User reverse management network method for multi-source heterogeneous network

Abstract

The invention provides a user reverse management network method for a multi-source heterogeneous network. In a multi-source heterogeneous network, different types of networks (such as a ground wired network, a wireless network, a satellite network and the like) are mutually interwoven, and the network environment is complex and changeable. Network service providers are responsible for managing and maintaining various network resources and expect to meet diversified requirements of users and maximize the resource utilization efficiency of the network service providers; however, the user hopes to obtain the network quality of service (QoS) meeting the self task requirement at the minimum cost when renting the network resources, and in order to more effectively realize the fine-grained on-demand distribution of a single small task in the multi-source heterogeneous network, the user equipment can interact the use experience of different network links, so that the user experience is improved, and the user experience is improved. And taking the result as a reference basis for network resource lease decision making. Through the interaction, the user can more comprehensively know the actual performance of the network link, so that a more reasonable routing decision is made.

Description

Technical Field

[0001] The present invention relates to a resource management method for multi-source heterogeneous networks. Background Art

[0002] With the continuous development of network technologies, people's demand for networks has become increasingly strong. Ubiquitous network coverage and low-cost, highly reliable network services have become important requirements for network services.

[0003] In existing single networks, the network provider has the management right of the network, while the users of network services can only passively receive network services and do not have the ability to control the network. When the number of network service providers increases, if users can access multiple networks simultaneously, users can adjust the selection and use of different types of networks according to their own demands for network services, and on this basis, achieve the goal of users' reverse management of the network.

[0004] In view of the problem of users' management of multi-source heterogeneous networks, the present invention provides a reverse management method for multi-source heterogeneous network resources: abstracting network resources as a set of data transmission channels with different bandwidths through a network coding method; realizing the on-demand allocation of network resources through the fine-grained on-demand selection of the network; realizing the integrated utilization of high-bandwidth services through the multi-path transmission of the network; and realizing the optimized use of network resources through the real-time switching of the network. Summary of the Invention

[0005] The present invention is a method for realizing user-side management of multi-source heterogeneous networks. The two key problems solved by this method are the abstraction of network resources and the on-demand allocation of network resources.

[0006] Network Coding Method

[0007] In the multi-source heterogeneous network resource management system of the present invention, the random linear network coding (RLNC) algorithm is adopted, which provides key support for the efficient operation of the entire system.

[0008] Random linear network coding (RLNC) is a coding method based on linear algebra theory and operates in a finite field. In a multi-source heterogeneous network, when a network node receives data packets from different data sources, these data packets are regarded as vectors in a finite field. The node randomly generates a set of coding coefficients and performs a linear combination operation on these data packet vectors to generate new coded data packets. This coding method enables each coded data packet to contain information from multiple original data packets. As long as the receiving end receives a sufficient number of coded data packets, it can restore the original data packets by solving a system of linear equations.

[0009] The RLNC algorithm can encode and fuse data packets from different links. Even if there are packet losses during transmission, the ground station can still recover the data according to the solution conditions of the linear equations, ensuring the reliability of the transmission. In the scenario of the space-ground integrated network, when multi-node collaborative communication occurs among vehicles, drones, etc., the RLNC algorithm can uniformly process the data from all parties and dynamically adjust the encoding strategy according to the real-time status of the links to ensure the efficient and accurate transmission of data in scenarios such as urban traffic monitoring.

[0010] Single small task fine-grained on-demand allocation method

[0011] In a multi-source heterogeneous network, different types of networks (such as terrestrial wired networks, wireless networks, satellite networks, etc.) are intertwined, and the network environment is complex and changeable. Network service providers are responsible for managing and maintaining various network resources. They expect to maximize the utilization efficiency of their own resources while meeting the diverse needs of users; while users hope to obtain the network service quality (QoS) that meets their task requirements at the lowest cost when leasing network resources. In order to more effectively achieve the fine-grained on-demand allocation of single small tasks in a multi-source heterogeneous network, user devices can interact their usage experiences of different network links, which can be used as a reference basis for network resource leasing decisions. This interaction enables users to more comprehensively understand the actual performance of network links, thereby making more reasonable routing decisions.

[0012] 1. Task and network parameter modeling

[0013] Consider that there are multiple user devices and various types of network links in a multi-source heterogeneous network. Let the set of user devices be UE = {1, 2, …, M}, and the total number of user devices be M; the set of network links be NL = {1, 2, …, N}, and the total number of network links be N.

[0014] For a single small task, define its task model as where Q m represents the data volume of the task, represents the maximum tolerable transmission delay of task J m .

[0015] Model the performance of network links, mainly considering three key factors: delay, bandwidth, and packet loss rate:

[0016] Delay: The delay T m for task J mn transmitted from user device m through network link n can be decomposed into transmission delay node processing delay and queuing delay That is where the transmission delay r mnIndicates the data transmission rate between user device m and network link n; node processing delay C n Represents the computing and processing capabilities of the nodes connected by network link n; queuing delay

[0017] Bandwidth: The available bandwidth of network link n is B n , task J m The bandwidth requirement during transmission is B mn , and it is necessary to satisfy B mn ≤B n .

[0018] Packet loss rate: The packet loss rate of task J m during transmission on network link n is L mn , which is an important indicator to measure the reliability of data transmission.

[0019] 2. Routing decision-making process

[0020] The routing decision-making process can be regarded as a game process between user devices and network resources. In this process, user devices select appropriate network links according to their own task requirements and the actual situation of network links, while network service providers adjust resource allocation strategies to meet user needs and maximize their own interests.

[0021] The decision-making goal of user devices is to minimize their own costs on the premise of meeting the task QoS requirements. The utility function U of user device m m can be modeled as:

[0022]

[0023] where α, β, and γ are weight coefficients, with values of 0.4, 0.5, and 0.4, respectively, indicating the degree of user attention to bandwidth, delay, and packet loss rate; B min,m is the minimum bandwidth requirement for task J m ; P mn is the cost per unit data volume for transmitting task J m using network link n.

[0024] The decision-making goal of network service providers is to maximize their own profits on the basis of meeting user task requirements. The utility function V of network link n n can be expressed as:

[0025]

[0026] where ζ is the energy consumption cost coefficient, with a value set to 0.6, and E n is the energy consumed by network link n for transmitting tasks.

[0027] Based on the above utility function, the user equipment and the network service provider gradually reach a balanced state by continuously adjusting their respective strategies. In this balanced state, the network link selected by the user equipment can maximize its own utility while meeting the task QoS requirements; the resources allocated by the network service provider can also maximize its own profit on the premise of meeting the user's needs.

[0028] 3. Dynamic Adjustment Mechanism

[0029] The state of the multi-source heterogeneous network is dynamically changing, and the performance of the network link (such as delay, bandwidth, packet loss rate) will change with time and network load. In order to adapt to this dynamic change, a dynamic adjustment mechanism needs to be established.

[0030] When the network state changes, the user equipment and the network service provider re-evaluate their respective utility functions and adjust their decision-making strategies accordingly. If the delay of a certain network link suddenly increases and exceeds the maximum delay tolerance of the task, the user equipment will recalculate the utility values of each network link and select a network link with lower delay for task transmission; the network service provider will adjust the resource allocation strategy according to the network load situation, such as adjusting the bandwidth allocation of the link or improving the processing capacity of the node, to optimize the network performance.

[0031] Define the dynamic adjustment evaluation function R mn , which is used to evaluate whether the transmission situation of task J m on network link n meets the requirements and the comprehensive performance of the link. This function is constructed based on the QoS requirements of the task and the real-time state of the network link. The formula is as follows:

[0032]

[0033] Among them, ω1, ω2, and ω3 are weight coefficients, which respectively represent the relative importance of delay, bandwidth, and packet loss rate in the evaluation, and ω1 + ω2 + ω3 = 1; is the maximum acceptable packet loss rate of task J m .

[0034] The setting of the weight coefficients ω1, ω2, and ω3 needs to be adjusted according to the type of task and the actual situation of the network. For tasks with extremely high real-time requirements, such as real-time video stream transmission, where latency has a huge impact on the user experience, ω1 can be set to a relatively large value at this time. For example, ω1 = 0.6, ω2 = 0.2, ω3 = 0.2, highlighting the importance of latency in the evaluation. For tasks that require strict data accuracy, such as financial transaction data transmission, the impact of the packet loss rate is more critical, and ω1 = 0.2, ω2 = 0.2, ω3 = 0.6 can be set. For some tasks with high bandwidth requirements, such as enterprise data backup and migration, the proportion of ω2 can be appropriately increased, such as ω1 = 0.2, ω2 = 0.6, ω3 = 0.2.

[0035] Single large task multi-path parallel transmission method

[0036] Suppose there are n links available for transmitting data packets from the sender to the receiver, and the network state attributes of each link are related to the end-to-end delay D, link bandwidth B, packet loss rate PLR, the number of data packets M transmitted on the link, network utilization U, and target delay L. Therefore, the network state attributes of each link can be described by S i ={D i ,B i ,PLR i ,M i ,U i ,L i}, i = 1, 2, …, n. Among them, the end-to-end delay D is a key factor in measuring the packet delivery situation and the network status of the transmission link. It refers to the time period from when the data packet starts to be sent from the sender to when the receiver receives the data packet, and is mainly affected by the propagation delay P, queuing delay Q, and transmission delay T, as shown in Equation (1):

[0037] D i =P i +Q i +T i , i = 1, 2, …, n (1)

[0038] During the data transmission process, the propagation delay P refers to the time consumed by the electromagnetic wave carrying the data packet information to propagate in the physical channel of the communication link. It is mainly affected by the actual physical conditions of the link. Specifically, it is determined by the length of the link transmission medium. The queuing delay Q is different. Its size mainly depends on the current network traffic of the link. In the specific scenario of multi-path parallel transmission, the queuing delays of each link are independent of each other and have no correlation with each other.

[0039] Based on the above situation, this paper further divides the end-to-end delay into the link absolute delay and the relative link delay T. Among them, the absolute link delay covers the propagation delay P and the queuing delay Q. In actual analysis, if the asymmetry of the end-to-end delay in the two processes of the data packet from the sender to the receiver and the acknowledgment of the received data packet from the receiver back to the sender is ignored, then it can be approximately considered that the propagation delay P is about half of the round-trip delay RTT, that is

[0040]

[0041] The magnitude of the queuing delay is often jointly determined by multiple factors, mainly including the size of the data packet, the utilization rate of the node, the transmission speed, and whether the data packet is sent in a periodic or bursty manner. In the actual scenario of real-time video transmission, the data packet is generally transmitted at a specific bit rate, and the traffic will continuously arrive at the node. At the same time, the buffer size of link i also needs to be considered. In view of this, the magnitude of the queuing delay can be expressed in the following way:

[0042]

[0043] In the formula: represents the average size of the link data packet. C i represents the buffer size of link i.

[0044] The relative link delay, that is, the transmission delay T, depends on the size of the data to be sent, the current bandwidth of the link network, and the additional transmission time of the link error correction mechanism, that is:

[0045]

[0046] In the formula, E corr is the additional transmission time coefficient of the link error correction mechanism, and its value range is [0, +∞). It depends on the complexity of the error correction algorithm adopted by the link. The more complex E corr the larger it is. For example, Bluetooth transmission often uses a simple cyclic redundancy check algorithm for error correction. Its calculation is simple and the additional transmission time is small. E corr is generally between 0.05 and 0.1. In satellite communication, in order to ensure data accuracy, a complex low-density parity-check code is used. Although its error correction ability is strong, the operation is complex, and E corr can reach 0.5–1.

[0047] Therefore, by combining equations (1)-(4), the end-to-end delay formula of the data packet of the i-th path in multi-path parallel transmission can be obtained:

[0048]

[0049] In a parallel transmission system, the bottleneck link in the multi-path transmission network is often the key factor affecting the transmission quality and system performance. The bottleneck link mentioned here refers to the one with the worst network state among the i links in the transmission network.

[0050] To ensure that the data packet can reach the receiving end smoothly within the target delay L and, on the premise of considering the reliability of the limited number of transmissions of the data packet, it is necessary to reserve a retransmission time T for the ***bottleneck link re . The purpose of doing this is to enable the end-to-end delay of the entire bottleneck link to meet the requirements of the target delay, that is, to satisfy:

[0051]

[0052] In the formula: sign(M i ) is the sign function:

[0053]

[0054] In the formula, I link is the importance degree of the link, and its value range is [0, 1]. The larger the value, the more important the link is. For example, one of the core businesses of ByteDance is short video content distribution, and the importance degree I link of its link can take 0.8. For edge services such as built-in games, the importance degree I link of its link can take 0.2.

[0055] Considering that different transmission service types have their own unique characteristics, in order to enable the multi-path parallel transmission system to better fit the actual situation and requirements of the service, it is necessary to constrain key factors such as the transmission delay, packet loss rate, and quality of service (QoS) of the system.

[0056] In view of this, this paper uses the method of linear programming to abstract the link into a mathematical model that includes the current bandwidth, end-to-end delay, network utilization rate, and target delay. By solving this linear programming problem, the maximum link switching time node allowed in the current network environment can be determined, and then the link combination that can minimize the transmission delay can be found. Specifically, the objective function and constraints of this linear programming problem are as follows:

[0057]

[0058] In the above formula, D0 represents the end-to-end delay when the link is in the idle state. In this state, the data packets transmitted on the link do not need to experience the queuing waiting process, and its value is the time required for the data packet to travel from the sending end to the receiving end without queuing waiting.

[0059] While τ is a variable parameter, its specific value is closely related to the specific requirements of the end-to-end delay for different real-time transmission service types. The following are examples of the typical value ranges of τ (unit: ms) corresponding to different real-time transmission service types: In voice communication, τ usually takes a value not exceeding 150; in the video call scenario, τ is generally controlled within 100; for industrial control services with extremely strict requirements for delay, τ needs to be controlled within 10.

[0060] It can be seen from Equation (8) that under the strict constraint of the target delay, if we want to make the most of the resource bandwidth of the link network as much as possible, while meeting the requirements of improving link utilization and ensuring transmission reliability through a limited number of retransmissions, there is an upper limit for the optimal handover time of the bottleneck link.

[0061] However, in the actual process of data packet transmission, the network conditions of the link are not static, but in a continuous dynamic change. The actual bandwidth, network utilization rate, and end-to-end delay of each link are not only different from each other, but also constantly changing. Therefore, for each link, the optimal link handover time point T for data packet retransmission re will also change continuously.

[0062] Driven by this optimization problem, T is calculated in real time through the feedback data packet information of the receiving end re , and according to the calculated T re and the last response time T 1st are compared, as shown in Algorithm 1.

[0063] Algorithm 1: Dynamic Link State Awareness Algorithm

[0064] Input:

[0065] Output: The network status of each link in real time, and the specific status types are:

[0066] active_stable: Indicates that the link is in a stable state and can reliably transmit data packets.

[0067] active_unstable: Means that the link state is unstable, which may affect the normal transmission of data packets, and the transmission strategy needs to be adjusted.

[0068] active_wary: Indicates that the link state is in a critical state, and it is necessary to closely monitor and appropriately adjust the data packet allocation to prevent the link performance from deteriorating.

[0069] 1. Initialization: Obtain the current state of link i, denoted as LinkState; according to the formula

[0070]

[0071] Calculate the value at the current moment T re ; Obtain the value at the current moment T 1st .

[0072] 2. Condition judgment:

[0073] If LinkState == active_stable, then make the following judgment:

[0074] If T re > T 1st , update LinkState to active_unstable.

[0075] Otherwise, if T re = T 1st , update LinkState to active_wary.

[0076] If none of the above conditions are met, keep LinkState as active_stable.

[0077] If LinkState is not active_stable, keep the current LinkState value unchanged.

[0078] 3. Return LinkState.

[0079] In the single large task multi-path parallel transmission method, the link weight factor α i plays a key role in the reasonable allocation of data packets. It comprehensively considers multiple important attributes of the link to reflect the superiority degree of the link in the transmission task. After analyzing factors such as the link bandwidth B i , packet loss rate PLR i , network utilization U i , etc., the formula is obtained:

[0080]

[0081] Among them, the link bandwidth B i directly affects the data transmission speed. The larger the bandwidth, the more data can be transmitted per unit time. In the formula, it is positively correlated with α i , that is, the larger the bandwidth, the larger α i , and the higher the weight of this link in the data packet allocation. The packet loss rate PLR i is an important indicator to measure the reliability of data transmission. The lower the packet loss rate, the higher the reliability of data transmission. Through the form of PLR i , the link with a lower packet loss rate has a higher α i value. The network utilization U i reflects the busyness of the link, Ui The lower it is, the more idle resources the link has, and the more suitable it is to undertake the data packet transmission task. In the formula, by in the form of, to avoid the situation where the denominator becomes zero when U i = 0, and at the same time, the lower the network utilization rate, the larger α i is.

[0082] T in Algorithm 1 1st represents the time period between the current moment and the moment when the last feedback data packet was received. Comparing this time period with the maximum link switching time node T re , the data packet transmission strategy can be dynamically adjusted according to the result:

[0083] If T re > T 1st , it indicates that the network condition of this link has deteriorated and it is no longer suitable to continue transmitting data packets. At this time, to ensure that the data packets can reach the receiving end within the target delay, the data packets that have not received the acknowledgment response message on this link need to be transferred to a link with good network condition for retransmission. The specific operation can refer to Algorithm 2.

[0084] If T re < T 1st , this means that the network condition of this link is good and it can continue to undertake the data packet transmission task, and data packets can be continuously allocated to it.

[0085] When T re = T 1st , it indicates that the network condition of this link is in the critical state of congestion. To avoid the link from getting congested, the number of data packets allocated to this link needs to be reduced until its network state improves. Algorithm 2: Multi-way shunt algorithm for dynamic links

[0086] Input: Real-time network state LinkState[i] of each link, T rei , T 1sti ,

[0087] Link bandwidth B i , packet loss rate PLR i , network utilization rate U i , i = 1, 2,..., n

[0088] Output: Link data packet allocation weight ω i Satisfying used to guide the allocation ratio of data packets on different links.

[0089] 1. Initialization: According to the formula

[0090]

[0091] Calculate the weight factor α of link i i 。

[0092] 2. Loop judgment (i from 1 to n):

[0093] If LinkState[i]!= active_stable, assign the weight ω of the data packet of link i i Set it to 0.

[0094] Otherwise, calculate the weight of the data packet assignment of link i

[0095] 3. Return ω i 。

[0096] Network Real-time Switching Method for Multi-task Parallelism

[0097] In a multi-task parallel scenario, when a high-priority task arrives, a combined algorithm based on AHP-TOPSIS is used for network real-time switching. First, determine the key indicators for network evaluation, such as bandwidth, latency, packet loss rate, and signal strength. Use the analytic hierarchy process to construct a judgment matrix for pairwise comparison of each indicator, calculate the weight of each indicator, and conduct a consistency test to ensure reasonable weight allocation. Then, collect the data of each indicator of each available network to form a decision matrix, perform standardization processing on this matrix to eliminate the influence of different indicator dimensions, and then combine the determined weights to obtain a weighted standardized matrix. After that, determine the positive ideal solution and the negative ideal solution. The positive ideal solution is composed of the optimal values of each indicator, and the negative ideal solution is composed of the worst values. Calculate the distances from each network to the positive and negative ideal solutions, and then obtain the closeness degree. The closer the closeness degree is to 1, the closer the network is to the optimal choice. Finally, sort all available networks according to the closeness degree, and switch the high-priority task to the network with the largest closeness degree to ensure the efficient and stable execution of the task.

[0098] The network switching algorithm based on AHP-TOPSIS has obvious advantages. In terms of decision-making, it comprehensively considers multiple indicators and reasonably allocates weights using AHP, which is more scientific and meets the requirements. It has strong adaptability, can adapt to dynamic networks, and flexibly cope with multi-network scenarios. It has high accuracy, and TOPSIS accurately sorts the networks to ensure the stable and efficient execution of high-priority tasks. Brief Description of the Drawings

[0099] Figure 1 It is a schematic diagram of the process of the present invention. Detailed Embodiments

[0100] (1) Task and Network Parameter Collection

[0101] In the entire data transmission process, data collection and encoding are extremely crucial steps. Verifying, time-synchronizing, and integrating the collected data are important measures to ensure data quality. Verifying data can remove outliers and guarantee data reliability; time synchronization makes different data consistent in terms of time, facilitating subsequent comprehensive analysis; integrating data aggregates scattered parameters to form a complete information data packet, providing convenience for subsequent processing and applications.

[0102] In the scenario of satellite-to-ground data transmission, it is necessary to encode the data. On the one hand, the satellite communication environment is complex, and signals are vulnerable to interference, resulting in easy loss or error of data packets. By adopting the Random Linear Network Coding (RLNC) algorithm, where each encoded data packet integrates the information of multiple original data packets, it can ensure that even if some data packets are lost, the ground data center can still solve the linear equations with a sufficient number of encoded data packets to restore the original data, greatly improving the transmission reliability. On the other hand, the satellite link bandwidth and transmission capacity are limited. Dividing the data into appropriately sized data packets for encoding can better adapt to the transmission characteristics of the satellite link, reducing the transmission delay and the risk of packet loss. In addition, adding identification information to the encoded data packets facilitates the ground data center to identify the data source, encoding method, and time sequence, which is conducive to the efficient decoding, classification, storage, and subsequent analysis and application of the data.

[0103] (2) Selection of transmission mode

[0104] In the process of satellite-to-ground data transmission, the selection of the transmission mode depends on the accurate monitoring of the network link status and the clear definition of task requirements. On the one hand, the ground data center uses professional network monitoring tools to conduct all-round and continuous monitoring of available links such as the ground wired network and 5G wireless network, and obtains key parameters such as delay, bandwidth, and packet loss rate in real time at minute intervals. By deeply analyzing historical data, mastering the fluctuation law of link performance, and combining the network topology structure and traffic distribution to predict future performance changes, it provides a comprehensive and accurate network status basis for the selection of the transmission mode. On the other hand, it is necessary to accurately evaluate the data volume before transmission, estimate the size of the data volume to be transmitted each time according to the characteristics of the data collected by the satellite, and at the same time determine the timeliness requirements according to the application scenario of the data. For example, meteorological warning data has extremely high timeliness requirements and needs to be transmitted within a few minutes; while data used for long-term climate research has relatively loose timeliness requirements and can be transmitted within a few hours. The clear definition of these task requirements is an important prerequisite for reasonably selecting the transmission mode.

[0105] (3) Application of the fine-grained on-demand allocation method for individual small tasks

[0106] Step 1: Network Link Information Collection. Determine the currently available network links, which may include terrestrial wired networks (such as fiber optic, Ethernet), 5G wireless networks, satellite communication links, etc. Through the status indicator lights of network devices, network management software, etc., confirm whether each link is working properly. Use network monitoring tools to collect parameters such as latency, bandwidth, and packet loss rate of each link in real time. To ensure data accuracy, conduct multiple measurements and take the average value.

[0107] Step 2: Link Utility Value Calculation and Selection. According to the characteristics of the task, determine the key factors affecting the transmission effect, usually including bandwidth, latency, and packet loss rate. For example, for tasks with high real-time requirements, the weight of latency may be relatively large; for large data volume transmission, the weight of bandwidth may be more important. Substitute the collected link parameters into the utility function constructed in the technical solution to calculate the utility value of each link. Compare the utility values of each link and select the link with the largest utility value as the data transmission link for this small task.

[0108] Step 3: Data Transmission and Monitoring. After determining the link, start data transmission. During the transmission process, continuously monitor the status of the selected link, and re-collect parameters such as latency, bandwidth, and packet loss rate of the link every certain period (such as 1 - 5 minutes). Use network monitoring software or custom scripts to achieve real-time monitoring.

[0109] Step 4: Dynamic Adjustment and Optimization. According to the link parameters obtained from real-time monitoring, recalculate the utility values of each link. If it is found that the status of the selected link has changed significantly, resulting in a decrease in its utility value, switching the link can be considered. Before switching the link, save the transmission progress, and then switch to the new link to continue the transmission.

[0110] Step 5: Transmission Completion and Verification. When all the data is transmitted to the target location, stop the transmission task. Check the log or status information of the transmission tool to confirm whether the transmission ends normally. Conduct integrity verification on the received data at the receiving end, and compare whether the hash value of the received data is consistent with the hash value calculated at the sending end. If the data is damaged or lost, request the sending end to re-transmit the corresponding data segment. Record the transmission process of this small task, including the selected link, transmission time, transmission rate, link status changes, etc. Summarize this transmission, analyze the problems that occurred during the transmission and the effects of optimization measures, and provide an experience reference for subsequent small task transmissions.

[0111] (4) Application of the Multi-Path Parallel Transmission Method for a Single Large Task

[0112] Step 1: Network Link Survey and Evaluation. Determine the current network links available for data transmission, which may include terrestrial wired networks (such as fiber optic, Ethernet), wireless networks (such as 5G, Wi-Fi), satellite communication links, etc. Confirm the availability of each link by checking the connection status of network devices, information of network service providers, etc. Use professional network monitoring tools to measure and count the key parameters of each link. Set the status of each link through Algorithm 1 in the technical solution.

[0113] Step 2: Path Planning and Strategy Formulation. Use the method of linear programming, combine the target delay, data volume of the task and the parameters of each link to establish a mathematical model. Take factors such as the bandwidth, delay, and packet loss rate of the link as constraint conditions, and take maximizing the transmission efficiency or minimizing the transmission cost as the objective function. Combine Algorithm 2 in the technical solution to determine the optimal link combination and data packet allocation scheme.

[0114] Step 3: Data Parallel Transmission. According to the determined link combination and data packet allocation scheme, start parallel transmission of data blocks through multiple links simultaneously. Use a transport protocol or tool that supports multi-path transmission to ensure that the data blocks can be accurately sent to the corresponding links. During the transmission process, monitor the transmission status of each link in real time. Use network monitoring software or custom scripts to obtain parameters such as the bandwidth usage, delay change, and packet loss rate of each link in real time. At the same time, monitor the transmission progress of the data blocks and record the number of transmitted and untransmitted data blocks. According to the results of real-time monitoring, dynamically adjust the data block allocation ratio of each link. If the bandwidth of a certain link suddenly increases, the number of data blocks allocated to this link can be appropriately increased; if a certain link appears congested or has serious packet loss, reduce the number of data blocks allocated to this link to ensure the overall transmission efficiency.

[0115] Step 4: Error Handling and Retransmission Mechanism. At the receiving end, perform error detection on the received data blocks. By comparing the check information of the data blocks, determine whether the data blocks are damaged during transmission. If it is found that the check information of the data block does not match, mark the data block as an error data block. The receiving end sends a retransmission request to the sending end, indicating the number of the data block that needs to be retransmitted. After receiving the retransmission request, the sending end checks the data blocks saved locally and retransmits the corresponding data blocks. To improve the retransmission efficiency, data blocks that have a greater impact on the overall data integrity can be preferentially retransmitted. If a certain link fails, immediately stop the data transmission on this link, and switch the untransmitted data blocks to other available links for transmission according to the link switching rules. At the same time, troubleshoot and repair the faulty link so that it can participate in the transmission again after returning to normal.

[0116] Step 5: Data Reorganization and Verification. After all data blocks are transmitted to the receiving end, reorganize the data blocks in the correct order according to the marking information of the data blocks to restore the original large file. Perform integrity verification on the reorganized large file, calculate the hash value of the file again, and compare it with the original hash value provided by the sending end. If the hash values are the same, it indicates that the file transmission is complete; if they are different, further check the transmission status of the data blocks, identify the data blocks that may have problems and retransmit them. Evaluate the performance of the multi-path parallel transmission of this large task, and analyze whether indicators such as the actual transmission time, bandwidth utilization rate, and packet loss rate meet the expected goals. Summarize the problems encountered and solutions during the transmission process to provide experience reference for subsequent similar tasks, so as to further optimize the transmission strategy and improve the transmission efficiency.

[0117] (5) Application of the Network Real-time Switching Method in Multi-task Parallelism

[0118] The network real-time switching method in multi-task parallelism is applicable to the situation where multiple transmission tasks need to be executed simultaneously.

[0119] Step 1: Determination of Evaluation Indicators and Calculation of Weights. According to the characteristics and requirements of the task, determine the key indicators for evaluating the network link. Common indicators include bandwidth, latency, packet loss rate, signal strength, etc. Use the Analytic Hierarchy Process (AHP) to construct a judgment matrix for pairwise comparison of each evaluation indicator. Invite experts in the relevant field or based on historical experience, evaluate and score the relative importance of each indicator. By calculating the eigenvector of the judgment matrix, obtain the weights of each evaluation indicator and conduct a consistency test to ensure the rationality of the weight allocation.

[0120] Step 2: Determination of the Positive Ideal Solution and the Negative Ideal Solution. Real-time collect the data of each evaluation indicator of each available network link to form a decision matrix. Standardize the decision matrix to eliminate the influence of different indicator dimensions. Multiply the standardized matrix by the weights calculated in Step 3 to obtain a weighted standardized matrix. From the weighted standardized matrix, find the optimal value of each evaluation indicator to form the positive ideal solution. Find the worst value of each evaluation indicator to form the negative ideal solution.

[0121] Step 3: Calculation of Closeness Degree and Network Selection. According to the weighted standardized matrix and the positive and negative ideal solutions, use the Euclidean distance formula to calculate the distance of each network link to the positive ideal solution and the negative ideal solution. According to the calculated distances, calculate the closeness degree of each network link. Compare the closeness degrees of each network link and select the network link with the largest closeness degree as the transmission network for the current task. Allocate high-priority tasks to this network link for transmission first.

[0122] Step 4: Task Completion and Summary. When all tasks have been successfully transmitted, verify the integrity and accuracy of each task. Check the checksum information of the data to ensure that no data is lost or damaged. Summarize the multi-task parallel transmission process, analyze the frequency of network switching, switching timing, and the usage efficiency of each network link, etc. Based on the summary results, optimize the evaluation metrics, weight allocation, threshold setting, etc., so as to enable more efficient real-time network switching in future multi-task parallel transmissions.

Claims

1. A user reverse management network method for multi-source heterogeneous networks, characterized by including the following steps: 1). Task and network parameter modeling There are multiple user devices and multiple types of network links in the multi-source heterogeneous network; let the user device set be UE = {1, 2,..., M}, and the total number of user devices be M; the network link set be NL = {1, 2,..., N}, and the total number of network links be N; For a single small task, define its task model as where Q m represents the data volume of the task, represents the maximum tolerable transmission delay of task J m ; Model the performance of the network link, considering three key factors: delay, bandwidth, and packet loss rate: Delay: The delay T m for task J mn transmitted from user device m through network link n can be decomposed into transmission delay node processing delay and queuing delay That is, where the transmission delay r mn represents the data transmission rate between user device m and network link n; the node processing delay C n represents the computing and processing power of the node connected by network link n; the queuing delay Bandwidth: The available bandwidth of network link n is B n , and the bandwidth requirement of task J m during transmission is B mn , and it is necessary to satisfy B mn ≤B n ; Packet loss rate: The packet loss rate of task J m transmitted on network link n is L mn , which is an important indicator to measure the reliability of data transmission; 2). Routing decision-making process The routing decision-making process is a game process between user devices and network resources; The decision-making goal of the user device is to minimize its own cost on the premise of meeting the task QoS requirements; the utility function U m of user device m can be modeled as: Among them, α, β, and γ are weight coefficients, taking values of 0.4, 0.5, and 0.4, respectively, representing the degree of user concern about bandwidth, latency, and packet loss rate; B min,m is the minimum bandwidth requirement for task J m ; P mn is the cost per unit data volume for transmitting task J m using network link n; The decision-making goal of the network service provider is to maximize its own profit on the basis of meeting the user's task requirements; the utility function V n of network link n is expressed as: Among them, ζ is the energy consumption cost coefficient, and the value is set to 0.6, and E n is the energy consumed by network link n to transmit the task; 3). Dynamic adjustment mechanism Define the dynamic adjustment evaluation function R mn , which is used to evaluate whether the transmission of task J m on network link n meets the requirements and the comprehensive performance of the link; this function is constructed based on the QoS requirements of the task and the real-time state of the network link, and the formula is as follows: Among them, ω1, ω2, and ω3 are weight coefficients, representing the relative importance of latency, bandwidth, and packet loss rate in the evaluation, and ω1 + ω2 + ω3 = 1; is the maximum acceptable packet loss rate of task J m ; Single large task multipath parallel transmission method There are n links that can be used for the sender to transmit data packets to the receiver. The network state attributes of each link are related to the end-to-end delay D, link bandwidth B, packet loss rate PLR, the number of data packets M transmitted on the link, network utilization U, and target delay L; therefore, the network state attributes of each link can be described by S i ={D i , B i , PLR i , M i , Y i , L i}, i = 1, 2,..., n. Among them, the end-to-end delay D refers to the time period from the start of data packet transmission by the sender to the receipt of the data packet by the receiver, which is affected by the propagation delay P, queuing delay Q, and transmission delay T, as shown in Equation (1): D i =P i +Q i +T i , i = 1, 2,..., n (1) During the data transmission process, the propagation delay P refers to the time taken for the electromagnetic wave carrying the data packet information to propagate in the physical channel of the communication link. The end-to-end delay is further subdivided into the link absolute delay and the link relative delay T. Among them, the link absolute delay includes the propagation delay P and the queuing delay Q. In actual analysis, the asymmetry of the end-to-end delay that may exist in the two processes of the data packet from the sending end to the receiving end and the confirmation of the received data packet returning from the receiving end to the sending end is ignored, and it is approximately considered that the propagation delay P is about half of the round-trip delay RTT, that is In view of this, the magnitude of the queuing delay can be expressed in the following way: In the formula: represents the average size of the link data packet; C i represents the buffer size of link i; The link relative delay, that is, the transmission delay T, depends on the size of the data being sent, the current bandwidth of the link network, and the additional transmission time of the link error correction mechanism, that is: In the formula, E corr is the additional transmission time coefficient of the link error correction mechanism; Therefore, by combining equations (1)-(4), the end-to-end delay formula for the data packet of the i-th path in multi-path parallel transmission is obtained: In a parallel transmission system, the bottleneck link in the multi-path transmission network is often the key factor affecting the transmission quality and system performance; the bottleneck link refers to the one with the worst network state among the i links in the transmission network. To ensure that the data packet can reach the receiving end smoothly within the target delay L and, on the premise of taking into account the reliability of the limited number of transmissions of the data packet, a retransmission time T needs to be reserved for the bottleneck link re ; that is, it satisfies: In the formula: sign(M i ) is the sign function: In the formula, I link is the importance degree of the link, and its value range is [0,1]. The larger the value, the more important the link; Using the method of linear programming, abstract the link into a mathematical model that includes the current bandwidth, end-to-end delay, network utilization rate, and target delay; by solving this linear programming problem, the maximum link switching time node allowed in the current network environment can be determined, and then the link combination that can minimize the transmission delay can be found; the objective function and constraints of this linear programming problem are as follows: In the above formula, D0 represents the end-to-end delay when the link is in the idle state; in this state, the data packets transmitted on the link do not need to go through the queuing waiting process, and its value is the time required for the data packet to go from the sender to the receiver without queuing waiting; And τ, as a variable parameter, its specific value is closely related to the specific requirements of the real-time transmission service type for the end-to-end delay; the following gives examples of the typical value ranges of τ corresponding to different real-time transmission service types, unit: ms; Calculate T in real time through the feedback data packet information of the receiving end re , according to the calculated T re and the last response time T 1st Compare them, as shown in Algorithm 1; Algorithm 1: Dynamic Link State Awareness Algorithm Input: D i , M i , B i , U i , RTT i , D0, τ, L, i = 1, 2, …, n Output: The network status of each link of link i, the real-time network status of each link, and the specific status types are: active_stable: indicates that the link is in a stable state and can reliably transmit data packets; active_unstable: means that the link state is unstable, which may affect the normal transmission of data packets, and the transmission strategy needs to be adjusted; active_wary: indicates that the link state is in a critical state, and it is necessary to closely monitor and appropriately adjust the data packet distribution to prevent the link performance from deteriorating; 1.

1. Initialization: Obtain the current state of link i, denoted as LinkState; according to the formula Calculate the value of the current moment T re ; Obtain the value of the current moment T 1st ; 1.

2. Condition judgment: If LinkState == active_stable, then make the following judgments: If T re >T 1st , update LinkState to active_unstable; Otherwise, if T re = T 1st , update LinkState to active_wary; If the above conditions are not all satisfied, keep LinkState as active_stable; If LinkState is not active_stable, keep the current value of LinkState unchanged; 1.

3. Return LinkState; In the single large task multi-path parallel transmission method, the link weight factor α i considers multiple important attributes of the link to reflect the degree of advantage of the link in transmitting tasks; after analyzing the link bandwidth B i , packet loss rate PLR i , network utilization U i , the formula is obtained: Among them, Link bandwidth B i directly affects the data transmission speed. The larger the bandwidth, the more data can be transmitted per unit time. In the formula, it is positively correlated with α i That is, the larger the bandwidth, the larger α i is, and the higher the weight of this link in packet distribution; Packet Loss Rate (PLR) i is an important indicator to measure the reliability of data transmission. The lower the packet loss rate, the higher the reliability of data transmission. Through the form of PLR i , a link with a low packet loss rate has a higher α i value; The network utilization rate U i reflects the busy degree of the link. The lower U i , the more idle resources the link has, and the more suitable it is to undertake the task of data packet transmission; In the formula, by in the form of, avoid the situation where the denominator is zero when U i = 0, and at the same time, the lower the network utilization rate, the larger α i is; T in Algorithm 1 1st represents the time period between the current moment and the moment when the last feedback data packet was received; comparing this time period with the maximum link switching time node T re and based on the result, the data packet transmission strategy can be dynamically adjusted: If T re > T 1st , it indicates that the network condition of this link has deteriorated and is no longer suitable for continuous transmission of data packets; at this time, to ensure that the data packets can reach the receiving end within the target delay, it is necessary to transfer the data packets on this link that have not received the acknowledgment response message to a link with good network condition for retransmission, and the specific operation is Algorithm 2; If T re <T 1st , this means that the link network is in good condition, can continue to undertake the task of data packet transmission, and continuously allocate data packets to it; When T re = T 1st it indicates that the network condition of this link is in the critical state of congestion; to avoid the link from getting congested, it is necessary to reduce the number of data packets allocated to this link until its network state is improved; Algorithm 2: Multiplexing algorithm for dynamic links Input: Real-time network status LinkState[i] of each link, T rei , T 1sti , Link bandwidth B i , Packet loss rate PLR i , Network utilization U i , i = 1, 2, …, n Output: Link data packet allocation weight ω i Satisfy Used to guide the allocation ratio of data packets on different links; 2.

1. Initialization: According to the formula Calculate the weight factor α of computing link i i ; 2.

2. Loop judgment, i ranges from 1 to n; If LinkState[i] != active_stable, then assign the weight ω to the data packets of link i i and set it to 0; otherwise, calculate the weight ω assigned to the data packets of link i 2.

3. Return ω i ; Network real-time switching method during multitask parallelism In the multitask parallel scenario, first determine the key metrics for network evaluation, including bandwidth, latency, packet loss rate, and signal strength. Use the analytic hierarchy process to construct a judgment matrix for pairwise comparison of each metric, calculate the weights of each metric and conduct a consistency test; then collect the metric data of each available network to form a decision matrix, standardize this matrix to eliminate the influence of different metric dimensions, and then combine with the determined weights to obtain a weighted standardized matrix; then determine the positive ideal solution and the negative ideal solution. The positive ideal solution is composed of the optimal values of each metric, and the negative ideal solution is composed of the worst values. Calculate the distances from each network to the positive and negative ideal solutions and then obtain the closeness degree. The closer the closeness degree is to 1, the closer the network is to the optimal choice; finally, sort all available networks according to the closeness degree, and switch high-priority tasks to the network with the largest closeness degree.

Images