Network congestion control method and device based on fuzzy rule

Abstract

The invention discloses a network congestion control method and device based on a fuzzy rule, and the method comprises the following steps: setting a corresponding fuzzy rule according to network types which comprise a data center network and a wide area network; network parameters are obtained, fuzzification processing, fuzzy reasoning and defuzzification are carried out on the network parameters, an adjustment factor delta F is obtained, and the network parameters comprise the bandwidth utilization rate, the ECN marking rate and the queue length; obtaining the size of the current congestion window, and adjusting the size of the congestion window in real time according to the delta F; a current network type is detected between the fuzzification processing step and the fuzzy reasoning step, a corresponding fuzzy rule is selected according to the detected network type in the fuzzy reasoning process, and a weight coefficient corresponding to each network parameter is dynamically adjusted. According to the network congestion control method based on the fuzzy rule, the network congestion window can be adjusted in real time, intelligence and automation of network congestion control are achieved, and the network management efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a network congestion control method and device based on fuzzy rules. Background Art

[0002] With the rapid development of Internet technology, the types and number of network services are constantly increasing, which has put forward new requirements for the real-time performance of data transmission. In this context, congestion control algorithms are an indispensable part of network transmission, and their performance directly affects the overall efficiency of the network and user experience.

[0003] Traditional network congestion control algorithms primarily rely on the actual transmission rate and the maximum port bandwidth to calculate bandwidth utilization, and use the packet loss rate to determine the congestion window value. This approach has several significant drawbacks, including computational complexity and the need for manual adjustment of control parameters. Furthermore, it is slow to adapt to network changes in dynamic network environments, particularly in complex environments such as wireless networks and satellite communications. The algorithm cannot be adjusted to suit different network environments, resulting in poor versatility and adaptability. Furthermore, traditional congestion control algorithms struggle to accurately distinguish between latency changes and the causes of decreased link throughput and bandwidth utilization, making it difficult to make appropriate and timely adjustments.

[0004] The above problems make it difficult for traditional congestion control algorithms to meet the transmission requirements of the current network environment, and a more intelligent and flexible solution is urgently needed.

[0005] The disclosure of the above background technology content is only used to assist in understanding the inventive concept and technical solution of the present invention. It does not necessarily belong to the prior art of the present application, nor does it necessarily provide technical guidance. In the absence of clear evidence that the above content has been disclosed before the filing date of the present application, the above background technology should not be used to evaluate the novelty and creativity of the present application. Summary of the Invention

[0006] The purpose of the present invention is to provide a network congestion control method and device based on fuzzy rules, which can adjust the size of the congestion window in real time according to the network status to achieve flexible adjustment under different network environments, thereby improving the timeliness and accuracy of network congestion control.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is as follows: A network congestion control method based on fuzzy rules comprises the following steps: Setting corresponding fuzzy rules according to network types, wherein the network types include data center networks and wide area networks; Obtaining network parameters and performing fuzzification, fuzzy reasoning, and defuzzification on the network parameters to obtain an adjustment factor ΔF, wherein the network parameters include bandwidth utilization, ECN marking rate, and queue length; Obtain the current congestion window size and adjust the congestion window size in real time according to the adjustment factor ΔF; Between the fuzzy processing step and the fuzzy reasoning step, there is also a step of detecting the current network type. During the fuzzy reasoning process, the corresponding fuzzy rule is selected according to the detected network type and the weight coefficients corresponding to various network parameters are dynamically adjusted. When the network type is a data center network, the weight coefficient of the ECN (Explicit Congestion Notification) marking rate is adjusted to be greater than the weight coefficient of the bandwidth utilization rate; when the network type is a wide area network, the weight coefficient of the bandwidth utilization rate is adjusted to be greater than the weight coefficient of the ECN marking rate.

[0008] The present invention introduces fuzzy rules and presets differentiated fuzzy rules based on the characteristics of different network types. It dynamically assigns weight coefficients based on network types and uses a fuzzy reasoning mechanism to adapt to the nonlinear characteristics of network states. By further using the adjustment factor ΔF obtained from the fuzzy rules to calculate the size of the new congestion window, this method helps overcome the lag of traditional linear algorithms in responding to network mutations, effectively solving the problem of traditional methods being slow to respond in dynamic network environments and improving the timeliness and accuracy of network congestion control. Furthermore, the control method of the present invention forms an integrated control chain of "perception-reasoning-decision-execution" through fuzzy logic and network state feedback loops, significantly improving congestion response efficiency in complex network environments and achieving flexible adjustment in different network environments. This effectively solves the problem of poor applicability of traditional methods in data centers and wide area networks, improving the algorithm's versatility and adaptability.

[0009] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the method for obtaining the current congestion window size and adjusting the congestion window size in real time according to the adjustment factor ΔF is: The size of the updated congestion window required for real-time adjustment is calculated according to the following formula: W new =W current +β×ΔF×W max Among them, W current The size of the current congestion window is obtained, W max is the maximum window threshold, β is the preset anti-oscillation coefficient, the value range of β is 30%~50%, ΔF is the adjustment factor, the value range of ΔF is -1~1; W new is the size of the updated congestion window; Then, according to the calculation results, the current congestion window W currentAdjust to the updated congestion window W in real time new , to control network congestion.

[0010] Specifically, W max Determined by the network bandwidth-delay product, that is, W max = Bandwidth (network bandwidth) × RTT (physical link limit); the anti-oscillation coefficient β is set to control the adjustment range, thereby preventing severe window fluctuations and achieving smooth window scaling, balancing throughput and stability. The adjustment factor ΔF ranges from -1 to 1. A positive value indicates that the network is idle and the window needs to be expanded; a negative value indicates that there is a risk of network congestion and the window needs to be reduced.

[0011] Further, based on any one of the above technical solutions or a combination of multiple technical solutions, the W new The value range is greater than or equal to 1 and less than or equal to W max Among them, W new The minimum value is set to 1 to ensure that the sender will not completely stop sending packets even in the most extreme congestion conditions, to ensure protocol liveness and prevent deadlock, which is the basic safety bottom line of congestion control.

[0012] Furthermore, based on any one of the technical solutions or a combination of multiple technical solutions described above, the bandwidth utilization, ECN marking rate, and queue length are all input parameters of fuzzy rules, and the fuzzy output of the fuzzy rules is a fuzzy adjustment level, and the fuzzy adjustment level includes a first level, a second level, a third level, a fourth level, and a fifth level.

[0013] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the method for setting fuzzy rules is: Each network parameter of the bandwidth utilization, ECN marking rate and queue length is set to five levels: low, lower, medium, higher and high. All combinations of the bandwidth utilization level, the ECN marking rate level and the queue length level and the fuzzy adjustment level corresponding to each combination are listed.

[0014] Furthermore, based on any one of the technical solutions or a combination of multiple technical solutions described above, the value range corresponding to the low level of bandwidth utilization is 0~0.3, the value range corresponding to the relatively low level of bandwidth utilization is 0.2~0.45, the value range corresponding to the medium level of bandwidth utilization is 0.35~0.65, the value range corresponding to the relatively high level of bandwidth utilization is 0.55~0.8, and the value range corresponding to the high level of bandwidth utilization is 0.7~1.

[0015] Furthermore, based on any one of the technical solutions or a combination of multiple technical solutions described above, the value range corresponding to the low level of the ECN marking rate is 0-1, the value range corresponding to the relatively low level of the ECN marking rate is 0.5-3, the value range corresponding to the medium level of the ECN marking rate is 2-6, the value range corresponding to the relatively high level of the ECN marking rate is 5-15, and the value range corresponding to the high level of the ECN marking rate is 10-100.

[0016] Furthermore, based on any one of the technical solutions or a combination of multiple technical solutions described above, the value range corresponding to the low level of the queue length is 0~5, the value range corresponding to the relatively low level of the queue length is 3~15, the value range corresponding to the medium level of the queue length is 10~30, the value range corresponding to the relatively high level of the queue length is 20~60, and the value range corresponding to the high level of the queue length is 50~70.

[0017] Furthermore, based on any one or more of the aforementioned technical solutions, when the network type is a data center network, the corresponding fuzzy rule is a first fuzzy rule; when the network type is a wide area network, the corresponding fuzzy rule is a second fuzzy rule, and the fuzzy adjustment levels corresponding to the same combination of bandwidth utilization level, ECN marking rate level, and queue length level in the first and second fuzzy rules may be different or the same. Because data center networks are characterized by high bandwidth and low latency, while wide area networks are characterized by high volatility and long latency, pre-setting differentiated fuzzy rules effectively balances bandwidth utilization efficiency and congestion prevention needs.

[0018] Furthermore, based on any one of the technical solutions or a combination of multiple technical solutions described above, the first fuzzy rule and the second fuzzy rule both include the same number of sub-rules, and the number of all combinations of the bandwidth utilization level, the ECN marking rate level, and the queue length level is equal to the number of sub-rules in the fuzzy rule.

[0019] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the method for detecting the current network type is: Obtain the current network's round-trip delay, delay jitter, bandwidth, queue change rate, and flow duration as input parameters, and set a first scoring variable and a second scoring variable. The first scoring variable is used for feature matching scoring of the data center network, and the second scoring variable is used for feature matching scoring of the wide area network. When the round-trip delay is less than 1ms and the delay jitter is less than 0.2ms, the weight of the first score variable is increased by 0.4; otherwise, the weight of the second score variable is increased by 0.4; When the bandwidth is greater than or equal to 10 Gbps and the queue change rate is greater than 1000 packets / ms, the weight of the first score variable is increased by 0.3; otherwise, the weight of the second score variable is increased by 0.3; When the flow duration is less than 0.1s, the weight of the first score variable is increased by 0.3, otherwise the weight of the second score variable is increased by 0.3; Finally, the network type is determined based on the size relationship between the first score variable and the second score variable. When the first score variable is greater than the second score variable, it is identified as a data center network; when the first score variable is less than the second score variable, it is identified as a wide area network.

[0020] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the method for fuzzy processing the network parameters is: Each network parameter is mapped to its respective level according to the obtained precise value of each network parameter and the fuzzy rule, wherein each network parameter corresponds to at least one level.

[0021] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the method for performing fuzzy reasoning on network parameters is: Preset original weight coefficients of bandwidth utilization, ECN marking rate, and queue length based on the detected current network type, and normalize each original weight coefficient to obtain the corresponding normalized weight coefficient; Calculate the membership degree corresponding to bandwidth utilization, ECN marking rate, and queue length respectively based on the membership function; The activation degree is calculated based on the normalized weight coefficients corresponding to bandwidth utilization, ECN marking rate, and queue length, and their corresponding membership degrees.

[0022] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the normalized weight coefficient is calculated by the following formula:

[0023] in, are the original weight coefficients of bandwidth utilization, ECN marking rate and queue length, respectively, w i Represents the normalized weight coefficient of the i-th network parameter, i=1, 2, 3.

[0024] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the membership degree is calculated by the following formula:

[0025] Where x represents the exact value of the network parameter, a is the width parameter, b is the steepness parameter, c is the center of the membership function, and μ is the membership corresponding to the network parameter. Specifically, the width parameter a can be used to distinguish the overlap width of the fuzzy sets, and the steepness parameter b can range from 2 to 3.

[0026] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, when the network type is a data center network, the is 0.2~0.3, is 0.4~0.6, is 0.2~0.3; when the network type is a wide area network, the is 0.4~0.6, is 0.2~0.3, In the present invention, during the fuzzy reasoning process, original weight coefficients corresponding to various network parameters are dynamically assigned based on different network types. Specifically, when the network type is a data center network, the original weight coefficient of the ECN marking rate is greater than the original weight coefficient of the bandwidth utilization rate; when the network type is a wide area network, the original weight coefficient of the bandwidth utilization rate is greater than the original weight coefficient of the ECN marking rate. This plays a weighted role in the fuzzy rules, allowing the fuzzy rules to better fit the application scenario.

[0027] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the activation degree is calculated by the following formula: α=w1×μ1+w2×μ2+w3×μ3 Where α is the activation degree, w1 is the normalized weight coefficient of bandwidth utilization, and μ1 is the membership degree of bandwidth utilization; w2 is the normalized weight coefficient of ECN marking rate, and μ2 is the membership degree of ECN marking rate; w3 is the normalized weight coefficient of queue length, and μ3 is the membership degree of queue length.

[0028] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the method for defuzzifying the network parameters is: Select the corresponding fuzzy rule based on the current network type detected. In the corresponding fuzzy rule, determine the corresponding sub-rule according to the level of each network parameter obtained during the fuzzification process. Convert the fuzzy adjustment level in the current sub-rule into the corresponding representative value, and then calculate the adjustment factor ΔF according to the following formula:

[0029] Among them, m represents the total number of sub-rules activated under the level combination of each network parameter currently input in the corresponding fuzzy rule, α jis the activation degree of the jth sub-rule among all activated sub-rules (m sub-rules), F j is the representative value of the fuzzy adjustment level corresponding to the j-th sub-rule among all activated sub-rules (m sub-rules).

[0030] Furthermore, based on any one of the technical solutions or the combination of multiple technical solutions described above, when the fuzzy adjustment level is the first level, the corresponding value of Fj is -1, when the fuzzy adjustment level is the second level, the corresponding value of Fj is -0.5, when the fuzzy adjustment level is the third level, the corresponding value of Fj is 0, when the fuzzy adjustment level is the fourth level, the corresponding value of Fj is 0.5, and when the fuzzy adjustment level is the fifth level, the corresponding value of Fj is 1.

[0031] Furthermore, any one of the above-mentioned technical solutions or a combination of multiple technical solutions further includes the following steps: The sender determines the number of packets to be sent simultaneously based on the updated congestion window size, then places the packets into a send queue awaiting transmission while monitoring the network status. The aforementioned steps are executed at preset intervals to obtain a new congestion window size, thereby continuously updating and adjusting the congestion window size in real time. In some embodiments of the present invention, the preset interval is 50-100ms. Furthermore, by determining the number of packets to be sent simultaneously based on the updated congestion window size, buffer pressure on intermediate devices (such as routers) can be adjusted to avoid continued packet loss and increased latency.

[0032] According to another aspect of the present invention, a network congestion control device for performing network congestion control using the network congestion control method described above is provided. The network congestion control device includes a fuzzy decision module and a data analysis module. The fuzzy decision module is configured to detect the current network type and determine the fuzzy adjustment level output in the fuzzy rule based on the acquired network parameters. The data analysis module is configured to receive data packets sent by the transport layer and parse the data packets.

[0033] The beneficial effects brought about by the technical solution provided by the present invention are as follows: This invention uses multi-dimensional network parameter fusion perception, combined with weighted moving average processing of historical data, to construct an accurate network status quantification model. It then applies fuzzy rules to process network status, achieving intelligent and automated network congestion control. This avoids the complexity and inconvenience of manual parameter adjustment in traditional methods, significantly improving network management efficiency. b. This invention pre-defines differentiated fuzzy rules corresponding to the different characteristics of data center networks and wide area networks. By leveraging a multi-judgment mechanism based on fuzzy rules and dynamically assigning weight coefficients based on network type, it achieves flexible adjustment in different network environments. This addresses the issues of traditional methods, such as their slow response in dynamic network environments and their poor applicability in data center networks and wide area networks, thereby improving the algorithm's versatility and adaptability. c. This invention uses nonlinear fuzzy reasoning to fuzzify network parameters into different levels. Fuzzy rules are then used for reasoning, ultimately outputting an adjustment factor ΔF for real-time adjustment of the congestion window size. This overcomes the lag inherent in traditional linear algorithms in responding to sudden network changes, enabling precise and flexible adjustment of the congestion window and significantly improving congestion response efficiency in complex network environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0035] Figure 1 A flowchart of a fuzzy rule-based congestion control method provided by an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0036] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0037] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0038] See also Figure 1 In one embodiment of the present invention, a network congestion control method based on fuzzy rules is provided, comprising the following steps: Step 1: Set the first fuzzy rule corresponding to the data center network and the second fuzzy rule corresponding to the wide area network, respectively. The first fuzzy rule and the second fuzzy rule both include the same number of sub-rules. For the first and second fuzzy rules, bandwidth utilization, ECN marking rate, and queue length are all input parameters, and the fuzzy output of each fuzzy rule is the fuzzy adjustment level.

[0039] Specifically, each network parameter including bandwidth utilization, queue length, and ECN marking rate is set to five levels: low, lower, medium, higher, and very high. The fuzzy adjustment levels include the first level, the second level, the third level, the fourth level, and the fifth level. Then, all combinations of bandwidth utilization levels, queue length levels, and ECN marking rate levels and the fuzzy adjustment levels corresponding to each combination are listed. The number of all combinations of bandwidth utilization levels, queue length levels, and ECN marking rate levels is equal to the number of sub-rules in the fuzzy rule. The first fuzzy rule in the present invention is specifically shown in Table 1, and the second fuzzy rule is specifically shown in Table 2. Both fuzzy rules include 125 sub-rules. As can be seen from Tables 1 and 2, the fuzzy adjustment levels corresponding to the same combination of bandwidth utilization level, queue length level, and ECN marking rate level in the first and second fuzzy rules may be different or the same. This is because the first and second fuzzy rules are set based on data center networks and wide area networks, respectively. These two network types have different characteristics. Data center networks are characterized by high bandwidth and low latency sensitivity, while wide area networks are characterized by high volatility and long latency. By presetting differentiated first and second fuzzy rules in advance, it is possible to effectively balance bandwidth utilization efficiency and congestion prevention needs.

[0040] Among them, for the first fuzzy rule and the second fuzzy rule, the value range of bandwidth utilization of the same level is the same: when the bandwidth utilization level is low, the value range corresponding to the bandwidth utilization level is 0~0.3, when the bandwidth utilization level is relatively low, the value range corresponding to the bandwidth utilization level is 0.2~0.45, when the bandwidth utilization level is medium, the value range corresponding to the bandwidth utilization level is 0.35~0.65, when the bandwidth utilization level is relatively high, the value range corresponding to the bandwidth utilization level is 0.55~0.8, and when the bandwidth utilization level is high, the value range corresponding to the bandwidth utilization level is 0.7~1.

[0041] For the first fuzzy rule and the second fuzzy rule, the value range of the ECN marking rate of the same level is the same: the value range corresponding to the low level of ECN marking rate is 0~1, the value range corresponding to the lower level of ECN marking rate is 0.5~3, the value range corresponding to the medium level of ECN marking rate is 2~6, the value range corresponding to the higher level of ECN marking rate is 5~15, and the value range corresponding to the high level of ECN marking rate is 10~100.

[0042] For the first and second fuzzy rules, the value ranges of queue lengths of the same level are the same: when the queue length level is low, the value range corresponding to the queue length level is 0~5, when the queue length level is relatively low, the value range corresponding to the queue length level is 3~15, when the queue length level is medium, the value range corresponding to the queue length level is 10~30, when the queue length level is relatively high, the value range corresponding to the queue length level is 20~60, and when the queue length level is high, the value range corresponding to the queue length level is 50~70.

[0043]

[0044]

[0045]

[0046]

[0047]

[0048]

[0049] Step 2: Obtain network parameters, and perform fuzzification, fuzzy reasoning and defuzzification on these network parameters to obtain the adjustment factor ΔF.

[0050] Step 2.1: Receive the data packet sent by the network transport layer and parse the data packet to obtain the network parameters of the data packet, including bandwidth utilization, ECN marking rate, and queue length.

[0051] Step 2.2: Fuzzify the network parameters.

[0052] Based on the precise value of each network parameter obtained and the fuzzy rules, each network parameter is mapped to its respective level. Each network parameter corresponds to at least one level. Since the value ranges of different levels for the same network parameter overlap, when the precise value of a network parameter obtained falls within the value ranges of two levels at the same time, it is necessary to list the sub-rules corresponding to the two levels for each network parameter and apply them to the subsequent calculation process. Similarly, regardless of whether one, two, or all three of the three network parameters obtained can correspond to different levels, it is necessary to list all the sub-rules separately.

[0053] Step 2.3: Detect the current network type.

[0054] The round-trip delay, delay jitter, bandwidth, queue change rate, and flow duration of the current network are obtained as input parameters, and a first score variable and a second score variable are set. The first score variable is used for the feature matching score of the data center network, and the second score variable is used for the feature matching score of the wide area network.

[0055] When the round-trip delay is less than 1ms and the delay jitter is less than 0.2ms, the weight of the first score variable is increased by 0.4; otherwise, the weight of the second score variable is increased by 0.4; when the bandwidth is greater than or equal to 10Gbps and the queue change rate is greater than 1000 packets / ms, the weight of the first score variable is increased by 0.3; otherwise, the weight of the second score variable is increased by 0.3; when the flow duration is less than 0.1s, the weight of the first score variable is increased by 0.3; otherwise, the weight of the second score variable is increased by 0.3.

[0056] Finally, the network type is determined based on the size relationship between the first score variable and the second score variable. When the first score variable is greater than the second score variable, it is identified as a data center network. When the first score variable is less than the second score variable, it is identified as a wide area network. In this way, the current network type is determined to facilitate the selection of corresponding fuzzy rules in subsequent calculations.

[0057] Step 2.4: Perform fuzzy inference on network parameters.

[0058] The original weight coefficients of bandwidth utilization, ECN marking rate, and queue length are preset based on the detected current network type. The original weight coefficients are normalized according to the following formula to obtain the corresponding normalized weight coefficients:

[0059] in, are the original weight coefficients of bandwidth utilization, ECN marking rate and queue length, respectively, wi Represents the normalized weight coefficient of the i-th network parameter, i=1, 2, 3.

[0060] When presetting the original weight coefficient, when the network type is a data center network, due to the high sensitivity of the queue, the weight coefficient of the ECN marking rate is adjusted to be greater than the weight coefficient of the bandwidth utilization, and 0.2~0.3, 0.4~0.6, is 0.2~0.3; when the network type is WAN, due to the need for long-term fairness, the weight coefficient of bandwidth utilization is adjusted to be greater than the weight coefficient of ECN marking rate, and 0.4~0.6, 0.2~0.3, The original weight coefficients corresponding to each network parameter are dynamically allocated according to different network types to play a weighted role in the fuzzy rules, so that the fuzzy rules can better fit the application scenario.

[0061] Then, based on the membership function, the membership corresponding to bandwidth utilization, ECN marking rate, and queue length is calculated using the following formula:

[0062] Among them, x represents the exact value of the obtained network parameter, a is the width parameter, which can be used to distinguish the overlapping width of the fuzzy set; b is the steepness parameter, and the value range of b is 2~3; c is the center of the membership function, and μ is the membership corresponding to the network parameter.

[0063] The activation degree is then calculated using the following formula based on the normalized weight coefficients corresponding to bandwidth utilization, ECN marking rate, and queue length, and their corresponding membership degrees: α=w1×μ1+w2×μ2+w3×μ3 Where α is the activation degree, w1 is the normalized weight coefficient of bandwidth utilization, and μ1 is the membership degree of bandwidth utilization; w2 is the normalized weight coefficient of ECN marking rate, and μ2 is the membership degree of ECN marking rate; w3 is the normalized weight coefficient of queue length, and μ3 is the membership degree of queue length.

[0064] Step 2.5: Defuzzify the network parameters.

[0065] The first fuzzy rule or the second fuzzy rule is selected according to the current network type detected. In the corresponding fuzzy rule, the corresponding sub-rule is confirmed according to the level corresponding to each network parameter obtained in the fuzzification process, and the fuzzy adjustment level in the current sub-rule is converted into a corresponding representative value: when the fuzzy adjustment level is the first level, the corresponding representative value is -1; when the fuzzy adjustment level is the second level, the corresponding representative value is -0.5; when the fuzzy adjustment level is the third level, the corresponding representative value is 0; when the fuzzy adjustment level is the fourth level, the corresponding representative value is 0.5; when the fuzzy adjustment level is the fifth level, the corresponding representative value is 1; and then the adjustment factor ΔF is calculated according to the following formula:

[0066] Among them, m represents the total number of sub-rules activated under the level combination of each network parameter currently input in the corresponding fuzzy rule, α j is the activation degree of the jth sub-rule among the m activated sub-rules, F j is the representative value of the fuzzy adjustment level corresponding to the j-th sub-rule among the m activated sub-rules.

[0067] Step 3: Get the current congestion window W current The size of the congestion window W required for real-time adjustment is calculated. new The size of the current congestion window W is calculated based on the result. current Adjust to the updated congestion window W in real time new , W new Calculated according to the following formula: W new =W current +β×ΔF×W max Among them, W current The size of the current congestion window is obtained, W max is the maximum window threshold, which is determined by the network bandwidth-delay product, that is, W max =Bandwidth×RTT; β is the preset anti-oscillation coefficient, and the value range of β is 30%~50%. Introducing the anti-oscillation coefficient β in this calculation formula can control the adjustment range, prevent the window from violent oscillation, and achieve smooth window scaling; ΔF is the adjustment factor, and the value range of ΔF is -1~1; W new is the updated congestion window size, W new The value range is greater than or equal to 1 and less than or equal to W max .

[0068] Step 4: Based on the updated congestion window size, the sender determines the number of packets to be sent simultaneously, places the packets in the send queue, and monitors the network status. Steps 2 and 3 are executed at a preset interval to obtain the new congestion window size, thereby continuously updating and adjusting the congestion window size in real time. The preset interval is 50-100ms.

[0069] Another embodiment of the present invention provides a device for network congestion control using the above-mentioned network congestion control method. By setting up a fuzzy decision module and a data analysis module, multi-dimensional network parameters such as bandwidth utilization, ECN marking rate, and queue length are analyzed in real time. Weight coefficients are dynamically assigned based on network type, thereby achieving precise and flexible adjustment of the congestion window.

[0070] Specifically, the fuzzy decision module in the network congestion control device is configured to detect the current network type and determine the fuzzy adjustment level output in the fuzzy rules based on the obtained network parameters. The fuzzy decision module includes a fuzzification processor, a knowledge base, an inference engine, and a defuzzifier. The fuzzification processor converts input network parameters (bandwidth utilization, ECN marking rate, and queue length) into linguistic variables (i.e., the corresponding levels of the network parameters: low, lower, medium, higher, and high). It also calculates the precise values of the input network parameters based on a preset membership function to obtain the corresponding membership. The knowledge base stores preset first and second fuzzy rules. The inference engine matches the input variables (membership values and normalized weight coefficients) with the fuzzy rules, calculates the activation degree of each sub-rule, and determines the output fuzzy set. The defuzzifier converts the fuzzy output (fuzzy adjustment level) into a clear numerical output (adjustment factor ΔF).

[0071] The data analysis module is configured to receive and parse data packets sent by the transport layer. It includes an ECN marking counter, a total ACK counter, an ECN marking rate calculator, a port utilization collector, an ECN field parser, a network layer round-trip delay and packet loss rate detector, and a queue smoothing calculator. The ECN marking counter continuously records the number of ECN-marked packets received per unit time. It parses the ECN field in the IP packet header to determine whether a packet is congestion-marked and accumulates the statistics within each time window. The total ACK (acknowledgement packet) counter records the total number of ACKs received during the current observation period, providing an accurate denominator for subsequent calculations of the ECN marking rate, reflecting the proportion of marked congestion in the network. The ECN marking rate calculator calculates the ECN marking rate per unit time (the number of ECN-marked packets divided by the total number of ACK packets). The port utilization collector obtains real-time bandwidth usage on ports of network devices (such as switches and routers) to reflect current link utilization. The ECN field parser is used to analyze the ECN field status in data packets in real time. By checking the field bits on each inbound or outbound packet, it records congestion trends and assists in determining path health and whether explicit congestion feedback exists on the current path. The network layer round-trip delay and packet loss rate acquirer estimates current delay and packet loss using methods such as the TCP (Transmission Control Protocol) timestamp option, ICMP (Control Message Protocol) echo requests, link-layer statistics, or stack measurements. This provides a basis for dynamic window adjustment, thereby monitoring and acquiring the round-trip delay and packet loss rate of the path. The queue smoothing calculator uses historical records to perform weighted smoothing on the queue length (using a weighted moving average to incorporate the current queue length) to mitigate misjudgments caused by transient fluctuations.

[0072] The fuzzy rule-based congestion control method of the present invention introduces fuzzy rules and pre-defines differentiated fuzzy rules based on the characteristics of different network types. It dynamically assigns weight coefficients based on network types and uses a fuzzy inference mechanism to adapt to the nonlinear characteristics of network states. By further using the adjustment factor ΔF derived from the fuzzy rules to calculate the new congestion window size, this method overcomes the lag inherent in traditional linear algorithms in responding to sudden network changes, effectively addressing the sluggish response of traditional methods in dynamic network environments and improving the timeliness and accuracy of network congestion control. Furthermore, by forming a closed loop with fuzzy logic and network state feedback, it significantly improves congestion response efficiency in complex network environments, enabling flexible adjustment in diverse network environments and enhancing the algorithm's versatility and adaptability.

[0073] Due to the advanced nature of this technical solution, the technology of the present invention can be widely used in scenarios such as cloud computing data centers, real-time video transmission, and large-scale Internet of Things. In particular, in 5G edge computing and SDN network architectures, it can significantly reduce network latency fluctuations and improve link utilization in high-concurrency scenarios. At the same time, through the dynamic window adjustment mechanism, it avoids the "sawtooth" throughput fluctuations of the traditional TCP protocol, providing a low-latency, high-throughput, and highly adaptive core congestion control solution for the new generation of intelligent networks.

[0074] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0075] The above is only a specific implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A network congestion control method based on fuzzy rules, characterized in that: The steps include: Setting corresponding fuzzy rules according to network types, wherein the network types include data center networks and wide area networks; Obtaining network parameters and performing fuzzification, fuzzy reasoning, and defuzzification on the network parameters to obtain an adjustment factor ΔF, wherein the network parameters include bandwidth utilization, ECN marking rate, and queue length; Obtain the current congestion window size and adjust the congestion window size in real time according to the adjustment factor ΔF; Between the fuzzy processing step and the fuzzy reasoning step, there is also a step of detecting the current network type. During the fuzzy reasoning process, the corresponding fuzzy rule is selected according to the detected network type and the weight coefficients corresponding to various network parameters are dynamically adjusted. When the network type is a data center network, the weight coefficient of the ECN marking rate is adjusted to be greater than the weight coefficient of the bandwidth utilization rate; when the network type is a wide area network, the weight coefficient of the bandwidth utilization rate is adjusted to be greater than the weight coefficient of the ECN marking rate.

2. The network congestion control method according to claim 1, characterized in that: The method for obtaining the current congestion window size and adjusting the congestion window size in real time according to the adjustment factor ΔF is: The size of the updated congestion window required for real-time adjustment is calculated according to the following formula: IN new =W current +β×ΔF×W max Among them, W current The size of the current congestion window is obtained, W max is the maximum window threshold, β is the preset anti-oscillation coefficient, the value range of β is 30%~50%, ΔF is the adjustment factor, the value range of ΔF is -1~1; W new is the size of the updated congestion window; Then, according to the calculation results, the current congestion window W current Adjust to the updated congestion window W in real time new , to control network congestion.

3. The network congestion control method according to claim 2, characterized in that: The W new The value range is greater than or equal to 1 and less than or equal to W max .

4. The network congestion control method according to claim 1, wherein: The bandwidth utilization, ECN marking rate and queue length are all input parameters of fuzzy rules, and the fuzzy output of the fuzzy rules is a fuzzy adjustment level, which includes a first level, a second level, a third level, a fourth level and a fifth level.

5. The network congestion control method according to claim 4, characterized in that: The method of setting fuzzy rules is: Each network parameter of the bandwidth utilization, ECN marking rate and queue length is set to five levels: low, lower, medium, higher and high. All combinations of the bandwidth utilization level, the ECN marking rate level and the queue length level and the fuzzy adjustment level corresponding to each combination are listed.

6. The network congestion control method according to claim 5, characterized in that: When the level of bandwidth utilization is low, the corresponding value range is 0~0.3; when the level of bandwidth utilization is relatively low, the corresponding value range is 0.2~0.45; when the level of bandwidth utilization is medium, the corresponding value range is 0.35~0.65; when the level of bandwidth utilization is relatively high, the corresponding value range is 0.55~0.8; when the level of bandwidth utilization is high, the corresponding value range is 0.7~1.

7. The network congestion control method according to claim 5, characterized in that: When the level of the ECN marking rate is low, the corresponding value range is 0~1; when the level of the ECN marking rate is relatively low, the corresponding value range is 0.5~3; when the level of the ECN marking rate is medium, the corresponding value range is 2~6; when the level of the ECN marking rate is relatively high, the corresponding value range is 5~15; when the level of the ECN marking rate is high, the corresponding value range is 10~100.

8. The network congestion control method according to claim 5, characterized in that: When the level of the queue length is low, the corresponding value range is 0~5; when the level of the queue length is relatively low, the corresponding value range is 3~15; when the level of the queue length is medium, the corresponding value range is 10~30; when the level of the queue length is relatively high, the corresponding value range is 20~60; when the level of the queue length is high, the corresponding value range is 50~70.

9. The network congestion control method according to claim 5, characterized in that: When the network type is a data center network, the corresponding fuzzy rule is the first fuzzy rule; when the network type is a wide area network, the corresponding fuzzy rule is the second fuzzy rule. The fuzzy adjustment levels corresponding to the same combination of bandwidth utilization level, ECN marking rate level, and queue length level in the first fuzzy rule and the second fuzzy rule are different or the same.

10. The network congestion control method according to claim 9, characterized in that: The first fuzzy rule and the second fuzzy rule both include the same number of sub-rules, and the number of all combinations of the bandwidth utilization level, the ECN marking rate level, and the queue length level is equal to the number of sub-rules in the fuzzy rule.

11. The network congestion control method according to claim 1, characterized in that: The method for detecting the current network type is: Obtain the current network's round-trip delay, delay jitter, bandwidth, queue change rate, and flow duration as input parameters, and set a first scoring variable and a second scoring variable. The first scoring variable is used for feature matching scoring of the data center network, and the second scoring variable is used for feature matching scoring of the wide area network. When the round-trip delay is less than 1ms and the delay jitter is less than 0.2ms, the weight of the first score variable is increased by 0.4; otherwise, the weight of the second score variable is increased by 0.4; When the bandwidth is greater than or equal to 10 Gbps and the queue change rate is greater than 1000 packets / ms, the weight of the first score variable is increased by 0.3; otherwise, the weight of the second score variable is increased by 0.3; When the flow duration is less than 0.1s, the weight of the first score variable is increased by 0.3, otherwise the weight of the second score variable is increased by 0.3; Finally, the network type is determined based on the size relationship between the first score variable and the second score variable. When the first score variable is greater than the second score variable, it is identified as a data center network; when the first score variable is less than the second score variable, it is identified as a wide area network.

12. The network congestion control method according to claim 5, characterized in that: The method for fuzzifying network parameters is as follows: Each network parameter is mapped to its respective level according to the obtained precise value of each network parameter and the fuzzy rule, wherein each network parameter corresponds to at least one level.

13. The network congestion control method according to claim 12, characterized in that: The method for performing fuzzy reasoning on network parameters is: Preset original weight coefficients of bandwidth utilization, ECN marking rate, and queue length based on the detected current network type, and normalize each original weight coefficient to obtain the corresponding normalized weight coefficient; Calculate the membership degree corresponding to bandwidth utilization, ECN marking rate, and queue length respectively based on the membership function; The activation degree is calculated based on the normalized weight coefficients corresponding to bandwidth utilization, ECN marking rate, and queue length, and their corresponding membership degrees.

14. The network congestion control method according to claim 13, characterized in that: The normalized weight coefficient is calculated by the following formula: ； in, are the original weight coefficients of bandwidth utilization, ECN marking rate and queue length, respectively, w i Represents the normalized weight coefficient of the i-th network parameter, i=1, 2, 3.

15. The network congestion control method according to claim 14, characterized in that: The membership degree is calculated by the following formula: ； Among them, x represents the exact value of the obtained network parameter, a is the width parameter, b is the steepness parameter, c is the center of the membership function, and μ is the membership corresponding to the network parameter.

16. The network congestion control method according to claim 14, characterized in that: When the network type is a data center network, the is 0.2~0.3, is 0.4~0.6, is 0.2~0.3; when the network type is a wide area network, the is 0.4~0.6, is 0.2~0.3, It is 0.2~0.

3.

17. The network congestion control method according to claim 15, wherein: The activation degree is calculated by the following formula: α=w1×μ1+w2×μ2+w3×μ3 Where α is the activation degree, w1 is the normalized weight coefficient of bandwidth utilization, and μ1 is the membership degree of bandwidth utilization; w2 is the normalized weight coefficient of ECN marking rate, and μ2 is the membership degree of ECN marking rate; w3 is the normalized weight coefficient of queue length, and μ3 is the membership degree of queue length.

18. The network congestion control method according to claim 17, characterized in that: The method for defuzzifying network parameters is: Select the corresponding fuzzy rule based on the current network type detected. In the corresponding fuzzy rule, determine the corresponding sub-rule according to the level of each network parameter obtained during the fuzzification process. Convert the fuzzy adjustment level in the current sub-rule into the corresponding representative value, and then calculate the adjustment factor ΔF according to the following formula: ； Among them, m represents the total number of sub-rules activated under the level combination of each network parameter currently input in the corresponding fuzzy rule, α j is the activation degree of the jth sub-rule among all activated sub-rules, F j is the representative value of the fuzzy adjustment level corresponding to the j-th sub-rule among all activated sub-rules.

19. The network congestion control method according to claim 18, characterized in that: When the fuzzy adjustment level is the first level, the corresponding Fj value is -1; when the fuzzy adjustment level is the second level, the corresponding Fj value is -0.5; when the fuzzy adjustment level is the third level, the corresponding Fj value is 0; when the fuzzy adjustment level is the fourth level, the corresponding Fj value is 0.5; when the fuzzy adjustment level is the fifth level, the corresponding Fj value is 1.

20. A network congestion control device, characterized in that: The network congestion control device performs network congestion control using the network congestion control method described in any one of claims 1 to 19. The network congestion control device includes a fuzzy decision module and a data analysis module. The fuzzy decision module is configured to detect the current network type and determine the fuzzy adjustment level output in the fuzzy rule based on the acquired network parameters. The data analysis module is configured to receive data packets sent by the transport layer and parse the data packets.

Images