Method and system for evaluating time service precision of network time server

Abstract

The invention provides a time service precision evaluation method and system for a network time server, and relates to the technical field of computer networks, and the method comprises the steps: 1, collecting the operation data of the network time server in a dynamic network environment in real time, including core index data and auxiliary index data, carrying out the normalization of the core index data, and carrying out the calculation of the auxiliary index data; generating a standardized core index data set; step 2, dynamically integrating the standardized core index data set according to the fluctuation characteristics and relevance of each core index to generate a comprehensive core index characterization value; and step 3, in combination with auxiliary index data, carrying out dynamic compensation correction on the comprehensive core index representation value, and generating a comprehensive time service precision evaluation value. According to the invention, comprehensive, accurate and dynamic evaluation and abnormity diagnosis of the time service precision of the network time server are realized, and the network time service quality and the operation and maintenance efficiency are effectively improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of computer networks, in particular to a network time server time service precision evaluation method and system. BACKGROUND

[0002] The time service precision of a network time server (NTS) is a core element for ensuring the collaborative operation of enterprise distributed systems. Traditional evaluation methods rely on a single index (such as average time deviation) to calculate errors by statistically analyzing the round-trip delay of time synchronization protocol (such as NTP, PTP) messages, but do not integrate the dynamic geometric characteristics of network topology, resulting in evaluation failure in burst traffic scenarios (such as DDoS attacks and large-scale data backups).

[0003] Specifically, in a dynamic network environment, the existing technology cannot combine the physical location and topological connection relationship of key nodes (core switches and boundary routers) to quantize the spatial propagation effect of network congestion on time service precision.

[0004] For example, when a transmission path is congested due to an attack, the traditional method cannot represent the congestion propagation effect through changes in the distance between nodes, symmetry attenuation and other geometric characteristics, and it is also difficult to locate abnormal core nodes, resulting in evaluation results deviating from the actual scenario and fault repair lagging behind. SUMMARY

[0005] The technical problem to be solved by the application is to provide a network time server time service precision evaluation method and system to quantize the influence of congestion propagation on time service precision in real time.

[0006] To solve the above technical problems, the technical solution of the application is as follows: In a first aspect, a network time server time service precision evaluation method is provided, which comprises: Step 1: Real-time collection of operation data of a network time server in a dynamic network environment, including core index data and auxiliary index data, and normalization of the core index data to generate a standardized core index data set; Step 2: Dynamic integration of the standardized core index data set according to the fluctuation characteristics and correlation of each core index to generate a comprehensive core index representation value; Step 3: Dynamic compensation and correction of the comprehensive core index representation value in combination with the auxiliary index data to generate a comprehensive time service precision evaluation value; Step 4: Obtaining the coordinates of the vertices of the quadrilateral formed based on four target detection points in the auxiliary index data, and calculating the length ratio change rate of the quadrilateral, the diagonal angle offset and the symmetry attenuation coefficient according to the real-time changes of the quadrilateral vertex coordinates to generate a dynamic geometric correction factor; Step 5, the dynamic geometric correction factor is fused with the comprehensive timing accuracy evaluation value, the confidence interval of the evaluation value is adjusted through the coverage range contraction rate and the vertex offset direction of the quadrilateral, and the optimized comprehensive timing accuracy evaluation value and the dynamic confidence range are generated; Step 6, based on the optimized comprehensive timing accuracy evaluation value and the dynamic confidence range, in combination with the preset congestion state grading threshold, the timing accuracy stability level of the network time server in the current dynamic network environment, the abnormal event warning information and the network topology abnormal area positioning report identified based on the vertex offset direction of the quadrilateral are output.

[0007] In a second aspect, a network time server timing accuracy evaluation system comprises: A data acquisition module is configured to acquire core index data and auxiliary index data of a network time server in a dynamic network environment in real time. A data processing module is configured to normalize the core index data to generate a standardized core index data set, and dynamically integrate the core index data according to the fluctuation characteristics and correlation of each core index to generate a comprehensive core index representation value. A dynamic compensation correction module is configured to combine the network delay distribution characteristics and clock frequency offset in the auxiliary index data to dynamically compensate and correct the comprehensive core index representation value to generate a comprehensive timing accuracy evaluation value. A dynamic geometric correction module is configured to calculate the length proportion change rate of the quadrilateral vertex coordinates, the diagonal angle offset and the symmetry decay coefficient based on the position information of the four target detection points in the auxiliary index data to generate a dynamic geometric correction factor. An evaluation optimization module is configured to fuse the dynamic geometric correction factor with the comprehensive timing accuracy evaluation value, adjust the confidence interval of the evaluation value through the coverage range contraction rate and the vertex offset direction of the quadrilateral, and generate an optimized comprehensive timing accuracy evaluation value and a dynamic confidence range. A report generation module is configured to output the timing accuracy stability level, the abnormal event warning information and the network topology abnormal area positioning report identified based on the vertex offset direction of the quadrilateral based on the optimized comprehensive timing accuracy evaluation value and the dynamic confidence range in combination with the preset congestion state grading threshold.

[0008] In a third aspect, a computer readable storage medium stores a program, which is executed by a processor to implement the method.

[0009] The above-mentioned scheme of the present application has at least the following advantages: Real-time acquisition of core indicators and auxiliary indicators data, and normalization of core indicators data, to ensure that comprehensive and uniform standard data is obtained. Core indicators data reflect the key factors of time service accuracy, and auxiliary indicators data provide related information such as network environment. Through analysis of the fluctuation characteristics and correlation of core indicators, dynamic integration is performed to generate comprehensive core indicator representation values, which can mine the hidden relationship between core indicators, reflect the comprehensive state of related core elements of time service accuracy as a whole, and avoid the limitations of single indicator analysis. Dynamic compensation and correction of the comprehensive core indicator representation values are performed in combination with auxiliary indicator data, considering the influence of factors such as network environment on time service accuracy, so that the comprehensive time service accuracy evaluation value is closer to the actual situation.

[0010] A quadrilateral is constructed based on four target detection points in the auxiliary indicator data, and a dynamic geometric correction factor is calculated by calculating related geometric parameters, introducing a new dimension for time service accuracy evaluation from a geometric perspective, which can capture the influence of network topology changes on time service accuracy and enhance the comprehensiveness of the evaluation method. The dynamic geometric correction factor and the comprehensive time service accuracy evaluation value are fused to adjust the confidence interval of the evaluation value, generate an optimized evaluation value and dynamic confidence range, and further optimize the evaluation result, so that the evaluation not only has numerical results, but also can give a confidence range, providing more information for decision-making. Based on the optimized result, in combination with a preset threshold, a time service accuracy stability level, abnormal event warning information and network topology abnormal area positioning report are output, the evaluation result is converted into specific and operable information, helping operation and maintenance personnel quickly understand the running state of the network time server, timely discover and handle abnormalities, ensure time service accuracy and improve network service quality. BRIEF DESCRIPTION OF DRAWINGS

[0011] Figure 1 is a flowchart of a network time server time service accuracy evaluation method provided by an embodiment of the present application.

[0012] Figure 2 is a schematic diagram of a network time server time service accuracy evaluation system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0013] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0014] As Figure 1 shown, an embodiment of the present application proposes a network time server time service accuracy evaluation method, which comprises the following steps: Step 1: collect the operation data of the network time server in a dynamic network environment in real time, including core indicator data and auxiliary indicator data, and normalize the core indicator data to generate a standardized core indicator data set; Step 2: Dynamically integrate the standardized core indicator data set based on the fluctuation characteristics and correlation of each core indicator to generate a comprehensive core indicator representation value; Step 3: Combined with the auxiliary indicator data, dynamically compensate and correct the comprehensive core indicator representation value to generate a comprehensive timing accuracy evaluation value; Step 4: Based on the four target detection points in the auxiliary indicator data, the vertex coordinates of the formed quadrilateral area are obtained. According to the real-time changes in the vertex coordinates of the quadrilateral, the change rate of the quadrilateral's side length ratio, the diagonal angle offset, and the symmetry attenuation coefficient are calculated to generate a dynamic geometric correction factor. Step 5: The dynamic geometry correction factor is integrated with the comprehensive timing accuracy evaluation value. The confidence interval of the evaluation value is adjusted by the coverage shrinkage rate of the quadrilateral and the vertex offset direction to generate the optimized comprehensive timing accuracy evaluation value and dynamic confidence range. Step 6: Based on the optimized comprehensive timing accuracy evaluation value and dynamic confidence range, combined with the preset congestion status classification threshold, output the timing accuracy stability level of the network time server in the current dynamic network environment, abnormal event warning information and network topology abnormal area positioning report based on quadrilateral vertex offset direction identification.

[0015] In an embodiment of the present invention, by real-time collection of core indicators (time deviation, jitter, synchronization cycle stability) and auxiliary indicators (network delay distribution, clock frequency offset), and combining with a dynamic compensation correction mechanism, the impact of short-term fluctuations in network delay and long-term drift of clock frequency on timing accuracy can be effectively distinguished.

[0016] A quadrilateral dynamic geometric model constructed based on four target detection points quantifies the spatial propagation effect of network congestion on the topological path through the rate of change of side length ratio, diagonal angle offset and symmetry attenuation coefficient, and realizes the real-time correlation analysis of timing accuracy anomalies and network topological geometric characteristics for the first time.

[0017] By fusing the dynamic geometric correction factor and the timing accuracy evaluation value, combining the coverage range shrinkage rate of the quadrilateral and the vertex offset direction, dynamically adjusting the confidence interval of the evaluation value, the false alarm and the missing alarm problems of the traditional static threshold method in the burst traffic scene are solved, and the abnormal detection sensitivity is improved. Based on the coupling relationship between the vertex offset direction of the quadrilateral and the network topology path, the core switch node or the boundary router node causing the timing accuracy degradation is located, a visual report containing an abnormal propagation path is generated, and the fault repair efficiency is improved by more than 50%. Through the sliding window smoothing algorithm, the dynamic baseline updating and the multi-time scale feature fusion technology, the network traffic mutation (such as DDoS attack, large-scale data backup) is responded in real time, and the timing accuracy evaluation is dynamically corrected within 5 seconds, so that the high real-time operation and maintenance demand is met.

[0018] In a preferred embodiment of the present application, the above step 1, the running data of the network time server in the dynamic network environment is collected in real time, including core index data and auxiliary index data, and the core index data is normalized to generate a standardized core index data set; the core index data includes time deviation sequence, time jitter parameter, synchronization period stability coefficient and maximum delay error value; the auxiliary index data includes network delay distribution characteristics, clock frequency offset and position information of four preset target detection points; the four target detection points are respectively located at the core switch node, the boundary router node, the key business server node and the node where the time server is located in the network topology, and can include: In the embodiment of the present application, the analysis process of the core index data is as follows: Time deviation sequence: in actual operation, a high-precision time measurement device such as a high-precision time synchronizer is used to establish a reliable connection with a high-precision atomic clock. During the operation of the network time server, the time of the network time server is obtained through a stable network communication protocol (such as NTP, PTP) at a set accurate time interval (such as every millisecond). Due to the influence of network transmission delay and device processing time, the transmission delay is calculated by measuring the round-trip time multiple times each time the time is obtained, and the delay is deducted when calculating the difference from the standard time. The time difference obtained each time is arranged in time sequence to form a time deviation sequence. During data recording, if abnormal data appears, for example, the difference suddenly jumps greatly and exceeds the normal fluctuation range, the adjacent data can be modified by interpolation method for reference, so as to ensure the accuracy of the sequence.

[0019] Time jitter parameter: Based on the obtained time deviation sequence, first, the sequence is cleaned up to remove invalid data points caused by network failures, equipment abnormalities, etc. such as continuous multiple same abnormal deviation values. In order to more accurately reflect the time jitter characteristics, the data is reasonably segmented according to the fluctuation period and trend of the time deviation sequence. For example, if the sequence has obvious periodic fluctuations, it can be segmented by period. In each segment of data, the variance or standard deviation is calculated. When calculating the variance, first, the square of the difference between each data point and the average value of the segment is calculated, and then the average value of these square values is calculated; the standard deviation is the square root of the variance. By comparing the calculation results with the different level thresholds set in advance, the severity level of time jitter is divided, such as slight jitter, moderate jitter and severe jitter, etc. to evaluate the instability of time deviation, and the smaller the value, the more stable the time.

[0020] Synchronization period stability coefficient: The definition of the synchronization period of the network time server is clear, which can be determined according to the synchronization protocol it uses (such as the default synchronization period or custom period of NTP). In each synchronization period, the change of time deviation is continuously monitored. In addition to calculating the difference of time deviation between adjacent synchronization periods, the change rate and fluctuation range of time deviation in the window are calculated using a sliding window method. Statistical analysis is performed on these differences and indicators, and in addition to calculating the mean and standard deviation, the median, coefficient of variation and other statistical quantities are also calculated. For example, the coefficient of variation can reflect the relative dispersion of the data, which helps to describe the stability of the time deviation in the synchronization period from different angles. A line chart of the stability coefficient changing with time is drawn, and the trend analysis algorithm (such as moving average method, exponential smoothing method, etc.) is used to predict the stability of the future synchronization period, and potential stability problems are discovered in time.

[0021] Maximum delay error value: In the process of time synchronization between the network time server and other nodes, high-precision timestamp recording equipment (such as hardware-based high-precision timestamp chip) is used to accurately record the time when sending time requests and receiving time responses. Due to various uncertain factors in network transmission, such as network congestion, route switching, etc., each synchronization not only records the single delay time, but also records related network state information, such as current network bandwidth utilization, packet loss rate, etc. The quality of the measurement data is evaluated, and when the delay time appears an abnormally high value, it is analyzed whether it is caused by network transient fault (such as temporary link interruption) or device performance bottleneck. By comparing the delay time distribution of multiple adjacent synchronization periods, it is determined whether the abnormally high value is within the normal fluctuation range. The moving average of the maximum delay error value in a period of time is calculated, and when the value exceeds a certain threshold, a network delay warning signal is sent to take timely measures to optimize network performance.

[0022] Analysis process of auxiliary index data: Network delay distribution characteristics: Select the appropriate probe packet type based on the network type (e.g., Ethernet, wireless) and measurement requirements (e.g., link delay measurement, end-to-end delay measurement, etc.). For example, when measuring link delay, use smaller ICMP probe packets; when measuring end-to-end delay, use TCP or UDP probe packets that resemble actual service data. The frequency and number of probe packets should be appropriately set. Using stratified sampling, probe packets should be sent during different time periods (e.g., peak and off-peak hours) and along different network paths. Each time a probe packet is sent, not only the round-trip delay time but also the routing information traversed by the probe packet is recorded. Statistical analysis should be performed on large amounts of delay data, calculating statistics such as the mean, median, standard deviation, and percentiles. Probability density function and cumulative distribution function graphs should be plotted for the delay data to visually demonstrate the distribution of network delay. Delay data should be categorized by different dimensions (e.g., source and destination node locations, network protocol type, etc.), and delay statistics should be calculated for each category to identify patterns and differences in delay distribution. This data should be compared with historical delay data to analyze changing trends in network delay distribution characteristics and promptly identify abnormal changes in network performance.

[0023] Clock frequency offset: Using high-precision frequency measurement equipment, such as an atomic frequency standard or a high-stability crystal oscillator, as a reference source, measure the network time server's clock frequency under stable environmental conditions (including constant temperature, humidity, and electromagnetic environment). Direct counting can be used, where a counter counts the cycles of the clock signal and calculates the frequency based on the count results. Alternatively, a phase comparison method can be used, where the network time server's clock signal is compared with the standard clock signal and the frequency offset is calculated based on the phase difference. When calculating frequency offset, consider the accuracy and stability of the measurement equipment, perform error analysis, and perform corrections. Record measurement results at regular intervals (e.g., once per minute), and plot the frequency offset over time. When the frequency offset exceeds a preset threshold, promptly implement frequency calibration measures, such as adjusting the clock's division factor through software or using hardware calibration equipment, to ensure the accuracy of the network time server's clock frequency.

[0024] Position information of the four target detection points: Obtain detailed network topology documents, including the model of network equipment, connection relationship, IP address and other information. For core switch nodes, border router nodes and key business server nodes, use network management software (such as SNMP management tools) to obtain the geographical location information of the equipment (if the equipment is configured with relevant information). If the equipment has no direct geographical location information, its approximate location can be inferred through the logical structure of the network topology and network traffic analysis. For the node where the time server is located, the accurate coordinate information is directly obtained from the physical installation location of the device. After obtaining the location coordinates, compare them with geographic information system (GIS) data, or use GPS positioning equipment to measure some nodes in the field to calibrate the coordinates. With the changes of network topology and the migration of equipment, a dynamic management mechanism is established to update the location information of the four target detection points in a timely manner, ensuring that these information can be accurately used in network analysis and time accuracy evaluation.

[0025] The normalization of core index data is to convert data of different magnitudes and units to the same range. For each core index, find the maximum and minimum values in the index data, and then map each data point to a specified interval such as [0, 1] or [-1, 1]. For example, for a index value , its normalized value can be calculated by the formula , where and are the minimum and maximum values in the index data, respectively. In this way, the time deviation sequence, time jitter parameter, synchronization period stability coefficient and maximum delay error value and other core index data are converted into unified dimensionless data to form a standardized core index data set.

[0026] In a preferred embodiment of the present application, the above step 2, according to the fluctuation characteristics and correlation of each core index, dynamically integrates the standardized core index data set to generate a comprehensive core index representation value, which can include: Step 200, sliding window segmentation is performed on the time deviation sequence, and the standard deviation of the short-term fluctuation amplitude and the positive or negative slope distribution of the trend direction in each window are analyzed to generate time deviation dynamic fluctuation characteristics; Step 201, based on the time deviation dynamic fluctuation characteristics, the peak value distribution characteristics of the time jitter parameter are extracted, the jitter events are filtered through a pre-set threshold, the jitter event frequency and intensity cumulative value are counted, and a time jitter abnormal event quantization parameter set is generated; Step 202, according to the time jitter abnormal event quantization parameter set, the synchronization period stability coefficient is converted from time domain to frequency domain, the main oscillation frequency and its harmonic component energy proportion are calculated, and a synchronization period stability frequency spectrum quantization index is generated; Step 203: combining the synchronization cycle stability spectrum quantitative index and the historical baseline data of the maximum delay error value, calculating the probability of triggering an extreme event exceeding the historical threshold, and correlating it with the real-time network load status to generate a maximum delay risk level parameter; Step 204: Input the dynamic fluctuation characteristics of time deviation, the quantitative parameter set of time jitter abnormal events, the quantitative index of synchronization cycle stability spectrum, and the maximum delay risk level parameter into the dynamic time window, calculate the time-varying correlation coefficient matrix between the parameters, and construct a multi-dimensional dynamic correlation model; Step 205 , performing feature space projection on the standardized core indicator dataset based on the multi-dimensional dynamic association model to extract dynamic feature vectors of different time granularities; Step 206 : Perform multi-time granularity superposition and normalization processing on the dynamic feature vector to generate a comprehensive core indicator representation value reflecting the coupling effect of the dynamic network environment.

[0027] In an embodiment of the present invention, first, the size and step size of the sliding window are determined. The window size is like the size of a "small window" for observing time deviations, for example, it is set to 100 consecutive time point data; the step size determines the distance the window slides each time, assuming that the step size is set to 10 time point data. Starting from the starting position of the time deviation sequence, the window is overlaid on the sequence, and the 100 time deviation data within the window are processed. The standard deviation of these 100 data is calculated. The standard deviation is calculated by first finding the average value of this set of data, then calculating the square of the difference between each data and the average value, then finding the average of these squared values, and finally taking the square root of this average value. The standard deviation can be used to measure the degree of dispersion of the time deviation data within the window, that is, the magnitude of the fluctuation. The larger the standard deviation, the more severe the fluctuation.

[0028] A linear fit is performed on the 100 time deviation data points within the window, essentially finding a straight line that best represents the data's changing trend. If the slope of this line is positive, it indicates an upward trend in time deviation during this window; a negative slope indicates a downward trend; and a slope close to zero indicates relatively stable time deviation. After completing the analysis for a window, the window is moved to the next position according to the set step size (10 time points), and the standard deviation calculation and slope analysis process is repeated until the window has traversed the entire time deviation sequence. The standard deviation calculated for each window and the distribution of positive and negative slopes in the trend direction are recorded. This information constitutes the dynamic fluctuation characteristics of time deviation.

[0029] Step 201, based on the time deviation dynamic fluctuation characteristics obtained in step 200, the time jitter parameter is analyzed point by point. In this process, an observation window is set, the size of this window depends on the stability of the network environment and the accuracy requirement of analysis. For example, in a stable enterprise network, 5-10 minutes is selected as an observation window; while in the fluctuation larger wide area network environment, it may need to be reduced to 1-2 minutes.

[0030] In each observation window, the peak value is determined by using sliding comparison. The specific method is to compare the jitter parameter value of the current point with a certain number of adjacent points (such as 5 points before and after) before and after it. If the value of the current point > all adjacent points, and exceeds a certain proportion (such as 1.5 times) of the average value in the window, it is confirmed as a local peak value. Record the specific time stamp of the occurrence of these peak values to the millisecond level, and record the corresponding peak value size, retain enough decimal places (such as 6 decimal places) to ensure accuracy. After completing the analysis of all observation windows, the obtained peak values are arranged in time sequence to form a peak value sequence.

[0031] The determination of the preset threshold is as follows: a variety of factors are considered, including network type, business demand and historical data performance. Data statistical method is adopted: that is, the time jitter parameter data in the past period (such as a week or a month) is analyzed, and the average value and standard deviation are calculated. The threshold can be set as the average value plus 3 times the standard deviation.

[0032] After determining the preset threshold and the peak value distribution characteristics, the jitter event is screened. For each peak point, its value is compared with the preset threshold, if it exceeds the threshold, it is marked as a potential jitter event. In order to avoid misjudgment, these potential jitter events need to be confirmed again. The specific method is to check the jitter parameter change in a period of time (such as 30 seconds) before and after the peak point. If in this period of time, the jitter parameter presents obvious rising and falling trend, and the duration exceeding the threshold reaches a certain length (such as 5 seconds), it is confirmed as a valid jitter event. For a plurality of peak points appearing continuously, if they all exceed the threshold and the time interval is short (such as less than 10 seconds), they are combined into a jitter event to avoid repeated counting of the same fluctuation process.

[0033] Jitter event frequency calculation: after confirming all jitter events, the total number of jitter events occurring in a certain period of time (such as 1 hour, 1 day) is counted. Divide this total number by the length of the time period to get the frequency of jitter event occurrence, which can be expressed in times / hour or times / day.

[0034] Jitter event intensity cumulative value calculation: for each confirmed jitter event, calculate its intensity. The method of intensity calculation is to integrate the part of the peak value in the event that exceeds the threshold, that is, to calculate the difference between the peak value and the threshold, and multiply it by the duration of the difference. Add up the intensity values of all jitter events to get the intensity cumulative value.

[0035] Organize the obtained jitter event occurrence frequency, intensity cumulative value, and peak value distribution characteristics, etc. information to form a complete set of time jitter abnormal event quantization parameters. This parameter set can be organized in the form of a table, and the specific content includes: Basic statistical information: including total observation time, total number of jitter events, average frequency, etc.

[0036] Peak value distribution characteristics: list the timestamp, amplitude value of all peak values, and statistical characteristics of peak values (such as maximum value, minimum value, average value, median, etc.).

[0037] Jitter event list: detailed record of the start time, end time, duration, peak size, amplitude exceeding the threshold, and calculated intensity value of each jitter event.

[0038] Summary index: including intensity cumulative value, frequency change trend, etc. comprehensive index.

[0039] Step 202, after obtaining the time jitter abnormal event quantization parameter set, convert the data of the synchronization period stability coefficient from the time domain to the frequency domain. The time domain describes the change of the synchronization period stability coefficient with time, while the frequency domain conversion can analyze the data from the frequency point of view, that is, to decompose a piece of sound into different frequency notes, and decompose the synchronization period stability coefficient into different frequency components. In the converted frequency domain data, find the frequency with the most concentrated energy, which is the main oscillation frequency, which represents the most important frequency component in the change of the synchronization period stability coefficient. Then, calculate the energy proportion of each harmonic component (that is, other frequency components except the main oscillation frequency). Record the main oscillation frequency and the energy proportion of each harmonic component, etc. information to generate the synchronization period stability spectrum quantization index.

[0040] Step 203, refer to the synchronization cycle stability spectrum quantization index obtained in step 202, and then retrieve the historical baseline data of the maximum delay error value. The historical baseline data is the range and statistical information of the maximum delay error value collected and recorded over a long period of time under the condition of normal and stable operation of the network. Compare the current maximum delay error value data with the threshold value in the historical baseline data, and count the number of times the maximum delay error value exceeds the historical threshold value within a certain time period. Divide the number of times by the total number of times to obtain the extreme event trigger probability of exceeding the historical threshold value. At the same time, obtain the real-time network load state information, such as the usage of network bandwidth, the size of data flow in the network, etc. According to the extreme event trigger probability and the real-time network load state, divide the maximum delay risk into different levels according to the pre-set risk level division rule, such as low risk, medium risk, high risk, etc., so as to generate the maximum delay risk level parameter.

[0041] Step 204, set a dynamic time window, the size of which can be adjusted according to actual needs, which is like an "observation frame" that moves with time. Put the time deviation dynamic fluctuation characteristics, time jitter abnormal event quantization parameter set, synchronization cycle stability spectrum quantization index and maximum delay risk level parameter obtained in steps 200-203 into this dynamic time window. In the window, calculate the time-varying correlation coefficient between any two parameters. The correlation coefficient is used to measure the degree of association between two parameters, and the value range is between -1 and 1. If the correlation coefficient is close to 1, the change trend of the two parameters is highly consistent; close to -1, the change trend of the two parameters is opposite; close to 0, there is almost no correlation between the two parameters. Calculate the correlation coefficient between each two parameters at different time points, and arrange these correlation coefficients into a matrix, which shows the mutual association of various parameters at different time points. Based on this time-varying correlation coefficient matrix, a multi-dimensional dynamic correlation model is constructed, which presents the complex dynamic relationship between various core indicators.

[0042] Step 205, use the multi-dimensional dynamic correlation model constructed in step 204 to regard the standardized core indicator data set as a whole. Project the data set in the feature space through the model, as if the data set is "mapped" from one space to another. In the new space, the features of the data will be presented in different forms. According to different time granularity requirements, such as short time (such as a few minutes), medium time (such as a few hours), and long time (such as a few days), extract the corresponding dynamic features from the projected results. These dynamic features are the key information that the data shows at different time scales. Arrange the dynamic features extracted at each time granularity into a vector form, and you get the dynamic feature vectors at different time granularities.

[0043] Step 206, superimpose the dynamic feature vectors of different time granularities obtained in step 205. This is like layering information of different time periods together and fusing them with each other to form a comprehensive vector containing information of multiple time granularities. Since the numerical size and range of the dynamic feature vectors of different time granularities may be different, normalization needs to be performed on the comprehensive vector. Normalization is to uniformly adjust the values in the vector to a fixed range, such as 0 to 1, according to certain rules. Through normalization, the dimensional differences between the dynamic feature vectors of different time granularities are eliminated, making them comparable. After the multi-time-granularity superposition and normalization, the final result is a comprehensive core index representation value reflecting the coupling effect of the dynamic network environment, which comprehensively considers the dynamic changes of each core index under different time granularities and the interaction relationship between them.

[0044] Carefully capturing the fluctuation and trend of the time deviation in a short time helps to have a clearer understanding of the dynamic characteristics of the time deviation, and helps to timely discover abnormal changes of the time deviation. Abstract time jitter is converted into specific quantifiable parameters, and the jitter abnormal event is accurately located, which facilitates the analysis of the influence of time jitter on the time accuracy and provides strong data support for evaluating and optimizing the time accuracy. From the frequency domain perspective, the stability of the synchronization period is analyzed to find potential periodic interference factors, which helps to more comprehensively and deeply evaluate the influence of the synchronization period on the time accuracy and helps to find the key frequency components affecting the stability of the synchronization period.

[0045] By combining historical data and real-time network status, the risk of maximum delay error can be predicted in advance, so that operation and maintenance personnel can take timely measures to deal with possible serious delay problems and ensure the stability of the time accuracy. The multi-dimensional dynamic correlation model constructed shows the mutual relationship and dynamic changes between the core indexes, breaks the limitation of isolated analysis of indexes, helps to understand the comprehensive influence mechanism of the indexes on the time accuracy, and provides a theoretical basis for accurate evaluation. The dynamic feature vectors of different time granularities are extracted to meet the demand for time scale in different analysis scenarios, whether focusing on short-term changes or long-term trends, the corresponding key information can be obtained, and the flexibility of evaluation is improved.

[0046] In a preferred embodiment of the present application, the step 3 above, in combination with auxiliary index data, performs dynamic compensation and correction on the comprehensive core index representation value to generate a comprehensive time accuracy evaluation value, which can include: Step 300, according to the burst fluctuation intensity and duration of the network delay distribution characteristics, set a dynamic abnormal jump threshold, identify the abnormal jump interval of the time deviation sequence that is greater than or equal to the dynamic abnormal jump threshold; Step 301, using a sliding window smoothing algorithm to segment and compensate the abnormal jump interval, and calculating the time deviation correction sequence after compensation. Step 302: Based on the compensated time deviation correction sequence, reversely calibrate the accumulated drift error of the synchronization period stability coefficient to generate a drift error compensation coefficient; Step 303: Superimpose the time offset correction sequence and the drift error compensation coefficient on the comprehensive core indicator representation value to generate a timing accuracy evaluation value after preliminary compensation. Based on the real-time fluctuation amplitude of the timing accuracy evaluation value, calculate the real-time change rate of the network delay distribution characteristics. Step 304 : dynamically adjust the window length and compensation strength of the sliding window smoothing algorithm according to the real-time change rate of the network delay distribution characteristics, and generate a comprehensive timing accuracy evaluation value including dynamic network delay correction and clock frequency offset suppression.

[0047] In this embodiment of the present invention, when analyzing network delay distribution characteristics, the first step is to determine the time range for data statistics. The selection of this range is critical. If the time range is too short, it may not capture the complete network fluctuation pattern; if the time range is too long, it may include too much irrelevant interference information. Taking the past 10 minutes as an example, during these 10 minutes, the network monitoring equipment continuously recorded each network delay data point. This data is like a string of beads, with each bead representing the network delay value at a specific point in time. When calculating basic statistics such as the mean and variance of the network delay data during this period, the mean is like the average weight of the beads, reflecting the general level of network delay. The variance reflects the degree of difference in the weight of the beads; the larger the variance, the more severe the fluctuation of network delay. The process of calculating the mean and variance is like weighing the beads on a scale and then analyzing the distribution of their weight.

[0048] Observing fluctuations in network latency data to determine if there are sudden fluctuations requires the same sensitivity as observing weather changes. For example, in an otherwise calm network environment, latency suddenly rises from an average of 50ms to 200ms, and this high latency persists for 10 seconds. This is like a clear sky suddenly covered with dark clouds and heavy rain; this noticeable change is a sudden fluctuation.

[0049] When setting the dynamic abnormal jump threshold based on the intensity and duration of sudden fluctuations, historical data should be used as a reference. Assume that long-term observations have shown that under normal circumstances, network latency increases generally do not exceed 30ms and last no longer than 2 seconds. When a sudden fluctuation occurs with an intensity exceeding 50ms and a duration exceeding 3 seconds, to accurately identify the anomaly while avoiding oversensitivity, we set the threshold slightly lower than the intensity of the sudden fluctuation, such as 45ms. This threshold acts as a warning line; when the data changes in the time deviation series cross this line, an anomaly is considered.

[0050] To identify abnormal jump intervals in a time-deviation series, we need to examine each data point in turn, much like searching for a unique symbol in a password. When the difference (i.e., the change) between a data point and the previous one is greater than or equal to the set dynamic abnormal jump threshold, we mark that point as an abnormal jump interval. This is equivalent to finding the first unique symbol in the password. We then continue checking until we find a point where the change is less than the threshold. The interval between these two points is the abnormal jump interval. If there are multiple such intervals in the time-deviation series, we need to identify them one by one, like a detective searching for clues.

[0051] In step 301, after determining the abnormal transition interval, selecting an appropriate sliding window size is crucial. The selection of the sliding window size requires a comprehensive consideration of the length of the abnormal transition interval and data fluctuations. For example, a window size encompassing data from 10 time points is chosen. This window acts like a movable magnifying glass, allowing for detailed observation of the data within the abnormal transition interval. The sliding window is positioned at the beginning of the abnormal transition interval and the data within the window is processed. The average value of the data within the window is calculated. This process is like mixing the 10 data points within the window in a large bowl to obtain a value representing the average of these 10 data points. This average value is then used to replace the original value of each data point within the window. This is equivalent to replacing beads of varying shapes with beads of the same size, thus achieving data smoothing. The window is then moved backward in steps of a specified size (e.g., one time point at a time). The above average calculation and data replacement process is repeated with each slide. This process is like slowly moving a magnifying glass across the abnormal transition interval, gradually smoothing the data. Until the sliding window covers the entire abnormal jump interval, the originally fluctuating data becomes relatively smooth.

[0052] For more precise processing, longer abnormal jump intervals can be divided into several smaller segments. For example, if an abnormal jump interval contains data from 50 time points, we can imagine it as a long rope and divide it into five equal segments, each with 10 time points. Then, a sliding window smoothing operation is performed on each segment in turn, just like straightening a long rope section by section, ultimately obtaining a compensated time deviation correction sequence.

[0053] In step 302, when analyzing the synchronization cycle stability coefficient based on the compensated time deviation correction sequence, we must treat each synchronization cycle as an independent, small world. Within each small world, we observe the changes in the time deviation correction sequence and calculate the cumulative change in the time deviation within each synchronization cycle. This is like recording the total weight change of a bead within each small world. These cumulative changes are compared with the synchronization cycle stability coefficient to determine whether the synchronization cycle stability coefficient has drifted. If the synchronization cycle stability coefficient continuously increases or decreases over multiple cycles, inconsistent with the changing trend of the time deviation correction sequence, it indicates that it is drifting, like a train that has deviated from its track. When calculating the cumulative drift error, we must find the sum of the differences between the actual and theoretical values ​​of the synchronization cycle stability coefficient. The theoretical value here is a reasonable value estimated based on the time deviation correction sequence, much like the normal weight estimated based on the variation pattern of the bead weight. By calculating the sum of the differences between the actual and theoretical values, we can determine the extent of the deviation of the synchronization cycle stability coefficient.

[0054] A drift error compensation coefficient is then generated based on the magnitude and direction of the accumulated drift error. If the accumulated drift error is positive, it indicates that the synchronization cycle stability coefficient is too large, just like a train running too fast and deviating from the track. In this case, the drift error compensation coefficient is a negative value, which is used to reduce the synchronization cycle stability coefficient and return it to normal track; vice versa.

[0055] Step 303 superimposes the compensated time deviation correction sequence and drift error compensation coefficient onto the comprehensive core indicator representation value. This is a data integration process. The time deviation-related portion of the comprehensive core indicator representation value is replaced with the time deviation correction sequence, much like replacing an old part with a new one. The portion related to the synchronization period stability coefficient is adjusted based on the drift error compensation coefficient, much like adjusting parameters in a machine. This generates a preliminary compensated timing accuracy evaluation value. The fluctuation amplitude of the preliminary compensated timing accuracy evaluation value over a short period of time (e.g., the past minute) can be calculated by, for example, calculating the difference between the maximum and minimum values. This is like measuring the undulation of ocean waves. The difference between the maximum and minimum values ​​is like the height difference between the highest and lowest points of the wave, while the variance more comprehensively reflects the severity of the wave fluctuations. Based on this fluctuation amplitude, combined with historical and current data on the network delay distribution characteristics, the real-time rate of change of the network delay distribution characteristics is calculated. If the fluctuation range of the evaluation value increases and the network delay data also shows that the delay value is increasing, just like the waves are becoming more and more turbulent, the real-time change rate is positive, indicating that the network delay distribution characteristics are deteriorating; conversely, if the fluctuation range of the evaluation value decreases and the real-time change rate is negative, it means that the network delay distribution characteristics are improving, just like the waves are gradually calming down.

[0056] Step 304 dynamically adjusts the window length and compensation strength of the sliding window smoothing algorithm based on the calculated real-time rate of change of the network delay distribution. If the real-time rate of change is large, it indicates that network delay is fluctuating dramatically, like waves in a storm. In this case, the sliding window length is increased (for example, from 10 time points to 20 time points) to better smooth the data, like using a larger net to catch the waves. At the same time, the compensation strength is increased, meaning the original data is replaced with the average value at a larger scale, to quickly eliminate the impact of abnormal fluctuations and restore data to normalcy as soon as possible.

[0057] On the contrary, if the real-time change rate is small, it means that the network delay is relatively stable, like the surface of a calm lake. In this case, reducing the length of the sliding window (for example, from 10 time points to 5 time points) increases sensitivity to data changes, just like using a finer net to catch small fish. Lowering the compensation intensity avoids excessive smoothing that may cause loss of real data change information and prevent the filtering out of small fish.

[0058] After dynamically adjusting the sliding window smoothing algorithm, the time offset correction sequence is processed again, combined with analysis of the clock frequency offset. The clock frequency offset is like the speed deviation of a small gear. Based on its magnitude, the timing accuracy evaluation value is fine-tuned to compensate for the error introduced by the clock frequency offset. Ultimately, a comprehensive timing accuracy evaluation value is generated that incorporates dynamic network delay correction and clock frequency offset suppression, resulting in a more accurate reflection of the timing accuracy of the network time server.

[0059] By setting a dynamic abnormal jump threshold based on the network delay distribution characteristics, it can flexibly adapt to fluctuations in different network environments and accurately identify abnormal jump intervals in the time deviation series, providing precise targets for subsequent corrections and avoiding misclassifying normal fluctuations as abnormalities, thereby improving the accuracy of the assessment. A sliding window smoothing algorithm is used to compensate for abnormal jump intervals in sections, effectively eliminating the impact of abnormal jumps on the time deviation series, making the time deviation data smoother and more stable, and reducing data noise interference, laying the foundation for more accurate subsequent timing accuracy assessments. Reverse calibration of the accumulated drift error of the synchronization period stability coefficient corrects for deviations caused by various factors, ensuring that it truly reflects the synchronization period stability of the network time server and improving the reliability of this indicator in timing accuracy assessments. The corrected time deviation series and compensation coefficient are superimposed on the comprehensive core indicator representation to generate a preliminary compensated timing accuracy evaluation value. The real-time change rate of the network delay distribution characteristics is then calculated, enabling preliminary correction of timing accuracy and real-time perception of network delay variations, providing a basis for further optimization and assessment. The sliding window smoothing algorithm is dynamically adjusted according to the real-time change rate of the network delay distribution characteristics, so that the algorithm can adapt to changes in the network environment and effectively correct time deviations under different network conditions, while suppressing the influence of clock frequency offsets, and ultimately generating a more accurate and reliable comprehensive timing accuracy evaluation value, thereby improving the adaptability and accuracy of the timing accuracy evaluation method in a dynamic network environment.

[0060] In a preferred embodiment of the present invention, the above step 4, based on the four target detection points in the auxiliary indicator data, obtains the coordinates of the vertices of the quadrilateral area formed, and calculates the quadrilateral side length ratio change rate, diagonal angle offset and symmetry attenuation coefficient according to the real-time changes of the quadrilateral vertex coordinates to generate the dynamic geometric correction factor, which may include: Step 400: Initialize the historical baseline data as the coordinate mean of the quadrilateral vertices in a non-congested network state, and dynamically update the historical baseline data according to the current network traffic state to generate an adaptive baseline reference value; Step 401, calculating the difference between the real-time side length and the adaptive baseline reference value to obtain the side length ratio change rate; Step 402: Based on the side length ratio change rate and the sudden increase in packet loss rate of the network topology transmission path, a baseline deviation of the diagonal angle is calculated. The baseline deviation is dynamically amplified or suppressed according to the real-time delay fluctuation amplitude to generate a diagonal angle offset. Step 403: Analyze the decay law of the quadrilateral vertex symmetry with the network congestion state, and generate a symmetry decay coefficient according to the degree of deviation between the vertex coordinates and the ideal symmetric position; Step 404, define the ideal symmetry position as the geometric center symmetry coordinate of the four vertices of the quadrilateral, and combine the diagonal angle offset to calculate the Euclidean distance between each vertex coordinate and the ideal symmetry position in real time to generate the symmetry deviation degree; Step 405, according to the historical data of network congestion events, the mapping relationship between the symmetry deviation degree and the congestion intensity is established to generate the symmetry attenuation coefficient; Step 406, the length ratio change rate, the diagonal angle offset and the symmetry attenuation coefficient are normalized, and based on the real-time transmission priority of the network topology path, the correction proportion of each parameter is dynamically adjusted to obtain the corrected parameter; Step 407, the corrected parameters are fused to generate a dynamic geometric correction factor reflecting the abnormal propagation path of the network topology.

[0061] In the embodiment of the application, in the initial stage of network use or in the period when the network is stable and not congested (such as the late night period when there is no business peak), the coordinate data of four target detection points (core switch nodes, border router nodes, key business server nodes and nodes where time servers are located) are continuously collected. After multiple collections, the mean value of each node coordinate on the horizontal and vertical coordinates is calculated to obtain the mean value of the four vertex coordinates of the quadrilateral, which is used as historical baseline data.

[0062] With the running of the network, the network traffic state is monitored in real time. When the network traffic changes, such as when the traffic increases significantly during the business peak period or a sudden traffic growth occurs (such as the start of a large-scale data download task), the historical baseline data is updated according to the current traffic. Weighted average can be used to give higher weight to recent coordinate data, so that the updated baseline data can better reflect the actual state of the network to generate adaptive baseline reference values. For example, the weight of recent data is 0.7 and the weight of historical data is 0.3. The current coordinate data and the historical baseline data are weighted and averaged according to the weight to obtain new baseline reference values.

[0063] Step 401, according to the real-time coordinates of the four target detection points, the distance between two points is calculated to calculate the real-time length of the four edges of the quadrilateral. For example, for one edge of the quadrilateral, the coordinates of the two end points are ( , ) and ( , ), and the length of the edge is the square sum of the difference between the horizontal and vertical coordinates of the two points. After calculating the real-time length of the four edges, the real-time length of each edge is compared with the length of the corresponding edge in the adaptive baseline reference value. When calculating the difference rate, the difference between the real-time length and the baseline length is divided by the baseline length to obtain the length ratio change rate of each edge. For example, the real-time length of a certain edge is , the length of which is , and the ratio of change of the side length is Through such calculation, the change of each side relative to the baseline state can be clearly understood, and the change of the network topology in the side length dimension can be determined.

[0064] In step 402, the real-time coordinates of the four vertices of the quadrilateral are known, such as vertices A(xA, yA), B(xB, yB), C(xC, yC), and D(xD, yD). The slope of the diagonal AC connecting vertices A and C is calculated according to the slope formula , and the slope of the diagonal BD connecting vertices B and D is also calculated . Then, the included angle between the two diagonals is calculated using the relationship between the slope and the included angle (such as the tangent function) . In this way, the real-time included angle obtained can accurately reflect the angle relationship between the diagonals of the current quadrilateral.

[0065] The calculated real-time included angle is compared with the included angle of the diagonal in the adaptive baseline reference value . The difference between them, i.e., the baseline deviation of the included angle, is calculated = |θ - θ0| This deviation value can measure the degree of change of the current quadrilateral diagonal included angle relative to the baseline state.

[0066] The packet loss rate of the network topology transmission path is monitored in real time. When a sudden increase in packet loss rate occurs, it indicates that the network transmission is abnormal. At this time, the number of unstable factors in the network increases, and the change of the diagonal included angle is more likely to be caused by network problems. Further analysis of the real-time delay fluctuation amplitude is performed. If the delay fluctuation amplitude is large, it means that there are many unstable factors in the network. In order to more accurately reflect the influence of the network topology change on the angle of the quadrilateral shape, the baseline deviation is amplified, such as multiplying the deviation value by a coefficient greater than 1; if the delay fluctuation amplitude is small, it means that the network is relatively stable, and the baseline deviation is suppressed, for example, the deviation value is multiplied by a coefficient less than 1. Through such dynamic adjustment of the baseline deviation, the diagonal included angle offset is finally generated, which can more accurately reflect the influence of the network topology change on the angle of the quadrilateral shape.

[0067] ​​​​​​​​Step 403: By observing and analyzing a large number of network congestion events, the changes in the vertex symmetry of the quadrilateral at different congestion levels are recorded. For example, when the network congestion level is low, the quadrilateral's shape is relatively regular, and the vertex symmetry is good. As the congestion level increases, the quadrilateral's shape may become distorted, and the vertex symmetry gradually deteriorates. Through repeated observations and summaries, a pattern is established in which the vertex symmetry of the quadrilateral decays with network congestion.

[0068] Calculate the degree of deviation of vertex coordinates from the ideal symmetric position: According to the real-time coordinates of the four target detection points, the ideal symmetrical position is the geometric center symmetrical coordinates of the quadrilateral vertices. Calculate the geometric center coordinates of the four vertices of the quadrilateral, that is, take the average of the four vertex horizontal coordinates as the center horizontal coordinate =4 + + ′+ ′, the average value of the vertical coordinate is taken as the central vertical coordinate =4 + + ′+ ′. For each vertex coordinate, calculate the Euclidean distance between it and the geometric center symmetric coordinate. Take vertex A ( , ) as an example, the Euclidean distance = + By calculating the Euclidean distance of each vertex, the degree of deviation of each vertex from the ideal symmetric position is obtained.

[0069] Based on the calculated deviation and the summarized attenuation rules, a preliminary symmetry attenuation coefficient is generated. The greater the deviation, the less symmetric the quadrilateral is, and the corresponding symmetry attenuation coefficient increases. For example, a relationship between the deviation and the symmetry attenuation coefficient is set. If the deviation exceeds a certain threshold, the symmetry attenuation coefficient is increased according to a specific rule, thus reflecting the structural changes in the network topology caused by congestion.

[0070] Step 404 emphasizes again that the ideal symmetric position is the geometric center symmetric coordinates of the quadrilateral vertices. By calculating the geometric center coordinates of the four vertex coordinates, a reference point is obtained, which represents the center position of the quadrilateral in the ideal symmetric state. For each vertex coordinate, the Euclidean distance between it and the geometric center symmetric coordinate is accurately calculated. The Euclidean distance here is obtained by calculating the square root of the sum of the differences between the horizontal and vertical coordinates. For example, for vertex B ( , ), Euclidean distance = + After calculating the Euclidean distance of all vertices, these distances reflect the deviation of each vertex from the ideal symmetric position.

[0071] In combination with the diagonal angle offset, the calculated Euclidean distance is comprehensively considered. Because the diagonal angle offset also reflects the shape change of the quadrilateral, it will affect the symmetry evaluation. For example, when the diagonal angle offset is large, even if the Euclidean distance of a certain vertex is relatively small, due to the change of the diagonal angle, it will also affect the symmetry deviation of the vertex. The Euclidean distance considering the diagonal angle offset is taken as the symmetry deviation of the vertex, and the symmetry deviations of multiple vertices collectively reflect the overall symmetry deviation of the quadrilateral. Through this comprehensive consideration, the symmetry of the quadrilateral can be more comprehensively evaluated.

[0072] Step 405, collect historical data of network congestion events, including the symmetry deviation of the quadrilateral under different congestion intensities. In actual network operation, record the congestion intensity (such as measured by network traffic size, packet loss rate, etc.) when each congestion event occurs, as well as the coordinate information of each vertex of the quadrilateral at that time, and then calculate the symmetry deviation. Analyze the collected data and establish the mapping relationship between the symmetry deviation and the congestion intensity through statistical methods (such as establishing a table of corresponding relationships, etc.).

[0073] For example, collect a large amount of relevant data of network congestion events, which cover network congestion situations under different time periods and different network environments. Divide the network congestion intensity, such as dividing the congestion intensity into several levels such as light congestion, medium congestion, and heavy congestion.

[0074] When analyzing the relationship between the symmetry deviation and the congestion intensity, for each congestion intensity level, a large amount of quadrilateral vertex coordinate data is processed. Taking the light congestion level as an example, the symmetry deviation of numerous quadrilaterals under this congestion level is counted, and the deviation of each quadrilateral vertex from the ideal symmetric position is recorded, such as some quadrilateral vertices having smaller Euclidean distance from the ideal symmetric position and some having larger Euclidean distance. Through the arrangement of these data, the range of the symmetry deviation under the light congestion level is obtained, such as the proportion of quadrilaterals with Euclidean distance within a certain interval.

[0075] For the medium congestion level, a large amount of quadrilateral vertex coordinate data is also analyzed. By calculating the Euclidean distance of each quadrilateral vertex from the ideal symmetric position, the distribution of the symmetry deviation under this congestion level can be obtained. For example, it is found that when the congestion is medium, the symmetry deviation of most quadrilaterals is concentrated in a certain specific numerical interval.

[0076] Under the heavy congestion level, a large amount of quadrilateral vertex coordinate data is analyzed in detail, and the range of symmetry deviation under the congestion level is obtained by calculating the Euclidean distance between the vertex and the ideal symmetry position and combining factors such as the diagonal angle offset.

[0077] Through the analysis of a large amount of quadrilateral vertex coordinate data under different congestion intensity levels, the corresponding relationship between each congestion intensity level and the symmetry deviation range can be determined. For example, under light congestion, the symmetry deviation is small; under moderate congestion, the symmetry deviation is moderate; and under heavy congestion, the symmetry deviation is large. After the symmetry deviation of a certain quadrilateral is calculated in real time, the corresponding relationship established previously can be used to accurately determine the congestion intensity of the current network. For example, if the symmetry deviation calculated in real time falls within the deviation range corresponding to the light congestion level, it can be judged that the current network is in a light congestion state.

[0078] According to the determined congestion intensity, the final symmetry attenuation coefficient can be generated. Since the attenuation law of the symmetry of the quadrilateral vertex under different congestion intensities is summarized, such as the symmetry of the quadrilateral vertex gradually deteriorates as the congestion intensity increases, the symmetry attenuation coefficient that can more accurately reflect the influence of the current network congestion state on the symmetry of the quadrilateral can be generated according to the current congestion intensity and in combination with these attenuation laws. This coefficient can effectively reflect the degree of change of the network topology, In step 406, the edge length ratio change rate, the diagonal angle offset, and the symmetry attenuation coefficient are normalized. Their values are mapped to a unified interval (such as [0, 1]), eliminating the differences in dimensions and numerical ranges between different parameters, so that they have comparability. For example, for the edge length ratio change rate , if its original value range is , , then the normalized value = − − . In this way, the three parameters are compared under the same scale.

[0079] Real-time transmission priority information of the network topology path is acquired in real time. Different service or data transmission paths can have different priorities, such as a critical service data transmission path having a higher priority and an ordinary data transmission path having a lower priority. The priority information of each path is recorded by a network management system. Different correction ratios are set for the three parameters according to the priority of each path. For a path with a high priority, the corresponding parameters have a greater weight in correction and have a greater impact on the final result. For example, for a critical service data transmission path, the correction ratio of the edge length ratio change rate is 0.4, the correction ratio of the diagonal line angle offset is 0.4, and the correction ratio of the symmetry decay coefficient is 0.2; for an ordinary data transmission path, the correction ratios can be 0.2, 0.3, and 0.5, respectively. Through such dynamic adjustment, the corrected edge length ratio change rate, diagonal line angle offset, and symmetry decay coefficient are obtained, so that these parameters can more accurately reflect the change of the network topology under different priority paths.

[0080] In step 407, the corrected edge length ratio change rate, diagonal line angle offset, and symmetry decay coefficient are fused according to certain rules. Here, a weighted summation method is used, and different weights are given according to the importance of each parameter in reflecting the abnormal propagation path of the network topology. For example, according to the actual situation of the network and the influence degree of the topology change, the weight of the edge length ratio change rate is 0.3, the weight of the diagonal line angle offset is 0.4, and the weight of the symmetry decay coefficient is 0.3.

[0081] The dynamic geometric correction factor is calculated as follows: The three parameters are multiplied by their respective weights and then added, that is, the dynamic geometric correction factor F = 0.3 × the corrected edge length ratio change rate + 0.4 × the corrected diagonal line angle offset + 0.3 × the corrected symmetry decay coefficient. A comprehensive value is obtained through this weighted summation method, and this value is the dynamic geometric correction factor. It can effectively reflect the change of the abnormal propagation path of the network topology, provide a comprehensive geometric dimension correction basis for the evaluation of the time server of the network, and make the evaluation result more accurately reflect the influence of the network topology change on the time service precision.

[0082] By initializing historical baseline data and dynamically updating it based on network traffic, the baseline reference value aligns with the actual network operating status, ensuring the accuracy and reliability of the assessment. By calculating the edge length proportional change rate, the system quantifies network topology changes from the edge length dimension, visually reflecting changes in network connectivity and helping to identify anomalies related to network transmission path length. By adjusting the diagonal angle offset based on packet loss rate and latency fluctuation, and comprehensively considering the impact of various network factors on topology angles, the system can more accurately capture angular changes in the network topology and identify potential network transmission issues. By analyzing the symmetry of quadrilateral vertices and mapping it to congestion intensity, the system generates a symmetry attenuation coefficient. This evaluates network topology from the perspective of structural symmetry, effectively reflecting the extent to which network congestion damages the topology structure and providing early warning of network failures. Parameters are normalized and dynamically adjusted to eliminate dimensionality differences. By integrating transmission priority, each parameter is appropriately adjusted, ensuring that the assessment results are more consistent with actual network conditions. The generated dynamic geometric correction factor integrates multiple parameters that reflect changes in network topology. It can effectively characterize the propagation path of network topology anomalies, provide a powerful basis for geometric dimension correction for the timing accuracy assessment of network time servers, improve the accuracy of assessment results, and help operation and maintenance personnel quickly locate areas of network topology anomalies.

[0083] In a preferred embodiment of the present invention, step 5 above, fusing the dynamic geometry correction factor with the comprehensive timing accuracy evaluation value, adjusting the confidence interval of the evaluation value by the quadrilateral coverage shrinkage rate and vertex offset direction, and generating an optimized comprehensive timing accuracy evaluation value and dynamic confidence range, may include: Step 500: Analyze the impact of the dynamic geometry correction factor on the expansion of the credible interval boundary of the comprehensive timing accuracy evaluation value based on the correlation between the coverage shrinkage rate of the quadrilateral and the sudden change of network traffic, and generate a boundary expansion coefficient; Step 501: Based on the boundary expansion coefficient, the mapping relationship between the vertex offset direction and the network topology anomaly propagation path is extracted, the anomaly propagation direction priority is identified, and based on the anomaly propagation direction priority, the dynamic distribution correction ratio of the comprehensive timing accuracy evaluation value within the credible interval is determined; In step 502, the credible interval of the comprehensive timing accuracy evaluation value is initially scaled based on the boundary expansion coefficient to obtain the credible interval after initial scaling. The credible interval after initial scaling is then subjected to secondary dynamic calibration in combination with the dynamic distribution correction ratio to generate an optimized comprehensive timing accuracy evaluation value and a dynamic confidence range.

[0084] In the embodiment of the present invention, during the network operation, a high-precision coordinate monitoring device is used to track the coordinates of the four vertices of the quadrilateral in real time. The vertex coordinates are recorded at a fixed time interval (e.g., every second). After recording the coordinates for a certain period of time (e.g., 1 minute), the area of ​​the quadrilateral is calculated. Assume that the coordinates of the four vertices of the quadrilateral at the initial moment are A( , )、B( , )、C( ′, ′)、D( ′, ′), and the initial area is obtained according to the quadrilateral area formula (for example, by dividing the quadrilateral into triangles to calculate the area). After a period of time (such as 1 minute), record the vertex coordinates again and calculate the area to obtain The shrinkage rate is calculated by the formula ×1 calculates the shrinkage rate of the quadrilateral coverage area.

[0085] Use network traffic monitoring tools (such as traffic sensors and network traffic analysis software) to monitor network traffic in real time. Not only should you record the traffic volume (e.g., the amount of data transmitted per second, in bytes), but you should also monitor for traffic bursts. A traffic burst is considered a traffic burst if the network traffic value increases significantly (e.g., by more than a certain percentage, such as 50% of the original traffic value) within a short period of time (e.g., 10 seconds).

[0086] Compare and analyze the calculated shrinkage rate of the quadrilateral coverage area with sudden changes in network traffic. By plotting a time series graph of shrinkage rate and traffic volume, we can observe the changing trend of the shrinkage rate during sudden changes in traffic volume (such as bursts). For example, if a sudden increase in network traffic is accompanied by a simultaneous increase in shrinkage rate, and this pattern persists across multiple sudden changes in traffic volume, this indicates a positive correlation between the shrinkage rate and sudden changes in network traffic volume. Conversely, if the shrinkage rate remains unchanged or even decreases when traffic volume increases, this indicates an insignificant or negative correlation between the two.

[0087] The dynamic geometry correction factor includes parameters such as the side length ratio change rate, diagonal angle offset, and symmetry decay coefficient. When analyzing its impact on the expansion of the credible interval bounds of the comprehensive timing accuracy evaluation value, it is first necessary to clarify the changes in each parameter. For example, the side length ratio change rate is obtained by comparing the length ratios of the quadrilateral's sides at different times. If the side length ratio change rate increases, it indicates that the quadrilateral's side lengths have changed significantly, which may affect the credible interval bounds of the timing accuracy evaluation value. The diagonal angle offset reflects the difference between the diagonal angle and the angle in the ideal symmetrical state. When the offset increases, it will also affect the credible interval bounds. The symmetry decay coefficient comprehensively considers the changes in the symmetry of the quadrilateral's vertices. An increase in the decay coefficient indicates a decrease in symmetry, which will also affect the credible interval bounds.

[0088] Based on the correlation analysis between the quadrilateral coverage shrinkage rate and network traffic mutations, combined with the parameter changes of the dynamic geometry correction factor, we determine their impact on the expansion of the credible interval boundary. For example, if the quadrilateral coverage shrinkage rate is positively correlated with network traffic mutations and the rate of change of the side length ratio in the dynamic geometry correction factor increases, then the credible interval boundary may expand outward. Conversely, if the shrinkage rate is negatively correlated with traffic mutations and the rate of change of the side length ratio decreases, the credible interval boundary may shrink inward.

[0089] By comprehensively considering the correlation between the shrinkage rate and sudden changes in network traffic volume, as well as the changes in the dynamic geometry correction factor parameters, a coefficient is determined to reflect the impact of the dynamic geometry correction factor on the expansion of the credible interval boundary, namely the boundary expansion coefficient. For example, when the correlation between the shrinkage rate and sudden changes in network traffic volume is high and the dynamic geometry correction factor parameters vary significantly, the boundary expansion coefficient may be large, such as 1.5. Conversely, when the correlation is low and the parameter changes are small, the boundary expansion coefficient may be small, such as 0.5. In this way, the boundary expansion coefficient can accurately reflect the impact of network state changes on the credible interval boundary.

[0090] Step 501: Continuously monitor the coordinate changes of the quadrilateral vertices, specifically calculate the difference in coordinates of each vertex at different times. Take vertex A as an example, assuming that at time The coordinates of , ), at the moment The coordinates of , ), then the difference is ( − , − ). Determine the offset direction of the vertex based on the sign and size of the difference. If − >0 and If > 0, the offset direction of vertex A is right-up; if If < 0 and If < 0 and If < 0, the offset direction of vertex A is left-down, and so on.

[0091] By monitoring the path changes of data transmission in the network, the network topology monitoring tool (such as network topology discovery software) is used to obtain the data transmission path between each node in the network. At the same time, the path with a high packet loss rate is focused on, and the packet loss rate refers to the proportion of the number of lost data packets to the total number of transmitted data packets within a certain time. When the packet loss rate of a certain path exceeds a certain threshold (such as 5%), it is determined that the path is an abnormal propagation path. By analyzing the changes of data transmission paths in the network topology and the packet loss rate, the abnormal propagation path of the network topology is determined.

[0092] The vertex offset direction is compared and analyzed with the abnormal propagation path of the network topology. For example, when a certain vertex offsets in a certain direction, it is checked whether there is an abnormal propagation path related to it in the network topology. If the vertex offset direction is consistent with the abnormal increase direction of the traffic of a certain path in the network, and the packet loss rate of the path is high, it is considered that there is a correlation between them. For example, vertex A offsets in a certain direction, and the traffic on the related path suddenly increases and the packet loss rate exceeds the threshold, so it can be determined that the offset direction of vertex A is mapped to the abnormal propagation path of the path.

[0093] According to the mapping relationship between the vertex offset direction and the abnormal propagation path of the network topology, the traffic size, packet loss rate, and the influence degree on the time accuracy of the path are comprehensively considered to prioritize the abnormal propagation direction. For example, if the data transmission on the network topology path corresponding to the offset direction of a certain vertex has a greater influence on the time accuracy (such as a larger change in time deviation on the path), and the packet loss rate of the path is high and the traffic is also large, then the priority of the abnormal propagation direction is high; on the contrary, if the path corresponding to the offset direction of a certain vertex has a smaller influence on the time accuracy, and the packet loss rate of the path is low and the traffic is also small, then the priority of the abnormal propagation direction is low. In this way, the priority of different abnormal propagation directions can be accurately identified.

[0094] ​​Based on the priority of the anomaly propagation direction and the boundary expansion coefficient, the dynamic distribution correction ratio of the comprehensive timing accuracy evaluation value within the credible interval is determined. For anomaly propagation directions with high priority, the proportion of their impact on the distribution of the comprehensive timing accuracy evaluation value within the credible interval is increased accordingly. For example, assume that the correction ratio corresponding to the anomaly propagation direction with high priority is 0.6, and the correction ratio corresponding to the anomaly propagation direction with low priority is 0.4. By setting different correction ratios, the distribution of the comprehensive timing accuracy evaluation value within the credible interval can more accurately reflect the impact of the network topology anomaly propagation path and vertex offset direction, thereby more accurately evaluating the timing accuracy.

[0095] Step 502: Initially scale the credible interval of the comprehensive timing accuracy evaluation value based on the boundary expansion coefficient. If the boundary expansion coefficient is large, it means that the network status changes significantly, and the credible interval needs to be expanded outward to increase the range of possible evaluation values. For example, the original credible interval is [ , ], the boundary expansion coefficient is 1.5, and the upper and lower limits are adjusted to [ − , + ],in It is determined based on the boundary expansion coefficient and the range of the original credible interval, such as =( − )×0.5 (the 0.5 here is adjusted based on the case where the boundary expansion factor is 1.5). Conversely, if the boundary expansion factor is small, it indicates that the network state is relatively stable, and the credible interval may need to be contracted inward, narrowing the range of the evaluation value. In this way, the credible interval after initial scaling can initially reflect the impact of network state changes on the range of the evaluation value.

[0096] The process of secondary dynamic calibration combined with the dynamic distribution correction ratio is as follows: Combined with the dynamic distribution correction ratio determined in step 501, a secondary dynamic calibration is performed on the initially scaled credible interval. A higher correction ratio is assigned to anomaly propagation directions with a higher priority, and the distribution of evaluation values ​​within the credible interval is adjusted based on this ratio. For example, if an anomaly propagation direction has a high priority and a correction ratio of 0.6, the distribution of evaluation values ​​within the credible interval is redistributed so that the evaluation values ​​are more likely to be in the area associated with that anomaly propagation direction.

[0097] Specifically, the confidence interval is divided into several sub-intervals, and the probability distribution of the evaluation value in these sub-intervals is adjusted according to the correction ratio. If the sub-interval correction ratio corresponding to the abnormal propagation direction is 0.6, a higher weight is given when calculating the probability of the evaluation value falling within the sub-interval, so that the evaluation value is more likely to take a value within the sub-interval. Through this secondary dynamic calibration, the optimized comprehensive timing accuracy evaluation value and dynamic confidence range are generated, so that the evaluation value can more accurately reflect the influence of network topology changes on timing accuracy, and the dynamic confidence range can more reasonably represent the reliability of the evaluation value, providing more reliable results for the timing accuracy evaluation of the network time server.

[0098] By analyzing the correlation between the quadrilateral coverage range contraction rate and the network traffic mutation, and the influence of the dynamic geometric correction factor on the boundary expansion of the confidence interval, the boundary expansion coefficient is generated. This enables the confidence interval of the comprehensive timing accuracy evaluation value to be reasonably adjusted according to the actual network situation, improving the reliability of the evaluation value and enabling it to more accurately reflect the influence of network topology changes on timing accuracy, avoiding errors in the evaluation value caused by network changes. The mapping relationship between the vertex offset direction and the network topology abnormal propagation path is extracted, the priority of the abnormal propagation direction is identified, and the dynamic distribution correction ratio of the comprehensive timing accuracy evaluation value within the confidence interval is determined, which helps to understand the influence mechanism of network topology abnormalities on timing accuracy and adjust the distribution of the evaluation value within the confidence interval according to the priority of the abnormal propagation direction, enabling the evaluation value to more accurately reflect the influence of the network topology abnormal propagation path. The confidence interval is initially scaled and combined with the dynamic distribution correction ratio for secondary dynamic calibration to generate the optimized comprehensive timing accuracy evaluation value and dynamic confidence range. This enables the comprehensive timing accuracy evaluation value to be dynamically adjusted according to network topology changes, and the dynamic confidence range can more reasonably represent the reliability of the evaluation value, thereby improving the reliability and effectiveness of timing accuracy evaluation and providing more accurate evaluation results for the operation of the network time server, which helps to timely discover and solve timing accuracy problems.

[0099] In a preferred embodiment of the present application, the step 6, based on the optimized comprehensive timing accuracy evaluation value and dynamic confidence range, in combination with the preset congestion state classification threshold, outputs the timing accuracy stability level of the network time server in the current dynamic network environment, abnormal event warning information and network topology abnormal area positioning report based on the identification of the quadrilateral vertex offset direction, which can include: Step 600, multi-dimensional matching of the dynamic confidence range of the optimized comprehensive timing accuracy evaluation value and the preset congestion state classification threshold to determine the timing accuracy stability level; Step 601, generating a level reliability parameter according to the deviation degree of the timing accuracy stability level and the dynamic confidence range; Step 602, based on the level of credibility parameter, set the dynamic abnormal event detection threshold, identify the abnormal event of evaluation value >= dynamic abnormal event detection threshold, and analyze the propagation direction and diffusion rate of abnormal events on the topological path in combination with the spatio-temporal correlation of the quadrilateral vertex offset direction; Step 603, according to the propagation direction and diffusion rate, combined with the coupling relationship between the vertex offset direction and the network topological path, locate the starting node and relay node of the abnormal propagation path, and associate the node performance baseline data; Step 604, based on the node performance baseline data and the diffusion rate, match the feature library of historical network congestion events, generate a repair priority list of core switch nodes and boundary router nodes; Step 605, according to the repair priority list, combined with the level of credibility parameter, generate a positioning report containing node identification, abnormal propagation path and repair suggestion.

[0100] In the embodiment of the application, the preset congestion state grading threshold is combed. These thresholds are set according to the demand of network service for timing accuracy, combined with historical network data and experience, for example, different threshold intervals corresponding to four levels of "excellent", "good", "medium" and "poor".

[0101] The dynamic confidence range of the optimized comprehensive timing accuracy evaluation value is matched with these thresholds in multiple dimensions. Not only whether the evaluation value itself falls within a certain threshold interval is compared, but also the relationship between the upper limit and lower limit of the dynamic confidence range and the threshold boundary is considered. For example, if the lower limit of the dynamic confidence range is higher than the threshold lower limit corresponding to the "good" level, and the upper limit is lower than the threshold upper limit corresponding to the "good" level, then the timing accuracy stability level is preliminarily determined as "good". If the lower limit of the dynamic confidence range is lower than the threshold lower limit corresponding to the "poor" level, even if the evaluation value itself is in other intervals, it may also be determined as "poor" according to the rules. Through this multi-dimensional matching, the evaluation value and its fluctuation range are considered comprehensively to finally determine the timing accuracy stability level.

[0102] Step 601, calculate the deviation degree of the timing accuracy stability level from the dynamic confidence range. For example, if the timing accuracy stability level is determined as "good", and the lower limit of the dynamic confidence range is close to the threshold lower limit of the "medium" level, and the upper limit is close to the threshold upper limit of the "excellent" level, it means that there is a certain uncertainty in the level determination. The level confidence parameter is generated by quantifying the deviation degree. A rule can be set, such as when the distance between the boundary of the dynamic confidence range and the corresponding level threshold boundary is less than a certain proportion (such as 10%), the confidence parameter is reduced; the greater the distance, the higher the confidence parameter. Assuming that the confidence is represented by a numerical interval of 0-1, if the deviation degree is small, the confidence parameter can be set to 0.8-1; if the deviation degree is large, the confidence parameter is set to 0-0.2. The final level confidence parameter can reflect the reliability of the current timing accuracy stability level determination.

[0103] Step 602, set the dynamic abnormal event detection threshold based on the level confidence parameter. The higher the confidence parameter, the more stringent the set threshold; the lower the confidence parameter, the more relaxed the threshold. For example, when the level confidence parameter is 0.9, the dynamic abnormal event detection threshold is set to a value close to the upper limit of the corresponding level threshold; when the level confidence parameter is 0.3, the threshold can be appropriately reduced. In the comprehensive timing accuracy evaluation value, find the case where the evaluation value is greater than or equal to the dynamic abnormal event detection threshold, and once it is found, it is identified as an abnormal event.

[0104] Analyze abnormal events in combination with the spatio-temporal correlation of quadrilateral vertex offset directions. Record the offset direction and time point of each vertex of the quadrilateral when the abnormal event occurs, and analyze the change of the vertex offset direction in the subsequent time. If multiple vertices continue to offset in the same direction within a certain time period, and the offset angle gradually increases, it means that the abnormal event has a specific propagation direction on the topology path; by calculating the distance or angle change of the vertex offset per unit time, the diffusion rate of the abnormal event is evaluated.

[0105] Step 603, according to the propagation direction and diffusion rate of the abnormal event obtained by analysis, combined with the coupling relationship between the vertex offset direction and the network topology path, the node positioning is carried out. Observe the consistency of the vertex offset direction and the data transmission path in the network, if the offset direction of a certain vertex is consistent with the abnormal increase direction of the data flow on a certain network topology path, and the path meets the propagation direction of the abnormal event, then the node on this path is likely to be a related node of the abnormal propagation path. By tracking the initial direction and change of the vertex offset in the propagation process of the abnormal event, the starting node of the abnormal propagation path is determined; then according to the change of the vertex offset direction and the propagation direction in the diffusion process, the nodes playing a relay role on the path are found. Correlate the performance baseline data of these nodes, including the historical normal running state data indexes of the nodes such as processing capacity, bandwidth utilization, delay, etc.

[0106] In step 604, the performance baseline data of the located abnormal propagation path nodes is combined with the diffusion rate of the abnormal event and matched against a signature library of historical network congestion events. The signature library stores information such as node performance change characteristics and propagation rates for various types of previous network congestion events. Through comparative analysis, it is determined which historical events in the signature library the current abnormal event is similar to, thereby determining the priority of key nodes such as core switch nodes and border router nodes during the repair process. For example, if the performance baseline data of a node shows that its processing capability has decreased after the abnormal event occurs, and the abnormal event has a rapid diffusion rate, and the node has a high match with the event characteristics in the signature library that caused serious network failures, then the node's repair priority is high; otherwise, the repair priority is low. Ultimately, a list containing the repair priorities of each key node is generated.

[0107] Step 605: Generate a positioning report based on the repair priority list and the level credibility parameters. The report clearly lists the identifiers of the abnormal nodes (such as the node's IP address, device name, etc.), and describes in detail the abnormal propagation path, including information such as the starting node, relay nodes, and data transmission direction. Specific repair suggestions are given for nodes of different priorities. For nodes with high repair priority, it is recommended to immediately check, restart, or replace the equipment; for nodes with low repair priority, it is recommended to conduct continuous monitoring or perform maintenance when the network load is low. At the same time, the credibility of the timing accuracy stability level is noted in the report, so that operation and maintenance personnel can fully understand the current status and potential risks of the network time server.

[0108] The timing accuracy stability level is determined through multi-dimensional matching, comprehensively considering the evaluation value and its fluctuation range. This ensures that the level determination is more accurate and responsive to the actual network status, providing operators with an intuitive and accurate reference for timing accuracy stability. A level confidence parameter is generated to quantify the reliability of the level determination, allowing operators to understand the reliability of the timing accuracy stability level and avoid misjudgments caused by inaccurate level determinations. Dynamically setting anomaly detection thresholds and analyzing anomaly propagation characteristics enables timely and accurate identification of anomalies and understanding their propagation patterns, facilitating proactive measures to prevent anomalies from escalating. Accurately locating the starting node and relay nodes along the anomaly propagation path and correlating them with performance baseline data provides clear targets and data support for rapid troubleshooting and resolution of network issues, improving fault location efficiency. A repair priority list is generated to rationally prioritize repairs based on historical event characteristics and current anomaly conditions, ensuring that critical nodes are repaired first, optimizing the network repair process and reducing the impact of network outages. Generates a detailed positioning report containing information such as nodes, paths, and repair suggestions, providing operations and maintenance personnel with a clear troubleshooting guide. It also indicates the level of credibility to assist operations and maintenance personnel in making scientific decisions and ensure the stable operation of the network time server in a dynamic network environment.

[0109] As Figure 2 indicated, the embodiment of the application also provides a network time server time service precision evaluation system, comprising: a data acquisition module, configured to acquire core index data and auxiliary index data of the network time server in a dynamic network environment in real time; a data processing module, configured to perform normalization processing on the core index data to generate a standardized core index data set, and dynamically integrate according to fluctuation characteristics and correlation of each core index to generate a comprehensive core index representation value; a dynamic compensation correction module, configured to perform dynamic compensation correction on the comprehensive core index representation value in combination with network delay distribution characteristics and clock frequency offset in the auxiliary index data to generate a comprehensive time service precision evaluation value; a dynamic geometric correction module, configured to calculate a length ratio change rate of a quadrilateral vertex coordinate, a diagonal angle offset amount and a symmetry decay coefficient based on position information of four target detection points in the auxiliary index data to generate a dynamic geometric correction factor; an evaluation optimization module, configured to fuse the dynamic geometric correction factor and the comprehensive time service precision evaluation value, adjust a credible interval of an evaluation value through a contraction rate of a coverage range of the quadrilateral and a vertex offset direction, and generate an optimized comprehensive time service precision evaluation value and a dynamic confidence range; a report generation module, configured to output a time service precision stability level, an abnormal event early warning information and a network topology abnormal area positioning report based on a vertex offset direction of the quadrilateral based on the optimized comprehensive time service precision evaluation value and the dynamic confidence range in combination with a preset congestion state grading threshold.

[0110] It should be noted that the system is a system corresponding to the above method, and all implementation manners in the above method embodiment are applicable to this embodiment and can achieve the same technical effects.

[0111] The embodiment of the application also provides a computer readable storage medium storing instructions, when the instructions are run on a computer, the computer executes the method as described above. All implementation manners in the above method embodiment are applicable to this embodiment and can achieve the same technical effects.

[0112] The above is the preferred embodiment of the application, and it should be noted that for ordinary skilled persons in the art, without departing from the principles of the application, several improvements and refinements can be made, and these improvements and refinements should be regarded as the protection scope of the application.

Claims

1. A method for evaluating the timing accuracy of a network time server, characterized in that: The method comprises: Step 1: collect the operation data of the network time server in a dynamic network environment in real time, including core indicator data and auxiliary indicator data, and normalize the core indicator data to generate a standardized core indicator data set; Step 2: Dynamically integrate the standardized core indicator data set based on the fluctuation characteristics and correlation of each core indicator to generate a comprehensive core indicator representation value; Step 3: Combined with the auxiliary indicator data, dynamically compensate and correct the comprehensive core indicator representation value to generate a comprehensive timing accuracy evaluation value; Step 4: Based on the four target detection points in the auxiliary indicator data, the vertex coordinates of the formed quadrilateral area are obtained. According to the real-time changes in the vertex coordinates of the quadrilateral, the change rate of the quadrilateral's side length ratio, the diagonal angle offset, and the symmetry attenuation coefficient are calculated to generate a dynamic geometric correction factor. Step 5: The dynamic geometry correction factor is integrated with the comprehensive timing accuracy evaluation value. The confidence interval of the evaluation value is adjusted by the coverage shrinkage rate of the quadrilateral and the vertex offset direction to generate the optimized comprehensive timing accuracy evaluation value and dynamic confidence range. Step 6: Based on the optimized comprehensive timing accuracy evaluation value and dynamic confidence range, combined with the preset congestion status classification threshold, output the timing accuracy stability level of the network time server in the current dynamic network environment, abnormal event warning information and network topology abnormal area positioning report based on quadrilateral vertex offset direction identification.

2. The method for evaluating the timing accuracy of a network time server according to claim 1, wherein: Based on the fluctuation characteristics and correlation of each core indicator, the standardized core indicator data set is dynamically integrated to generate a comprehensive core indicator representation value, including: The time deviation series is segmented into sliding windows, and the standard deviation of the short-term fluctuation amplitude and the positive and negative slope distribution of the trend direction in each window are analyzed to generate the dynamic fluctuation characteristics of the time deviation; Based on the dynamic fluctuation characteristics of time deviation, the peak distribution characteristics of time jitter parameters are extracted, jitter events are screened using preset thresholds, the frequency and intensity accumulation values ​​of jitter events are statistically analyzed, and a quantitative parameter set of time jitter abnormality events is generated; Based on the quantitative parameter set of time jitter anomaly events, the synchronization period stability coefficient is converted from time domain to frequency domain, the main oscillation frequency and the energy proportion of its harmonic components are calculated, and the synchronization period stability spectrum quantitative index is generated; Combined with the synchronization cycle stability spectrum quantitative indicator, based on the historical baseline data of the maximum delay error value, the probability of triggering extreme events exceeding the historical threshold is calculated, and the maximum delay risk level parameter is generated by correlating it with the real-time network load status; The dynamic fluctuation characteristics of time deviation, the quantitative parameter set of time jitter abnormal events, the quantitative index of synchronization cycle stability spectrum and the maximum delay risk level parameter are input into the dynamic time window, and the time-varying correlation coefficient matrix between each parameter is calculated to construct a multi-dimensional dynamic correlation model. Based on the multi-dimensional dynamic association model, the standardized core indicator dataset is projected into feature space to extract dynamic feature vectors at different time granularities. The dynamic feature vectors are superimposed and normalized at multiple time granularities to generate comprehensive core indicator representation values ​​that reflect the coupling effects of the dynamic network environment.

3. The method for evaluating the timing accuracy of a network time server according to claim 2, wherein: Combined with auxiliary indicator data, dynamic compensation and correction are performed on the comprehensive core indicator representation value to generate a comprehensive timing accuracy evaluation value, including: According to the sudden fluctuation intensity and duration of the network delay distribution characteristics, a dynamic abnormal jump threshold is set to identify the abnormal jump intervals in the time deviation sequence that are greater than or equal to the dynamic abnormal jump threshold; The sliding window smoothing algorithm is used to compensate the abnormal jump interval in sections and calculate the time deviation correction sequence after compensation; Based on the compensated time deviation correction sequence, the accumulated drift error of the synchronization period stability coefficient is reversely calibrated to generate the drift error compensation coefficient; The time deviation correction sequence and the drift error compensation coefficient are superimposed on the comprehensive core indicator representation value to generate the timing accuracy evaluation value after preliminary compensation. Based on the real-time fluctuation amplitude of the timing accuracy evaluation value, the real-time change rate of the network delay distribution characteristics is calculated; The window length and compensation strength of the sliding window smoothing algorithm are dynamically adjusted according to the real-time change rate of the network delay distribution characteristics to generate a comprehensive timing accuracy evaluation value that includes dynamic network delay correction and clock frequency offset suppression.

4. The method for evaluating the timing accuracy of a network time server according to claim 3, wherein: Based on the four target detection points in the auxiliary indicator data, the vertex coordinates of the formed quadrilateral area are obtained. According to the real-time changes in the vertex coordinates of the quadrilateral, the change rate of the quadrilateral's side length ratio, the diagonal angle offset, and the symmetry attenuation coefficient are calculated to generate a dynamic geometric correction factor, including: Initialize the historical baseline data as the coordinate mean of the quadrilateral vertices in the non-congested network state, and dynamically update the historical baseline data according to the current network traffic state to generate an adaptive baseline reference value; Calculate the difference between the real-time side length and the adaptive baseline reference value to obtain the side length ratio change rate; Based on the rate of change of the side length ratio and the sudden increase in packet loss rate along the network topology transmission path, the baseline deviation of the diagonal angle is calculated. This baseline deviation is dynamically amplified or suppressed based on the real-time delay fluctuation amplitude to generate the diagonal angle offset. Analyze the decay law of quadrilateral vertex symmetry with network congestion, and generate the symmetry decay coefficient according to the deviation degree between the vertex coordinates and the ideal symmetric position; The ideal symmetric position is defined as the geometric center symmetric coordinates of the quadrilateral vertices, and combined with the diagonal angle offset, the Euclidean distance between each vertex coordinate and the ideal symmetric position is calculated in real time to generate the symmetry deviation; Based on historical data of network congestion events, a mapping relationship between symmetry deviation and congestion intensity is established to generate a symmetry attenuation coefficient. Normalize the side length ratio change rate, diagonal angle offset, and symmetry attenuation coefficient, and dynamically adjust the correction ratio of each parameter based on the real-time transmission priority of the network topology path to obtain the corrected parameters. The corrected parameters are fused to generate a dynamic geometric correction factor that reflects the propagation path of network topology anomalies.

5. The method for evaluating the timing accuracy of a network time server according to claim 4, wherein: The four target detection points are respectively located at a core switch node, a border router node, a key business server node and a node where a time server is located in the network topology.

6. The method for evaluating the timing accuracy of a network time server according to claim 5, wherein: The dynamic geometry correction factor is integrated with the comprehensive timing accuracy evaluation value. The confidence interval of the evaluation value is adjusted by the quadrilateral coverage shrinkage rate and vertex offset direction to generate the optimized comprehensive timing accuracy evaluation value and dynamic confidence range, including: According to the correlation between the coverage shrinkage rate of the quadrilateral and the sudden change of network traffic, the influence of the dynamic geometric correction factor on the expansion of the credible interval boundary of the comprehensive timing accuracy evaluation value is analyzed, and the boundary expansion coefficient is generated. Based on the boundary expansion coefficient, the mapping relationship between the vertex offset direction and the network topology anomaly propagation path is extracted, the anomaly propagation direction priority is identified, and based on the anomaly propagation direction priority, the dynamic distribution correction ratio of the comprehensive timing accuracy evaluation value within the credible interval is determined; Based on the boundary expansion coefficient, the credible interval of the comprehensive timing accuracy evaluation value is initially scaled to obtain the credible interval after initial scaling. Combined with the dynamic distribution correction ratio, the credible interval after initial scaling is subjected to secondary dynamic calibration to generate the optimized comprehensive timing accuracy evaluation value and dynamic confidence range.

7. The method for evaluating the timing accuracy of a network time server according to claim 6, wherein: Based on the optimized comprehensive timing accuracy evaluation value and dynamic confidence range, combined with the preset congestion status classification threshold, the system outputs the timing accuracy stability level of the network time server in the current dynamic network environment, abnormal event warning information, and a network topology abnormal area location report based on quadrilateral vertex offset direction identification, including: The dynamic confidence range of the optimized comprehensive timing accuracy evaluation value is matched with the preset congestion status classification threshold in multiple dimensions to determine the timing accuracy stability level; Generate a level confidence parameter based on the degree of deviation between the timing accuracy stability level and the dynamic confidence range; Based on the level credibility parameter, a dynamic abnormal event detection threshold is set to identify abnormal events with an evaluation value ≥ the dynamic abnormal event detection threshold. In addition, the propagation direction and diffusion rate of abnormal events on the topological path are analyzed by combining the spatiotemporal correlation of the quadrilateral vertex offset direction. Based on the propagation direction and diffusion rate, combined with the coupling relationship between vertex offset direction and network topology path, the starting node and relay nodes of the abnormal propagation path are located, and the node performance baseline data is correlated; Based on node performance baseline data and diffusion rate, the feature library of historical network congestion events is matched to generate a repair priority list for core switch nodes and border router nodes. Based on the repair priority list and combined with the level credibility parameters, a positioning report containing node identification, anomaly propagation path and repair suggestions is generated.

8. The method for evaluating the timing accuracy of a network time server according to claim 7, wherein: The core indicator data includes time deviation series, time jitter parameters, synchronization cycle stability coefficient and maximum delay error value; the auxiliary indicator data includes network delay distribution characteristics, clock frequency offset and position information of four preset target detection points.

9. A network time server timing accuracy evaluation system, the system implementing the method according to any one of claims 1 to 8, characterized in that: include: Data collection module, used to collect core indicator data and auxiliary indicator data of network time server in dynamic network environment in real time; The data processing module is used to normalize the core indicator data to generate a standardized core indicator data set, and dynamically integrate the fluctuation characteristics and correlation of each core indicator to generate a comprehensive core indicator representation value; The dynamic compensation correction module is used to dynamically compensate and correct the comprehensive core indicator representation value by combining the network delay distribution characteristics and clock frequency offset in the auxiliary indicator data to generate a comprehensive timing accuracy evaluation value; The dynamic geometry correction module is used to calculate the side length ratio change rate, diagonal angle offset and symmetry attenuation coefficient of the quadrilateral vertex coordinates in real time based on the position information of the four target detection points in the auxiliary indicator data, and generate a dynamic geometry correction factor; The evaluation and optimization module is used to integrate the dynamic geometry correction factor with the comprehensive timing accuracy evaluation value, adjust the confidence interval of the evaluation value by the quadrilateral coverage shrinkage rate and vertex offset direction, and generate the optimized comprehensive timing accuracy evaluation value and dynamic confidence range; The report generation module is used to output the timing accuracy stability level, abnormal event warning information and network topology abnormal area positioning report based on the optimized comprehensive timing accuracy evaluation value and dynamic confidence range, combined with the preset congestion status classification threshold.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program, which implements the method according to any one of claims 1 to 8 when executed by a processor.

Images