RP-CP-NP congestion signal enhancement method and system for large-scale power data transmission, medium and processor

Abstract

The invention discloses an RP-CP-NP congestion signal enhancement method and system for large-scale power data transmission, a medium and a processor, relates to the technical field of congestion control, and solves the problems of poor adaptability, slow response and inaccurate evaluation of an existing method. The method comprises the following steps: constructing an RP-CP-NP architecture containing an intelligent network card; the CP node collects and preprocesses queue length, RTT and link utilization rate signals; the CP node calculates a congestion evaluation value, grades and differentiated ECN marks; the intelligent network card generates and filters CNP; and the intelligent network card dynamically calls the CNP sending frequency according to the congestion level. The system comprises a construction module, an acquisition module, a marking module, a generation module and an adjustment module. The method improves the congestion assessment accuracy, accelerates the response speed, adapts to the power data transmission characteristics, guarantees the transmission real-time performance and stability, and is suitable for a large-scale power data transmission scene.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of congestion control, in particular to an RP-CP-NP congestion signal enhancement method, system, medium and processor for large-scale power data transmission. BACKGROUND

[0002] With the rapid development of power communication technology, large-scale power data transmission puts forward higher requirements on network performance. In order to meet the demand of high bandwidth and low delay, remote direct data storage (RDMA) technology is widely used in data center network (DCN). However, in the process of RDMA transmission, even a single data packet loss will greatly reduce the network throughput, seriously affecting the performance of application services. In order to ensure efficient and reliable RDMA data transmission, data center Ethernet network widely deploys priority-based flow control (PFC) mechanism to prevent buffer overflow.

[0003] However, the existing congestion control method still has some problems in practical application. The traditional congestion control algorithm cannot adapt to complex and interference information, the rate-based algorithm will sacrifice the delay, the adjustment time is long, and the machine learning-based algorithm needs to be solved in calculation and data transmission in network. In addition, the existing congestion control protocol is mainly based on reactive congestion detection, and the network congestion detection speed is slow, and it is difficult to obtain the appropriate rate in each process.

[0004] Therefore, an RP-CP-NP congestion signal enhancement method, system, medium and processor for large-scale power data transmission are needed. SUMMARY

[0005] In view of the slow congestion detection speed and the problem of being unable to adapt to complex and interference information in the prior art, the present application provides an RP-CP-NP congestion signal enhancement method, system, medium and processor for large-scale power data transmission, which can improve the congestion evaluation accuracy, optimize the congestion early warning efficiency, speed up the congestion response speed, and adapt to the dynamic change of congestion. The specific technical solutions are as follows: An RP-CP-NP congestion signal enhancement method for large-scale power data transmission, comprising: S1: constructing an RP-CP-NP congestion signal enhancement system architecture, and setting an intelligent network card at an NP node; S2: collecting and preprocessing multi-dimensional congestion signals at a CP node; S3: obtaining congestion evaluation values based on the multi-dimensional congestion signals at the CP node, and dividing congestion levels, and performing ECN differentiated marking of data packets according to the congestion levels; S4: generating CNP and filtering according to the ECN marking of data packets by the intelligent network card; S5: The intelligent network card dynamically adjusts the frequency of sending CNP to the RP node according to the congestion level.

[0006] Further, in step S2, the collection and preprocessing of the multi-dimensional congestion signals at the CP node include the following steps: S21: Real-time collection of queue length signals, RTT signals and link utilization signals related to network congestion at the intermediate network node; S22: Preprocessing of the collected signals at the intermediate network node to eliminate the influence of noise and outliers on subsequent congestion judgment.

[0007] Further, in step S3, the CP node obtains congestion evaluation values based on multi-dimensional congestion signals and divides congestion levels, and performs ECN differentiated marking of data packets according to the congestion levels, including the following steps: S31: Calculate the congestion evaluation value by taking the preprocessed multi-dimensional congestion signal as input, and divide the congestion level according to the congestion evaluation value; S32: According to different congestion levels, formulate differentiated ECN marking rules to realize the optimization of ECN marking strategy.

[0008] Further, in step S31, the congestion evaluation value formula is as follows: ; Wherein, is the normalized queue length signal; is the normalized RTT signal; is the normalized link utilization rate signal; is a nonlinear signal enhancement function, ; are the queue length signal, RTT signal and link utilization rate signal change rates respectively; is a scene correction coefficient.

[0009] Further, the nonlinear signal enhancement function The formula is as follows: ; Wherein, is the normalized queue length signal, RTT signal or link utilization rate signal.

[0010] Further, the calculation formula of the signal change rate is as follows: ; Wherein, is the sampling period of the corresponding signal; ; is the signal change rate, .

[0011] Further, in step S5, the intelligent network card dynamically adjusts the frequency of sending CNP to the RP node according to the congestion level, including the following steps: S51: setting an initial CNP sending frequency for different congestion levels according to the initial state of the network; S52: the intelligent network card monitors the change of network congestion state in real time, and dynamically adjusts the CNP sending frequency, and the calculation formula of the CNP sending frequency is as follows: is the final CNP sending frequency; is the initial CNP sending frequency reference value; is the proportion of data packets with ECN marking; is the target ECN marking proportion corresponding to the congestion level; is the rounding to the nearest integer of the calculation result.

[0012] A RP-CP-NP congestion signal enhancement system for large-scale power data transmission, applied to the above-mentioned RP-CP-NP congestion signal enhancement method for large-scale power data transmission, comprising: A construction module for constructing the RP-CP-NP congestion signal enhancement system architecture, and setting an intelligent network card at the NP node; An acquisition module for acquiring and preprocessing multi-dimensional congestion signals at the CP node; A marking module for the CP node to obtain congestion evaluation values based on multi-dimensional congestion signals and divide congestion levels, and to perform ECN differentiated marking on data packets according to the congestion levels; A generation module for the intelligent network card to generate CNP and filtering according to the data ECN marked packets; An adjustment module for the intelligent network card to dynamically adjust the frequency of sending CNP to the RP node according to the congestion level.

[0013] A computer readable storage medium, comprising a stored program, wherein the program controls the device where the computer readable storage medium is located to execute the above-mentioned RP-CP-NP congestion signal enhancement method for large-scale power data transmission when the program is running.

[0014] A processor for running a program, wherein the program performs the above-mentioned RP-CP-NP congestion signal enhancement method for large-scale power data transmission when the program is running.

[0015] Compared with the prior art, the beneficial effects of the present application are: 1. Improve congestion assessment accuracy: Break through the limitations of existing single signal judgment, collect multi-dimensional signals such as queue length, RTT, and link utilization rate, eliminate noise after preprocessing, and calculate congestion assessment value through formula containing nonlinear enhancement, signal change rate, and scene correction coefficient. The accuracy of congestion level determination is more than 95% (traditional <80%), avoiding misjudgment.

[0016] 2. Optimize congestion warning efficiency: Abandon fixed ECN marking rules, according to no / light / medium / heavy congestion levels, respectively use no marking, 20% / 50% / 100% probability marking strategy, balance warning timeliness and transmission efficiency, adapt to differentiated needs of power data.

[0017] 3. Speed up congestion response speed: NP node sets intelligent network card, 1us hardware generates CNP (traditional software needs 10us+), real-time filters invalid CNP, 1000 end concurrent scene, effective CNP amount increases by 50%+, guarantees sending end timely congestion notification.

[0018] 4. Adapt to dynamic changes in congestion: Initially set 10 / 30 / 50 CNP sending frequency per second according to congestion level, then combine congestion assessment value change and sending end response, dynamically adjust frequency through formula, avoid insufficient or excessive notification, reduce rate shock.

[0019] 5. Adapt to power scene needs: Scene correction coefficient adapts to power data burstiness, differentiated strategy guarantees core data transmission, makes power data transmission delay reduce by 30%+, packet loss rate <0.5% (traditional >2%), meets low delay and high reliability requirements. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or prior art description. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, each element or part is not necessarily drawn according to the actual proportion.

[0021] Figure 1 It is a flowchart of a RP-CP-NP congestion signal enhancement method for large-scale power data transmission; Figure 2 It is a structural diagram of a RP-CP-NP congestion signal enhancement system for large-scale power data transmission. DETAILED DESCRIPTION

[0022] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are some of the embodiments of the present application but not all the embodiments of the present application. Based on the embodiments of the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative effort should fall within the scope of the present application.

[0023] It should be understood that the terms "comprising" and "including" as used in the present application indicate the presence of the described features, integers, steps, operations, elements, and / or components but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0024] It should also be understood that the terms used in the present application are merely for the purpose of describing particular embodiments and are not intended to limit the present application. As used in the specification and the appended claims of the present application, the singular forms "a", "an" and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0025] It should be further understood that the term "and / or" as used in the present application means any combination of one or more of the associated listed items and all possible combinations thereof, and includes these combinations.

[0026] Embodiment one The RP-CP-NP mechanism is a framework for data center network congestion control, commonly used in RoCE (RDMA over Converged Ethernet) networks, such as the DCQCN congestion control protocol. RP-CP-NP represents different network node roles, and their specific meanings are as follows: RP (Reaction point): reaction node, usually the sender. After receiving the CNP (Congestion Notification Packet) sent by the NP, it will passively accept the CNP packet and implement flow control measures according to a specific algorithm to avoid congestion, such as reducing the sending rate.

[0027] CP (Congestion point): congestion point, generally refers to a switch. When the network is congested, the CP will use specific congestion detection algorithms, such as marking data packets as ECN (Explicit Congestion Notification) using the RED (Random Early Detection) function when the length of the export queue exceeds the set threshold.

[0028] NP (Notification point): notification node, usually the receiver. After the NP receives the interface marked with the ECN field of the IP packet, it actively sends the CNP packet to the opposite end, i.e. the sender RP, to tell the opposite end that congestion occurs in the network and congestion control is needed.

[0029] The RP-CP-NP mechanism realizes effective control of network congestion through the cooperative work of the three, enables the sender to adjust the sending rate according to the congestion state of the network, thereby avoiding the aggravation of network congestion and guaranteeing the efficient and stable operation of the network.

[0030] As shown in Figure 1 The present application provides a RP-CP-NP congestion signal enhancement method for large-scale power data transmission, and the specific implementation steps are as follows: S1: Construct the RP-CP-NP congestion signal enhancement system architecture, and set an intelligent network card at the NP node. Build an RP-CP-NP congestion signal enhancement system including a data sending end (RP), a data receiving end (NP), an intermediate network node (CP) and an intelligent network card. The data sending end is used for sending large-scale power data; the data receiving end is used for receiving power data and feeding back related congestion information; the intermediate network node (such as a router or a switch) is used for forwarding power data and collecting network congestion related signals; the intelligent network card is integrated in the data receiving end and is used for realizing the rapid generation, filtering and dynamic adjustment of the sending frequency of the CNP (Congestion Notification Packet). S2: Collect and pretreat multi-dimensional congestion signals at the CP node. S21: Deploy a signal collection module in the intermediate network node to collect multi-dimensional signals related to network congestion in real time, specifically including queue length signals, RTT (Round Trip Time) signals and link utilization rate signals. Queue length signal collection: through the queue management module of the intermediate network node, the length data of the data packet queue in the node is obtained in real time, and the sampling frequency is set to once every 10 ms to ensure that the change of the queue length can be captured in time. RTT signal collection: the data sending end sends a probe data packet with a timestamp, the data receiving end records the receiving time after receiving the probe data packet, and returns the receiving time information to the data sending end through a feedback data packet, the data sending end calculates the RTT according to the difference between the sending time and the receiving time, and then sends the RTT data to the signal collection module of the intermediate network node, and the RTT collection period is set to once every 50 ms. Link utilization signal acquisition: The link monitoring module of the intermediate network node is used to monitor the bandwidth usage of the network link in real time, calculate the ratio of the actual bandwidth used by the link to the total bandwidth of the link, and obtain the link utilization. The sampling frequency is once every 20ms. S22: Set up a signal preprocessing module in the intermediate network node to preprocess the collected multi-dimensional congestion signals to eliminate the impact of noise and outliers on subsequent congestion judgment. For the queue length signal, a moving average filtering algorithm is used, with a sliding window size of 5. The average of the 5 consecutively collected queue length data is calculated to obtain the filtered queue length signal. For RTT signals, outliers are removed using the 3σ criterion. This involves calculating the mean μ and standard deviation σ of the RTT data, identifying RTT data that exceed the range of [μ-3σ, μ+3σ] as outliers and removing them, and then performing linear interpolation on the remaining RTT data to obtain a complete RTT signal sequence. For the link utilization signal, a median filtering algorithm is used to select the median of seven consecutive link utilization data as the link utilization value at the current moment, so as to reduce the interference of sudden noise on the link utilization signal. S3: CP nodes obtain congestion assessment values ​​based on multi-dimensional congestion signals and classify congestion levels, and perform ECN (Explicit Congestion Notification) differential marking of data packets according to the congestion level.

[0031] S31: Use the preprocessed multi-dimensional congestion signal as input to calculate the congestion assessment value, and classify the congestion level according to the congestion assessment value. First, the signal for each dimension is normalized, mapping its value range to the interval [0,1]. The normalization formula is as follows: ; Where x is the original signal value; This is the minimum value of the signal; This is the maximum value of the signal. The congestion assessment value is calculated using the following formula: ; in, This is the normalized queue length signal; The normalized RTT signal; This is the normalized link utilization signal; It is a nonlinear signal enhancement function. Enhance signal sensitivity in high-congestion-risk areas; These are the rates of change of queue length signal, RTT signal, and link utilization signal, respectively, reflecting the dynamic trend of congestion. This is a scenario correction coefficient to adapt to the differences between "burst traffic" and "stable traffic" in power data transmission.

[0032] Nonlinear signal enhancement function To address different congestion characteristics in power data transmission, such as queue backlog, RTT surge, and link saturation, a piecewise function is used to amplify high-risk signals and avoid misjudging low-congestion intervals. ; Example: If (Moderate queue congestion), then Compared to linear mapping, it more accurately reflects the state of "queue saturation", which aligns with the document's requirement to "avoid misjudging congestion from a single signal".

[0033] The rate of change of the signal reflects the dynamic trend of congestion, and the formula is as follows: ; The sampling period of the corresponding signal (unit: ms), such as , , Result unit: Positive numbers indicate an increasing signal (congestion worsens), while negative numbers indicate a decreasing signal (congestion eases). .

[0034] Example: If time , ms ,but The value 's' indicates that RTT is rising rapidly, and the risk of congestion needs to be closely monitored.

[0035] Scene correction coefficient To address the discrepancy between "burst flow (such as fault monitoring data)" and "stable flow (such as routine monitoring data)" in power data transmission, signal fusion bias is corrected to avoid missed detections in extreme scenarios. ; Principle: The "maximum / minimum normalized signal difference" is used to determine whether the scenario is "extreme"—the larger the difference (e.g., a sudden increase in one signal while others remain unchanged), the more likely it is to be an extreme scenario. The closer the value is to 1.2, the larger the evaluation value should be to avoid missed detections; the smaller the difference (the more consistent the signal trend), the better. Approaching 1, the evaluation value remains stable.

[0036] Example: If , , (If the queue surges while other signals lag), then By strengthening the impact of queue congestion through the correction coefficient, it reflects the core logic of "multi-dimensional signal collaborative judgment of congestion".

[0037] Congestion levels are determined based on congestion assessment values: when C < 0.3, it is classified as no congestion; when 0.3 ≤ C < 0.6, it is classified as mild congestion; when 0.6 ≤ C < 0.8, it is classified as moderate congestion; and when C ≥ 0.8, it is classified as severe congestion. S32: Formulate differentiated ECN (Explicit Congestion Notification) marking rules based on different congestion levels to optimize the ECN marking strategy. No congestion level (C<0.3): No ECN marking is applied to data packets, ensuring efficient data transmission. Mild congestion level (0.3≤C<0.6): Set the ECN marking probability to 20%, that is, randomly select 20% of the packets for ECN marking, to give a slight indication of potential congestion. Moderate congestion level (0.6≤C<0.8): Increase the ECN marking probability to 50%. By increasing the proportion of ECN-marked packets, the congestion warning is enhanced, prompting appropriate adjustments to the sending rate. Severe congestion level (C≥0.8): Set the ECN marking probability to 100%, mark all data packets passing through intermediate network nodes with ECN, strongly notify the network that it is in a state of severe congestion, and require it to significantly reduce the sending rate. S4: The smart NIC generates a CNP (Congestion Notification Packet) based on the data ECN tag packet and filters it.

[0038] ECN marking is performed by intermediate network nodes (CPs, such as switches). When a CP determines network congestion based on multi-dimensional signals (queue length, RTT, link utilization), it marks data packets with ECN according to a differentiated strategy (e.g., marking 20% ​​of packets with mild congestion) to indicate that the packet has passed through a congested node. ECN-marked data packets are first transmitted to the data receiver (NP) or smart NIC, rather than directly to the sender. After receiving the ECN-marked data packet, the NP or smart NIC generates a Congestion NP (CNP) using hardware acceleration and filters out invalid CNPs (such as checksum errors or invalid IPs). The CNP is then forwarded to the sender (RP). By receiving the CNP, the sender is aware that network congestion has occurred and adjusts its transmission rate based on the congestion level information (e.g., mild or severe) in the CNP.

[0039] S41: When the smart network card receives a data packet, it quickly extracts key information (such as source IP address, destination IP address, port number, ECN tag, etc.) from the data packet through hardware circuitry, and automatically generates a CNP according to the proportion of data packets with ECN tags and a preset CNP format template. The default CNP format template includes a version number, protocol type, source IP address field, destination IP address field, port number field, congestion level field, and checksum field. When generating a CNP, the smart NIC hardware module directly extracts the source IP address, destination IP address, and port number from ECN-tagged packets and fills them into the corresponding fields. It determines the congestion level based on the proportion of packets with ECN tags and fills it into the congestion level field. It automatically calculates the checksum and fills it into the checksum field, completing the rapid generation of the CNP in less than 1μs, far faster than traditional software-generated CNPs (typically over 10μs). S42: The smart NIC filters the generated CNPs, removing invalid CNPs to prevent them from consuming network bandwidth resources. The unpredictability of invalid CNPs: The invalidity of CNPs (such as checksum errors or invalid target IPs) often stems from "occasional anomalies during the generation process," rather than the triggering conditions themselves. For example, when a smart network interface card (NIC) generates a CNP, a momentary circuit interference may cause a checksum calculation error, or the target terminal may suddenly go offline (such as a substation equipment restart), causing the IP to become invalid—these anomalies are unpredictable before CNP generation and can only be detected through verification after generation. If one attempts to determine this before generation, an additional "pre-verification step" is required. However, the logical complexity of pre-verification is comparable to that of the generation process, which would increase hardware latency from 1μs to over 3μs, contradicting the original intention of "hardware acceleration."

[0040] Invalid CNP determination: Check if the checksum in the CNP is correct. If the checksum is incorrect, the CNP is determined to be invalid and discarded directly. At the same time, check if the destination IP address and port number in the CNP exist in the preset list of valid targets. If they do not exist, the CNP is also determined to be invalid and discarded. Furthermore, the smart network interface card can adopt the following existing models: MellanoxConnectX-6Dx: Features: Supports hardware-accelerated congestion control, enabling rapid generation and processing of Congestion Notification Messages (CNPs). Possesses powerful packet processing capabilities, suitable for congestion control applications in data center networks. Application Scenarios: Suitable for data center networks requiring high-performance congestion control and low-latency communication.

[0041] Intel Ethernet 800 series: Features: Supports hardware-accelerated congestion control, enabling rapid generation and processing of Congestion Passives (CNPs). Offers high throughput and low latency, suitable for large-scale power line data transmission. Applications: Suitable for data center networks requiring high throughput and low latency, especially power communication networks.

[0042] Broadcom BCM57500: Features: Supports hardware-accelerated congestion control, enabling rapid generation and processing of Congestion Passives (CNPs). Offers high throughput and low latency, making it suitable for congestion control applications in data center networks. Application Scenarios: Suitable for data center networks requiring high-performance congestion control and low-latency communication.

[0043] S5: The smart NIC dynamically adjusts the frequency of sending CNPs to the RP node based on the congestion level. The sending end (RP) mainly determines the network congestion situation based on CNP (Congestion Notification Message).

[0044] S51: Based on the initial network state, set the initial CNP transmission frequency for different congestion levels. Specifically, when there is no congestion, the CNP transmission frequency is set to 0 (no CNP is transmitted); when there is mild congestion, the initial CNP transmission frequency is set to 10 CNPs / second; when there is moderate congestion, the initial CNP transmission frequency is set to 30 CNPs / second; and when there is severe congestion, the initial CNP transmission frequency is set to 50 CNPs / second. S52: The smart network card monitors changes in network congestion in real time and dynamically adjusts the CNP transmission frequency. Furthermore, the smart network card obtains the congestion assessment value C output by the congestion level determination model in step S3 in real time. When the congestion assessment value C increases by more than 0.1 within three consecutive sampling periods, it indicates that the network congestion is worsening, and the current CNP sending frequency is increased by 20%. When the congestion assessment value C decreases by more than 0.1 within three consecutive sampling periods, it indicates that the network congestion is easing, and the current CNP sending frequency is reduced by 20%. Meanwhile, the smart NIC determines whether the data sender is responding effectively to CNP by measuring the proportion of packets with ECN tags. If the data sender's transmission rate does not decrease as expected within five consecutive CNP transmission cycles (e.g., under severe congestion, the expected transmission rate decrease is 30%, but the actual decrease is only 10%), the CNP transmission frequency will be increased by 30%. If the data sender's transmission rate has decreased to a reasonable range (e.g., under mild congestion, the transmission rate decreases by 10%-15%), the CNP transmission frequency will be maintained at the current level or appropriately reduced by 10%.

[0045] Furthermore, the CNP transmission frequency can be determined by combining the congestion assessment value C and the proportion of packets with ECN tags, as shown in the following formula: ; This is the final CNP transmission frequency; The initial CNP transmission frequency reference value; The proportion of packets with ECN tags; The target ECN labeling ratio corresponding to the congestion level; To round the calculation results to the nearest integer, ensuring that the frequency is an integer; To ensure the frequency is non-negative (when there is no congestion) The output is still 0).

[0046] Preset according to congestion level: ; ; ) S6: System Collaboration and Effect Verification S61: The data transmitter sends large-scale power data. Intermediate network nodes collect multi-dimensional congestion signals in real time and preprocess them. Based on the preprocessed signals, the data packets are marked with ECN using an optimized ECN marking strategy. When the data packets with ECN tags are transmitted to the data receiver or pass through the smart network card, the smart network card generates CNP through hardware acceleration and filters them. Then, it dynamically adjusts the CNP transmission frequency according to the network congestion status and the response of the data transmitter, and sends the effective CNP to the data transmitter. The data transmitter adjusts the power data transmission rate according to the CNP, so as to realize the coordinated work of the entire system and achieve the purpose of congestion signal enhancement. S62: By building an experimental environment simulating large-scale power data transmission, the performance indicators of the traditional RP-CP-NP mechanism and the RP-CP-NP congestion signal enhancement method proposed in this invention are compared in terms of congestion detection accuracy, CNP supply efficiency, and real-time performance and stability of power data transmission. Congestion detection accuracy: The correct judgment rate of the method of this invention for different congestion levels (mild, moderate and severe) under the two mechanisms was statistically analyzed. The correct judgment rate of the method of this invention should not be less than 95%, which is much higher than that of the traditional mechanism (usually less than 80%). CNP supply efficiency: In large-scale incast (convergence congestion) scenarios (such as 1000 data senders sending data to one data receiver simultaneously), the number of effective CNPs sent per unit time under the two mechanisms is statistically analyzed. The number of effective CNPs sent per unit time by the method of this invention should be more than 50% higher than that of the traditional mechanism, ensuring sufficient CNP supply. Real-time performance and stability of power data transmission: The transmission delay and packet loss rate of power data under two mechanisms were measured. The transmission delay of the method of the present invention should be reduced by more than 30%, and the packet loss rate should be controlled below 0.5%. The traditional mechanism has a higher transmission delay and a packet loss rate of more than 2%. This verifies that the method of the present invention can effectively ensure the real-time performance and stability of power data transmission.

[0047] Beneficial effects: I. Multi-dimensional signal fusion solves the problem of "misjudgment based on a single signal" and improves the accuracy of congestion assessment. Existing technologies generally rely on a single signal (such as queue length or RTT alone) to determine congestion, which is susceptible to problems such as "sudden traffic interference" and "signal lag" in power data transmission, leading to misjudgments (such as mistaking short-term queue fluctuations for congestion or ignoring the risk of sudden surges in link utilization). This solution achieves a breakthrough through multi-dimensional signal collaborative evaluation: The signal acquisition dimensions are more comprehensive: three core signals, namely queue length, RTT (round-trip time), and link utilization, are collected simultaneously, covering the three key congestion characteristics of "queue backlog", "transmission delay" and "bandwidth occupancy", avoiding the one-sidedness of a single signal; More precise signal processing: Through preprocessing operations such as moving average filtering, 3σ outlier removal, and median filtering, electromagnetic interference and signal noise caused by equipment transient failures in the power network are eliminated, ensuring the reliability of input data; The assessment model is more scientific: Based on the nonlinear signal enhancement function (segmented amplification of signal sensitivity in high-congestion intervals), signal change rate (reflecting the dynamic trend of congestion), and scenario correction coefficient (adapting to the difference between "burst / stable flow" of power), a congestion assessment value formula is constructed, realizing a three-dimensional assessment of "static signal + dynamic trend + scenario adaptation", and the accuracy of congestion level classification is improved to over 95% (traditional methods are usually below 80%).

[0048] II. Differentiated ECN labeling strategy to solve the "one-size-fits-all" labeling problem and balance congestion warning and transmission efficiency. Existing technologies mostly employ fixed ECN (Explicit Congestion Notification) marking rules (such as marking 50% of packets regardless of congestion level), leading to either over-marking during mild congestion (wasting bandwidth) or under-marking during severe congestion (delayed early warning), failing to meet the dual requirements of "low latency" and "high reliability" for power data. This solution optimizes the marking process through differentiated marking linked to congestion levels: No congestion (C<0.3): No ECN is marked to ensure the transmission efficiency of routine power data (such as monitoring logs) and avoid invalid marking occupying resources; Mild congestion (0.3≤C<0.6): Marked with a 20% probability, slightly indicating potential risks, suitable for the transmission needs of "non-critical data" (such as ambient temperature and humidity) in the power industry, reducing the impact on normal business; Moderate congestion (0.6≤C<0.8): Marked with a 50% probability to enhance the warning strength and adapt to the reliability requirements of "important data (such as equipment status monitoring)"; Severe congestion (C≥0.8): 100% probability marking, strong triggering of congestion control to ensure that "core data (such as fault trip signals)" are not lost. This strategy achieves a precise match between "congestion level and early warning strength", finding the optimal balance between early warning timeliness and transmission efficiency.

[0049] 3. Smart NIC hardware acceleration solves the "CNP generation delay" problem and improves congestion response speed. Existing technologies rely on software to generate CNPs (Congestion Notification Messages), which typically take more than 10μs to generate and require scheduling by the operating system kernel. In large-scale power data scenarios (such as simultaneous transmission from tens of thousands of terminals), insufficient CNP supply can easily occur, causing the sending end (RP) to be unable to adjust its rate in time, exacerbating congestion. This solution achieves a breakthrough by deploying smart network interface cards (NICs) on the NP nodes: Hardware-based CNP generation: Smart NICs extract ECN tag data packet information through dedicated hardware circuits and complete CNP generation within 1μs (more than 10 times faster), avoiding scheduling delays at the software level; Real-time filtering of invalid CNPs: The hardware automatically verifies the checksum of CNPs and the validity of the target IP, and removes erroneous or invalid packets (such as IP failures caused by power terminal offline) to avoid invalid CNPs consuming bandwidth; Adapted to high-concurrency scenarios: In Incast scenarios where 1000 power data transmitters transmit simultaneously, the effective CNP transmission volume per unit time is increased by more than 50% compared to traditional methods, ensuring that the transmitter (RP) can receive congestion notifications in real time and quickly adjust the rate.

[0050] IV. Dynamic CNP transmission frequency solves the problem of "fixed notification frequency" and adapts to dynamic changes in congestion. Existing technologies often employ a fixed CNP transmission frequency (e.g., sending 20 notifications per second regardless of congestion level), leading to either insufficient notifications when congestion worsens or excessive notifications when congestion eases. For example, a fixed frequency cannot quickly transmit risks during severe congestion, while high-frequency notifications waste resources during mild congestion. This solution optimizes this by adjusting the frequency in conjunction with congestion status. Initial frequency tier settings: No transmission when there is no congestion; initial frequencies of 10 / 30 / 50 packets / second are set for light / moderate / severe congestion respectively to adapt to the baseline requirements of different scenarios. Real-time dynamic adjustment: Combining changes in congestion assessment values ​​(e.g., if C increases by more than 0.1 for three consecutive cycles, the frequency is increased by 20%) and the sender's response (e.g., if the speed does not decrease as expected, the frequency is increased by 30%), the frequency is precisely and dynamically optimized through formulas. Avoid over-control: When the sending rate drops to a reasonable range (e.g., a 10%-15% reduction during mild congestion), the frequency is automatically reduced by 10% to avoid rate fluctuations caused by "over-notification" and ensure the stability of power data transmission.

[0051] V. Adapt to large-scale power data scenarios, solve the problem of "incompatibility of general algorithms", and ensure transmission reliability. Existing congestion control algorithms (such as machine learning-based methods) are mostly designed for general data center scenarios, failing to consider the "burst nature" (e.g., sudden increases in fault monitoring data), "variability" (different priorities for core / non-core data), and "high reliability requirements" (e.g., zero loss of control commands) of power data. In power scenarios, these algorithms are prone to problems such as "computational complexity leading to latency" and "inability to distinguish data priorities." This solution achieves deep adaptation through scenario-based design: Scenario correction coefficient adapts to sudden events: The congestion assessment value is corrected by the γ(t) coefficient. When a certain signal suddenly increases (such as a sudden increase in queue length and RTT lag), the assessment value is amplified to avoid missed judgments and adapt to sudden traffic during power failures. Differentiated strategy ensures priority: Combining ECN marking probability and CNP frequency, the transmission path of core data (such as tripping instructions) is marked first and CNP is sent first to ensure that critical services are not affected by congestion; Low packet loss and high real-time performance: Experimental data shows that the power data transmission latency is reduced by more than 30% under this solution, and the packet loss rate is controlled below 0.5% (the packet loss rate of traditional methods is usually over 2%), meeting the core requirements of "low latency and high reliability" for large-scale power data.

[0052] Example 2 like Figure 2 As shown, an RP-CP-NP congestion signal enhancement system for large-scale power data transmission, applied to the aforementioned RP-CP-NP congestion signal enhancement method for large-scale power data transmission, includes: The building module is used to construct the RP-CP-NP congestion signal enhancement system architecture and set up smart network interface cards on the NP node; The acquisition module is used to collect and preprocess multi-dimensional congestion signals at the CP node. The tagging module is used by CP nodes to obtain congestion assessment values ​​based on multi-dimensional congestion information and classify congestion levels, and to perform ECN differential tagging of data packets according to the congestion level. The generation module is used by the smart network interface card to generate and filter CNPs based on data ECN-tagged packets; The adjustment module is used by the smart network interface card to dynamically adjust the frequency of sending CNPs to the RP node according to the congestion level.

[0053] Example 3 A computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the RP-CP-NP congestion signal enhancement method for large-scale power data transmission described above.

[0054] Example 4 A processor for running a program, wherein the program executes the above-described RP-CP-NP congestion signal enhancement method for large-scale power data transmission.

[0055] This application discloses an RP-CP-NP congestion signal enhancement method, system, medium, and processor for large-scale power data transmission, relating to the field of congestion control technology, and solving the problems of poor adaptability, slow response, and inaccurate assessment in existing methods. The method steps are as follows: constructing an RP-CP-NP architecture with smart network interface cards (NICs); CP nodes collect and preprocess queue length, RTT, and link utilization signals; CP nodes calculate congestion assessment values, classify congestion levels, and differentiate ECN tags; smart NICs generate and filter CNPs; smart NICs dynamically adjust the CNP transmission frequency according to the congestion level. The system includes modules for construction, collection, tagging, generation, and adjustment. This invention improves the accuracy of congestion assessment, accelerates response speed, adapts to the characteristics of power data transmission, ensures real-time transmission and stability, and is suitable for large-scale power data transmission scenarios.

[0056] Those skilled in the art will recognize that the units of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the invention.

[0057] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical functional division. In actual implementation, there may be other division methods, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored.

[0058] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0059] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of this application.

Claims

1. A method for enhancing RP-CP-NP congestion signals for large-scale power data transmission, characterized in that, include: S1: Construct an RP-CP-NP congestion signal enhancement system architecture and set up smart network interface cards (NICs) on the NP nodes; S2: Collect multi-dimensional congestion signals at the CP node and perform preprocessing; S3: CP nodes obtain congestion assessment values ​​based on multi-dimensional congestion information and classify congestion levels, and perform ECN differential marking on data packets according to the congestion level; S4: The smart network interface card generates and filters CNPs based on the data ECN tag packets; S5: The smart NIC dynamically adjusts the frequency of sending CNPs to the RP node based on the congestion level.

2. The RP-CP-NP congestion signal enhancement method for large-scale power data transmission according to claim 1, characterized in that, In step S2, the process of collecting and preprocessing multi-dimensional congestion signals at the CP node includes the following steps: S21: Real-time collection of queue length signals, RTT signals, and link utilization signals related to network congestion at intermediate network nodes; S22: Perform preprocessing operations on the acquired signals at intermediate network nodes.

3. The RP-CP-NP congestion signal enhancement method for large-scale power data transmission according to claim 1, characterized in that, In step S3, the CP node obtains a congestion assessment value based on multi-dimensional congestion signals and classifies congestion levels. It then performs ECN differentiation marking on data packets according to the congestion level, including the following steps: S31: Use the preprocessed multi-dimensional congestion signal as input to calculate the congestion assessment value, and classify the congestion level according to the congestion assessment value; S32: Formulate differentiated ECN marking rules based on different congestion levels to optimize the ECN marking strategy.

4. The RP-CP-NP congestion signal enhancement method for large-scale power data transmission according to claim 3, characterized in that, In step S31, the congestion assessment value is calculated using the following formula: ； in, This is the normalized queue length signal; The normalized RTT signal; This is the normalized link utilization signal; It is a nonlinear signal enhancement function. ; These are the queue length signal, RTT signal, and link utilization signal change rate, respectively. This is a scene correction factor.

5. The RP-CP-NP congestion signal enhancement method for large-scale power data transmission according to claim 4, characterized in that, The nonlinear signal enhancement function The formula is as follows: ； in, This refers to the normalized queue length signal, RTT signal, or link utilization signal.

6. The RP-CP-NP congestion signal enhancement method for large-scale power data transmission according to claim 4, characterized in that, The formula for calculating the rate of change of the signal is as follows: ； in, The sampling period of the corresponding signal; ; For the rate of change of the signal, .

7. The RP-CP-NP congestion signal enhancement method for large-scale power data transmission according to claim 4, characterized in that, In step S5, the smart network interface card dynamically adjusts the frequency of sending CNPs to the RP node according to the congestion level, including the following steps: S51: Based on the initial network state, set the initial CNP transmission frequency for different congestion levels; S52: The smart network interface card monitors changes in network congestion in real time and dynamically adjusts the CNP transmission frequency. The formula for calculating the CNP transmission frequency is as follows: This is the final CNP transmission frequency; The initial CNP transmission frequency reference value; The proportion of packets with ECN tags; The target ECN labeling ratio corresponding to the congestion level; To round the calculation result to the nearest integer.

8. An RP-CP-NP congestion signal enhancement system for large-scale power data transmission, characterized in that, The RP-CP-NP congestion signal enhancement method for large-scale power data transmission as described in any one of claims 1 to 7 includes: The building module is used to construct the RP-CP-NP congestion signal enhancement system architecture and set up smart network interface cards on the NP node; The acquisition module is used to collect and preprocess multi-dimensional congestion signals at the CP node. The tagging module is used by CP nodes to obtain congestion assessment values ​​based on multi-dimensional congestion information and classify congestion levels, and to perform ECN differential tagging of data packets according to the congestion level. The generation module is used by the smart network interface card to generate and filter CNPs based on data ECN-tagged packets; The adjustment module is used by the smart network interface card to dynamically adjust the frequency of sending CNPs to the RP node according to the congestion level.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device containing the computer-readable storage medium to perform the RP-CP-NP congestion signal enhancement method for large-scale power data transmission as described in any one of claims 1 to 7.

10. A processor, characterized in that, The processor is used to run a program, wherein the program executes the RP-CP-NP congestion signal enhancement method for large-scale power data transmission as described in any one of claims 1 to 7.

Images