy parameters for the data flow, including: Based on the capability profile of the information technology innovation environment and the data flow profile, candidate strategy templates that match the current capability characteristics and data flow profile characteristics of the information technology innovation environment are selected from the preset strategy template library, and an initial strategy parameter set is generated according to the candidate strategy templates. The running state vector and the initial policy parameter set are used as joint inputs. The initial policy parameter set is adjusted and calculated using a preset multi-layer parameter mapping model to obtain an intermediate policy parameter set that represents the target policy value in the current transmission stage. The multi-layer parameter mapping model includes a rule-based mapping sub-model and a learning-based parameter correction sub-model. Based on preset policy constraint rules and security compliance rules, the intermediate policy parameter set is constrained, verified, and corrected. Policy values ​​that do not match the capability profile of the information technology innovation environment are removed, and policy values ​​that violate security compliance rules are adjusted to generate a transmission policy parameter set that meets the constraint conditions.

6. The method according to claim 1, characterized in that, The step of performing fragmentation, connection and queue configuration, cryptographic processing configuration, and transmission scheduling on the data stream corresponding to the stream identifier according to the transmission policy parameter set includes: Based on the fragmentation configuration parameters in the transmission strategy parameter set, the original data stream corresponding to the stream identifier is segmented, and the stream identifier, sequence number mark and security attribute mark are added to each data fragment to generate a data fragment sequence to be sent. Based on the concurrent transmission configuration parameters and connection configuration parameters in the transmission strategy parameter set, multiple transmission channels are established on the domestic computing node, the processing thread and network interface queue corresponding to each transmission channel are determined, the data fragment sequence to be sent is divided into multiple sub-sequences according to the preset mapping rules, and each sub-sequence is assigned to different transmission channels and network interface queues respectively. Based on the cryptographic processing path configuration parameters in the transmission strategy parameter set, a hierarchical cryptographic processing pipeline is constructed. Based on the capability profile of the information technology innovation environment, the corresponding cryptographic operation is selected to be performed in the central processing unit or the cryptographic acceleration unit. The data fragments to be sent are queued and dequeued according to the depth and parallelism of the hierarchical cryptographic processing pipeline. Based on the retransmission and redundancy control parameters and priority scheduling parameters in the transmission strategy parameter set, a transmission scheduling sequence is generated for multiple transmission channels and network interface queues. A queue scheduling algorithm is used to select the target data fragment from the ready data fragments to be sent. According to the transmission scheduling sequence, the transmission interface is called on the corresponding transmission channel to send the target data fragment to the target node.

7. The method according to claim 1, characterized in that, The step of collecting transmission performance metrics corresponding to the transmission strategy parameter set during transmission, and associating the transmission performance metrics with the transmission strategy parameter set, includes: Select throughput, latency, retransmission and redundancy usage, processing thread load, and cryptographic operation usage indicators corresponding to the target data stream from the preset performance monitoring indicator library, and generate a performance monitoring configuration corresponding to the transmission strategy parameter set; According to the performance monitoring configuration, during the execution of the data stream fragmentation transmission, confirmation feedback reception and password processing pipeline, multi-dimensional transmission performance observation data corresponding to the stream identifier is collected in a sliding time window manner. The multidimensional transmission performance observation data is normalized and outlier filtered out. The processed multidimensional transmission performance observation data is then converted into a transmission performance feature vector that characterizes the performance characteristics of the current transmission stage using a preset multidimensional feature mapping model. The transmission performance feature vector is combined and encoded with the corresponding transmission strategy parameter set, the information technology innovation environment capability profile, and the data flow profile to generate a strategy performance association record, and the strategy performance association record is stored in the strategy performance sample library.

8. A dynamic adaptive data stream acceleration transmission device for the information technology innovation environment, characterized in that, include: The detection module is used to detect the capabilities of the central processing unit, operating system, network interface and cryptographic acceleration unit in the domestic IT innovation computing node, obtain environmental capability information that characterizes computing capabilities, network transceiver capabilities and cryptographic processing capabilities, and organize the environmental capability information into a domestic IT innovation environment capability profile. The parsing module is used to parse the service type, data size, timeliness level and security level of the data stream when a data stream transmission request is received, generate a data stream profile corresponding to the data stream in combination with preset classification rules, and assign a stream identifier to the data stream to identify the data stream profile. The acquisition module is used to collect operational status information corresponding to the information technology innovation environment capability profile and the data flow profile across layers during data flow transmission, and organize the operational status information into an operational status vector. The determination module is used to determine the set of transmission strategy parameters for the data flow by executing a preset strategy decision-making process based on the capability profile, data flow profile and running status vector of the information technology innovation environment. The sending module is used to perform fragmentation processing, connection and queue configuration, password processing configuration and sending scheduling on the data stream corresponding to the stream identifier according to the transmission strategy parameter set, and send the data fragments according to the fragmentation configuration parameters and concurrent transmission configuration parameters in the transmission strategy parameter set; The update module is used to collect transmission performance indicators corresponding to the transmission strategy parameter set during the transmission process, associate the transmission performance indicators with the transmission strategy parameter set, update the transmission strategy parameter set based on the transmission performance indicators, and use the updated transmission strategy parameter set for accelerated transmission control of subsequent data streams with the same or similar data stream profiles. The update module is used to select a target sample set from the strategy performance sample library that matches the information technology innovation environment capability profile, data flow profile, and transmission strategy parameter set corresponding to the target data flow. Each sample in the target sample set includes a transmission performance feature vector and a corresponding transmission strategy parameter set stored in the strategy performance association record. The module performs aggregation statistics and multi-dimensional similarity measurement on the transmission performance feature vectors in the target sample set, and determines a candidate strategy parameter set that meets a preset performance threshold based on preset performance evaluation criteria. The candidate strategy parameter set and the current transmission strategy parameter set are used together as the strategy update input. The strategy update input is used as the input to a preset strategy update model, which includes a rule-based parameter adjustment sub-model and a reinforcement learning-based sub-model. The parameter search sub-model corrects policy values ​​that do not conform to the constraint rules through the parameter adjustment sub-model. The parameter search sub-model iteratively searches the transmission policy parameter set within the parameter space that satisfies the constraint rules to obtain the updated transmission policy parameter set. The updated transmission policy parameter set is indexed with the corresponding information technology innovation environment capability profile and data flow profile to construct a policy index structure for retrieval by data flow profile. When a subsequent transmission request for a data flow with the same or similar data flow profile is received, the updated transmission policy parameter set matching the data flow profile is retrieved according to the policy index structure. The retrieved updated transmission policy parameter set is used as the initial configuration of the preset policy decision process for the accelerated transmission control of subsequent data flows.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images