Communication network performance monitoring optimization method and system

Abstract

The invention relates to the technical field of communication network optimization, and discloses a communication network performance monitoring optimization method and system. The method comprises the following steps: acquiring performance nodes influencing transmission quality in a communication network, and constructing a network transmission quality model; configuring a network monitoring probe model, performing performance test, acquiring the time delay detection capability and the packet loss rate detection precision, and integrating the network monitoring probe model and the packet loss rate detection precision into a network transmission quality model; calculating the total data flow load in a preset time window, obtaining a theoretical monitoring load in combination with the time delay detection capability of the probe, generating a data sampling frequency control instruction based on the theoretical monitoring load and the packet loss rate detection precision, and dynamically adjusting the working frequency of a probe data acquisition module according to the data sampling frequency control instruction. And the optimization effect of the method on the network transmission quality is evaluated. According to the method, by dynamically adjusting the sampling frequency, accurate adaptation to the network state is achieved, the real-time performance of monitoring and the resource utilization rate are improved, meanwhile, multi-dimensional performance indexes are integrated, and comprehensive support is provided for network transmission quality optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of communication network optimization, and in particular to a communication network performance monitoring and optimization method and system. Background Art

[0002] With the rapid development of communications technology, various network application scenarios continue to expand, from traditional voice communications to high-definition video transmission, IoT data exchange, and cloud computing services. The data traffic carried by communication networks is growing exponentially, and network structures are becoming increasingly complex. In this context, the stability and reliability of network transmission quality have become key factors affecting user experience and business continuity. However, existing communication network performance monitoring methods often have many limitations when facing dynamically changing network environments. Currently, most network monitoring systems rely on a fixed sampling frequency for data collection, making it difficult to adapt to drastic fluctuations in network traffic. When the network is under high load, a fixed sampling frequency can lead to redundant monitoring data, occupying excessive network resources and impacting the transmission of normal services. During periods of low load, insufficient sampling can lead to missing critical performance anomalies, resulting in delayed fault detection. Furthermore, traditional monitoring models often treat performance indicators such as latency and packet loss rate in isolation, lacking a comprehensive consideration of the relationships between various performance nodes. This makes it difficult to build a comprehensive network transmission quality assessment system, and the formulation of optimization strategies lacks accurate data support. The performance tests of existing monitoring probes are insufficiently integrated with the actual network environment. The probes' latency detection capabilities and packet loss rate accuracy are not dynamically correlated with the network's real-time load, significantly reducing the effectiveness and relevance of the monitoring data. These issues collectively make it difficult for communication network performance monitoring to meet the real-time, accuracy, and resource utilization requirements of complex network environments. A monitoring optimization method is needed that can dynamically adapt to network conditions and integrate multi-dimensional performance indicators. Summary of the Invention

[0003] The purpose of the present invention is to provide a communication network performance monitoring and optimization method and system to solve the problems raised in the above background technology.

[0004] To achieve the above objectives, the present invention provides a communication network performance monitoring and optimization method, the method comprising: Collecting performance nodes that affect transmission quality in the communication network, and building a network transmission quality model based on the performance nodes; Configuring a network monitoring probe model, performing a performance test operation on the network monitoring probe model, and obtaining the delay detection capability and packet loss rate detection accuracy of the network monitoring probe model; Integrating the network monitoring probe model into the network transmission quality model; Calculate the total data traffic load of the communication network within a preset time window, and calculate the theoretical monitoring load of the network monitoring probe model within the preset time window in combination with the delay detection capability of the network monitoring probe model; Based on the theoretical monitoring load and packet loss rate detection accuracy, generating a data sampling frequency control instruction for the network monitoring probe model; Dynamically adjust the operating frequency of the data acquisition module of the network monitoring probe model according to the data sampling frequency control instruction; Evaluate the transmission quality optimization effect of the network monitoring probe model on the communication network.

[0005] Preferably, collecting performance nodes that affect transmission quality in the communication network and constructing a network transmission quality model based on the performance nodes includes: Extracting topological structure information of a communication network and dividing the communication network into a plurality of logical subnets; Identify key performance nodes in each logical subnet and generate a set of performance nodes for the communication network; Measuring the bandwidth carrying capacity of each performance node and analyzing external interference factors acting on the performance node; Based on the bandwidth carrying capacity and external interference factors, a data transmission balance equation of the communication network is established.

[0006] Preferably, establishing a data transmission balance equation of the communication network based on the bandwidth carrying capacity and external interference factors includes: Analyze the data transmission coupling effect between each performance node and quantify the impact of the external interference factors on each performance node; The data transmission balance equation is constructed by combining the bandwidth carrying capacity of the performance node, the data transmission coupling effect, and the influence intensity of external interference factors.

[0007] Preferably, the configuring of the network monitoring probe model and performing the performance test operation of the network monitoring probe model include: The delay response characteristics and packet loss detection accuracy of the network monitoring probe model under high load, low load and fluctuating load conditions are tested respectively to obtain the delay detection capability and packet loss rate detection accuracy.

[0008] Preferably, integrating the network monitoring probe model into the network transmission quality model includes: Analyze the impact weight of each performance node in the network transmission quality model and locate the target performance node with the largest impact weight; The network monitoring probe model is deployed in the logical subnet corresponding to the target performance node.

[0009] Preferably, the calculating the total data traffic load of the communication network within a preset time window includes: Aggregate the data traffic generated by each performance node within the preset time window to obtain the total data traffic load; The theoretical monitoring load of the computing network monitoring probe model includes: Obtaining the basic monitoring overhead of the network monitoring probe model within a preset time window, and calculating the theoretical monitoring load in combination with the total data traffic load; The generating data sampling frequency control instruction comprises: Based on the theoretical monitoring load and packet loss rate detection accuracy, the optimal sampling frequency of the network monitoring probe model within a preset time window is determined.

[0010] Preferably, the generating of the data sampling frequency control instruction includes: The data throughput of the network monitoring probe model at a unit sampling frequency is tested, and a frequency switching time plan of the data acquisition module is generated in combination with the optimal sampling frequency.

[0011] Preferably, the dynamically adjusting the operating frequency of the data acquisition module of the network monitoring probe model includes: Dividing the preset time window into discrete time periods, and evenly allocating the frequency switching time plan to each discrete time period; After a single discrete time period ends, verifying whether the actual data throughput of the data acquisition module meets expectations; The frequency switching time plan of subsequent discrete time periods is compensated and corrected according to the verification results until the adjustment operation of the entire preset time window is completed.

[0012] Preferably, the evaluating the transmission quality optimization effect of the network monitoring probe model on the communication network includes: Obtain the change in delay and packet loss rate of the communication network before and after optimization; Calculating an actual monitoring load of the communication network based on the delay variation and the packet loss rate variation; Comparing the deviation between the actual monitoring load and the theoretical monitoring load to generate a frequency control authority index of the data acquisition module; The network transmission quality optimization effect is evaluated according to the frequency control authority index.

[0013] Preferably, the present invention further includes a communication network performance monitoring and optimization system for implementing the communication network performance monitoring and optimization method as described above, the system comprising: A transmission quality modeling module is used to collect performance nodes that affect transmission quality in the communication network and build a network transmission quality model based on the performance nodes; The probe performance test module is used to configure the network monitoring probe model, perform performance test operations on the network monitoring probe model, and obtain the delay detection capability and packet loss rate detection accuracy; A model fusion module, used to integrate the network monitoring probe model into the network transmission quality model; A monitoring load calculation module, configured to calculate the total data traffic load of the communication network within a preset time window and calculate the theoretical monitoring load in combination with the delay detection capability; A sampling control module, configured to generate a data sampling frequency control instruction for the network monitoring probe model based on the theoretical monitoring load and packet loss rate detection accuracy; A dynamic adjustment module, configured to dynamically adjust the operating frequency of the data acquisition module according to the data sampling frequency control instruction; The optimization evaluation module is used to evaluate the transmission quality optimization effect of the network monitoring probe model on the communication network.

[0014] Compared with the prior art, the present invention has the following beneficial effects: By collecting performance nodes that affect transmission quality and building a network transmission quality model, we achieve a systematic analysis and integration of key network links, breaking the limitations of isolated performance indicators in traditional monitoring and providing a more comprehensive reflection of the overall network operation status. This multi-node model construction method effectively explores the correlations between various performance parameters, providing a more realistic analytical foundation for subsequent monitoring and optimization. Configuring a network monitoring probe model and conducting performance testing to determine its latency detection capability and packet loss accuracy helps clarify the probe's performance boundaries and applicable scenarios, providing a precise basis for integrating the probe with the network transmission quality model. Integrating the probe model into the transmission quality model creates a synergistic mechanism, allowing the probe's monitoring data to directly inform the model's assessment of network status, eliminating data silos and improving the integrity and coherence of the monitoring system. By calculating the total data traffic load within a preset time window and combining it with the probe's latency detection capabilities, the theoretical monitoring load is derived. This allows the monitoring load to be determined more closely than simply empirical values, effectively correlating it with the actual network load. Based on this theoretical load and the accuracy of packet loss detection, data sampling frequency control instructions are generated, enabling dynamic adjustment of the sampling frequency. This reduces unnecessary sampling during high loads to conserve resources, while increasing sampling during low loads to capture more detail. This ensures effective monitoring while improving network resource utilization. Dynamically adjusting the operating frequency of the data acquisition module enables the monitoring system to adapt to changes in network traffic in real time, enhancing monitoring flexibility and adaptability and avoiding the monitoring redundancy or insufficiency that can occur with a fixed sampling frequency. The evaluation of network transmission quality optimization can verify the effectiveness of the entire monitoring and optimization process, providing feedback for continuous improvement of the method, forming a closed loop of "monitoring-optimization-evaluation" and promoting the continuous improvement of communication network performance monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a working principle diagram of the communication network performance monitoring and optimization method according to the present invention; Figure 2 Flowchart for performance node acquisition and network transmission quality model construction; Figure 3 Flowchart constructed for data transfer balance equation; Figure 4 Flowchart for network monitoring probe model integration; Figure 5 Flowchart generated for the data sampling frequency control instruction. DETAILED DESCRIPTION

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0017] See also Figure 1 The present invention provides a communication network performance monitoring and optimization method, the method comprising: Collect performance nodes that affect transmission quality in the communication network and build a network transmission quality model based on these performance nodes. Configure a network monitoring probe model, execute performance testing operations on the model, and obtain the latency detection capability and packet loss rate detection accuracy. Integrate the network monitoring probe model into the network transmission quality model. Calculate the total data traffic load of the communication network within a preset time window. Combined with the latency detection capability, calculate the theoretical monitoring load of the network monitoring probe model within the preset time window. Based on the theoretical monitoring load and packet loss rate detection accuracy, generate data sampling frequency control instructions for the network monitoring probe model. Based on the data sampling frequency control instructions, dynamically adjust the operating frequency of the data acquisition module of the network monitoring probe model. Evaluate the effectiveness of the network monitoring probe model in optimizing the transmission quality of the communication network.

[0018] Example 1: See Figure 2, extract the topological structure information of the communication network, and divide the communication network into multiple logical subnets by analyzing the connection relationships and routing paths of network devices. The logical subnets are divided according to the physical location of the equipment, business functions, or traffic boundaries, forming independent data transmission units. Identify the key performance nodes in each logical subnet, screen them based on parameters such as node data forwarding volume, number of connected devices, and historical failure rate, and generate a performance node set for the communication network. This set includes key network devices such as router core nodes, gateway aggregation nodes, and backbone switch nodes. Measure the bandwidth carrying capacity of each performance node, perform actual data transmission tests, and record the maximum throughput rate of the node under saturation. Analyze external interference factors acting on performance nodes, including dynamic factors such as electromagnetic environment noise, adjacent channel crosstalk, and sudden high-priority service preemption. Use a spectrum analyzer to collect environmental interference signal characteristics and count the trigger frequency of abnormal traffic samples.

[0019] Based on bandwidth carrying capacity and external interference factors, a data transmission balance equation for the communication network is established. The interactions between the various parameters in the data transmission balance equation are analyzed, and a mathematical model describing the relationship between node resource allocation and interference cancellation is constructed. The data transmission state is represented using a matrix equation, with the rows corresponding to the logical subnets and the columns corresponding to the performance node attributes. Filled values ​​include measured bandwidth, interference attenuation coefficients, and node buffer capacity. After the logical subnets are deployed, packet marking techniques are used to collect actual data transmission path parameters, including the number of transmission hops, path switching frequency, and inter-subnet latency. When validating the matrix equation, the values ​​of the variables are initialized synchronously, and the logical subnet topology constraints are input.

[0020] Obtain the change in latency and packet loss rate of the communication network before and after optimization. Perform equal-length network status snapshot collection before and after deploying the probe, compare the difference in the end-to-end latency mean of the same service flow, and calculate the absolute change value. Count the change in the number of lost packets with the same source and destination within the preset packet loss detection period to generate the relative change in packet loss rate. Calculate the actual monitoring load of the communication network. This load value is determined by the total amount of network resources occupied by the probe model during operation, including the monitoring data replication bandwidth, analysis and calculation processing overhead, and result reporting traffic. Compare the deviation between the actual monitoring load and the theoretical monitoring load, use the standard deviation algorithm to quantify the degree of dispersion of the two, and match the preset correction coefficient table according to the deviation value interval.

[0021] Generates a frequency control authority index for the data acquisition module. The authority index is calculated by weighting three parameters: deviation value, network service type weight, and current load urgency, and outputs a normalized value between 0 and 1. The index generation process includes initializing the weight coefficient, loading the subnet service priority configuration table, and performing multi-source data fusion operations. The optimization effect is evaluated based on the frequency control authority index. The authority index is positively correlated with the transmission quality optimization effect. The higher the index value, the more significant the probe model's effect on improving network performance. This process involves defining a coefficient update mechanism for the data transmission balance equation, which automatically triggers equation parameter recalibration when the external interference intensity changes exceed a threshold. The dynamic weight adjustment of external interference factors uses a sliding window algorithm to update the weight distribution ratio based on historical interference impact records.

[0022] When constructing the network transmission quality model, the inter-subnet node correlation is considered, logical subnet boundary nodes with data dependencies are marked, and cross-subnet transmission compensation parameters are added to the data transmission balance equation. The model covers the entire set of performance nodes, establishes an independent status monitoring unit for each node, and uploads node traffic load and bit error rate data in real time. When evaluating optimization results, resource allocation and adjustment priorities are adjusted based on the frequency control authority index, with regions with high authority indexes receiving preferential probe resource allocation. Changes in network transmission stability are quantified, extracting indicators such as transmission path jitter rate, retransmission request frequency, and bandwidth utilization fluctuation amplitude to establish a stability change baseline. During the model integration phase, bandwidth carrying capacity measurements are synchronously calibrated, and bandwidth stress testing is periodically performed to refresh node maximum throughput records. The calibration operation is linked to the analysis results of external interference factors, and a bandwidth retest mechanism is triggered when new interference types are added.

[0023] After the frequency control authority index is generated, deviation threshold detection is performed and two-level deviation alarm thresholds are set. The primary threshold triggers the model parameter fine-tuning instruction, and the advanced threshold activates the topology reconstruction process. The need for re-optimization of model parameters is managed by configuring the version number, marking the set of nodes to be optimized and pushing it to the network management platform. The solution of the data transmission balance equation is compared with the actual state of the network. When three consecutive abnormal deviations occur, the model output is frozen and the manual intervention verification process is initiated. After the logical subnet division results are implemented, the traffic balance between subnets is continuously monitored. When the cross-subnet traffic ratio exceeds the design threshold, the subnet boundary reconstruction is triggered. The dynamic maintenance of the performance node set includes the new node admission detection and the automatic removal of failed nodes. The node status changes are synchronized to the data transmission balance equation variable set in real time.

[0024] Example 2: See Figure 3, identifying the data transmission coupling effect between each performance node. This process analyzes network routing protocols and traffic engineering strategies to map the data flow dependencies between nodes. Coupling effect analysis includes three association modes: hard coupling caused by direct physical connections, soft coupling formed by shared transmission links, and logical coupling caused by cross-device business logic dependencies. The actual forwarding data packet sequence between nodes is collected, and the source and destination node pairing frequency and path overlap parameters are extracted. Each coupling mode is quantified. Hard coupling records the physical port rate matching degree, soft coupling measures the shared buffer contention ratio, and logical coupling analyzes the application layer session binding strength. The output is a correlation strength value matrix, with rows and columns representing performance node indices. The values ​​of the matrix elements reflect the interaction frequency and impact depth between nodes.

[0025] Quantify the impact of external interference factors on each performance node. First, establish a classification system for external interference factors, including four categories: environmental electromagnetic interference, equipment heat dissipation fluctuations, concurrent protocol conflicts, and management command interruptions. Design an intensity measurement plan for each type of interference source: for environmental electromagnetic interference, use a field strength probe to capture the radiation value of a specific frequency band; for heat dissipation fluctuations, record the correlation curve between temperature sensor data and fan speed; for protocol conflicts, detect the number of ARP broadcast storm triggers; and for management command interruptions, analyze the CPU cycles occupied by SNMP polling. Normalize the original measurement value to an impact intensity coefficient in the range of 0-1, where 0 represents no impact and 1 represents the maximum interference state. The time attenuation factor is introduced into the intensity coefficient calculation, and recent interference events are given a higher weight.

[0026] The data transmission balance equation is constructed by combining the bandwidth carrying capacity of performance nodes, the data transmission coupling effect, and the impact of external interference factors. The bandwidth carrying capacity serves as the basic input vector, with the elements corresponding to the measured maximum throughput of the node. Vector normalization is performed before construction to eliminate dimensional differences between nodes. The coupling effect matrix is ​​multiplied by the bandwidth vector to generate the node load transfer result, reflecting the additional pressure on each node to carry the forwarding load of other nodes. The impact of external interference factors is expressed as a diagonal matrix. The main diagonal elements are the independent interference coefficients of each node. These are dot-producted with the load transfer result to produce the interference-corrected load distribution.

[0027] The equation construction process forms a three-layer stacked structure: the first layer is the bandwidth base capacity layer, the second layer is the coupling diffusion layer, and the third layer is the interference attenuation layer. The output vector of the structure is the predicted value of the data transmission equilibrium state, and the vector dimension is consistent with the number of performance nodes. During the numerical initialization phase, the measured bandwidth baseline value is loaded, the initial elements of the coupling matrix are set to equal weight coefficients, and the interference coefficient is preset based on historical averages. Data processing performs step-by-step operations, first calculating the product of the bandwidth vector and the coupling matrix to generate an intermediate variable. This intermediate variable is then multiplied element-by-element by the interference diagonal matrix to output the final equilibrium state prediction value. After each operation, the output vector boundaries are checked to limit the value range to within the node's physical capabilities.

[0028] The construction of the transmission balance equation includes defining boundary constraints. A minimum available bandwidth threshold is set for each performance node, which is dynamically adjusted according to the key business traffic requirements carried by the node. A maximum interference tolerance upper limit is set, and interference conditions exceeding the upper limit trigger an alarm mechanism. Boundary conditions are used as hard constraints when solving the equation and are enforced through the Lagrange multiplier method. During the construction process, actual node traffic data is continuously collected and compared with the equation prediction value on a rolling basis. When the deviation exceeds the allowable range, the parameter optimization mechanism is activated: the coupling matrix weight is dynamically refreshed according to the changes in the measured traffic path, the interference coefficient is recalibrated based on real-time monitoring data, and the bandwidth base value updates the stress test results according to the preset period.

[0029] When iteratively solving the equations, the coupling effect weights are adjusted based on the actual data flow efficiency between nodes. When the forwarding delay of a node to other nodes continues to increase, the corresponding coupling influence coefficient is reduced. When a node has redundant path backups, its coupling relationship is split into multiple alternative paths. The adaptive adjustment amplitude of the weights is proportional to the rate of change of traffic, and a maximum adjustment step size limit is set to prevent oscillation. The quantified results of external interference factors are directly connected to the real-time alarm flow of the network management system. If the interference intensity suddenly changes and exceeds the preset gradient threshold, the equation solution process is immediately frozen and the calculation is resumed after the interference event analysis report is received.

[0030] After the equation is generated, a simulated traffic stress test is run on an independent verification platform. A virtual topology mirror environment is constructed and multiple sets of mixed service flow data packets are injected. The status data of each node in the simulation environment is recorded, including the cache queue length, transmission delay distribution, and the location of packet loss events. The recorded data is then compared with the equation prediction output for trend matching analysis to verify the equation's prediction accuracy. For nodes with high prediction deviations, the coupling relationship identification process and interference quantification data are traced back to perform positioning and repair operations. The equation parameters that pass the verification are solidified into a configuration file and synchronized to all related network management components. In subsequent operation cycles, the configuration version number is bound to the network change record to support historical status backtracking analysis. The solution of the data transmission balance equation serves as the decision input for the deployment location of the network monitoring probe model and also provides load distribution benchmark parameters for sampling frequency control.

[0031] Example 3: See Figure 4The performance test operation of the network monitoring probe model is completed by simulating the network environment under different load conditions. The load conditions are divided into three types: high load, low load and fluctuating load, and each type corresponds to traffic characteristics. The high load state simulates the situation where the network traffic is close to or reaches the capacity of the physical link, the low load state reflects the idle or lightly loaded operation state of the network, and the fluctuating load state presents periodic or random traffic changes. Configure the traffic generator before the test, and set the benchmark parameters including packet size distribution, sending interval and protocol type ratio. The traffic generator is connected to the monitoring interface of the probe model under test to form a closed-loop test environment.

[0032] During high-load testing, a traffic generator sends 90% to 95% of the maximum allowed data volume at a constant rate, long enough for the network to reach a stable state. The timestamp sequence of the packet captured by the probe model is recorded, and the statistical distribution of end-to-end latency is calculated. Latency detection capability is characterized by the following metrics: average latency reflects overall response speed, latency standard deviation demonstrates detection stability, and 99th percentile latency indicates performance limits under extreme conditions. The packet loss detection accuracy test uses active tagging, adding detection tags to a specific percentage of packets at the sending end. The receiving end then calculates the success rate and false positive rate of tag matches.

[0033] The low-load test reduces the traffic generator's transmission rate to below 10% of the link capacity, maintaining a test duration no less than the high-load test period. During this test, the probe model's basic processing latency and idle resource usage are observed. The probe model's minimum detection latency under no-contention conditions is recorded as a performance benchmark. The packet loss detection accuracy test injects a controlled number of abnormal packet loss events under low-load conditions to verify the probe's sensitivity to sparse packet loss.

[0034] The fluctuating load test utilizes a composite traffic pattern, maintaining a base traffic flow of 30% to 50% of the link capacity, superimposed with periodic burst traffic pulses. The pulse amplitude varies between 20% and 80% of the link capacity, with pulse widths ranging from milliseconds to seconds. Pulse parameters are dynamically adjusted during the test to cover scenarios with varying frequency and amplitude combinations. The latency response characteristics are characterized by a latency curve within the pulse period, focusing on the response lag between rising and falling edges. The packet loss detection accuracy test injects packet loss events during the pulse peak period to analyze the probe's detection consistency under varying load change rates.

[0035] The impact weight of each performance node in the network transmission quality model is determined through a multi-dimensional assessment. Evaluation parameters include the proportion of traffic handled by the node, the business criticality of the connected device, historical fault records, and the importance of the topological location. A judgment matrix is ​​constructed using the Analytic Hierarchy Process (AHP), and weights are calculated through eigenvector calculation. A consistency check is incorporated into the weight calculation process, and parameter scaling is rescaled when the consistency ratio of the judgment matrix exceeds a threshold.

[0036] An iterative screening process is used to identify the target performance nodes with the greatest impact. The first round of screening retains nodes with weights exceeding two standard deviations above the average weight. A second round compares these nodes' traffic handling capabilities and fault impact, ultimately determining the highest-weighted node. The logical subnets corresponding to the target nodes are automatically associated based on the network topology, with subnet boundaries defined by routing policies and VLAN divisions.

[0037] The network monitoring probe model is deployed in the logical subnet corresponding to the target performance node, involving two levels: physical connection and logical configuration. The physical connection ensures that the probe's monitoring port is connected to the core switch mirror port or splitter of the target subnet to obtain complete data traffic. The logical configuration includes setting filtering rules to focus on the target node-related traffic, adjusting the sampling rate to match the subnet traffic characteristics, and configuring the reporting path to avoid the key forwarding links of the target node. After the deployment is complete, a baseline test is performed to verify that the probe monitoring range covers all key interfaces of the target node. The delay detection capability of the probe model is quantified using the following formula: in, represents the mean square error of delay detection, N is the total number of test samples, is the delay value of the i-th data packet recorded by the probe, The actual latency value measured by the benchmark test system. This formula is specifically used for evaluating latency detection accuracy in this embodiment and has no sign overlap with the mathematical expressions in other embodiments. The calculated result is used to calibrate the probe's internal clock compensation parameters, reducing systematic latency measurement deviations.

[0038] Packet loss rate detection accuracy is assessed using the receiver verification method. The transmitter records complete packet sequence numbers, while the receiver calculates the percentage of missing sequence numbers as the true packet loss rate and compares this to the packet loss rate reported by the probe on a cycle-by-cycle basis. Accuracy metrics include the missed negative rate (missing negatives) and the false positive rate (false positives). The missed negative rate reflects the percentage of actual packet loss that the probe fails to detect, while the false positive rate measures the percentage of normal packets that the probe mistakenly identifies as lost. During testing, the network packet loss rate is dynamically adjusted, covering varying intensity ranges from 0.1% to 10%.

[0039] Monitoring data from target performance nodes is transmitted to the analysis system via independent channels, avoiding additional load on the existing network. Data transmission utilizes compression and differential encoding technologies to reduce monitoring overhead. The analysis system integrates latency and packet loss data reported by the probes and combines it with network topology information to generate a quality heat map, visually displaying the performance status of the target node and its associated paths. The heat map's update frequency is synchronized with the probe sampling rate, triggering real-time alerts when changes in key parameters exceed thresholds.

[0040] During long-term operation on the target subnet, the probe model continuously adapts to changes in network traffic patterns. This adaptive mechanism includes dynamically adjusting sampling strategies, optimizing detection algorithm parameters, and learning a baseline for normal latency. When a network upgrade or configuration change is detected, a calibration process is automatically triggered to re-establish the performance baseline. The effectiveness of the probe deployment is indirectly assessed by comparing the fault recovery time and performance fluctuations of the target nodes before and after deployment. The probe's own resource consumption is also monitored to ensure it remains within the designed range.

[0041] Traffic characteristic analysis of the logical subnet provides a basis for probe model configuration. Deep packet inspection technology is used to identify the main application protocols within the subnet and their timing characteristics, and differentiated detection sensitivities are set for different protocol types. Fine-grained latency monitoring is used for real-time voice and video traffic, while packet loss statistics are focused on bulk data transmission. Protocol analysis results are also used to optimize probe filtering rules, reducing processing overhead for irrelevant traffic. Periodic reanalysis of subnet traffic characteristics ensures that probe configuration is synchronized with actual network usage patterns.

[0042] The performance data of target nodes continuously interacts with the network transmission quality model. Real-time metrics collected by the probes are fed into the model for status prediction, and the model's output guides the probes in adjusting their monitoring priorities. When the model predicts performance degradation on a particular path, the probes increase monitoring density for that path. When the model detects an abnormal pattern, the probes initiate detailed diagnostic data collection. This two-way interaction forms a closed-loop optimization system, focusing monitoring resources on the areas of the network that require the most attention. The frequency of data exchange is strictly limited during this interaction to prevent the monitoring activity itself from becoming a source of network load.

[0043] Example 4: See Figure 5 , within the preset time window, the data traffic generated by each performance node in the communication network is collected through distributed counters. Each node deploys a lightweight traffic statistics agent to record the number and number of bytes of inbound and outbound data packets at a fixed time granularity. The time window is divided using a sliding window mechanism, and the window length is dynamically adjusted according to the network scale, and is typically set to range from 5 minutes to 1 hour. The traffic statistics agent summarizes the data at the end of the window and transmits it to the aggregation node through a compression protocol. The aggregation node performs deduplication verification to eliminate duplicate reports caused by network delays, and merges the traffic data of all performance nodes to generate the total data traffic load. The load value contains two dimensions: the total number of data packets and the total number of bytes, which are used for different types of monitoring load calculations.

[0044] The basic monitoring overhead of the network monitoring probe model is determined through offline calibration experiments. This calibration process is performed in an isolated test environment to eliminate the impact of network fluctuations. The probe model's CPU utilization, memory consumption, and network throughput are measured under zero load. These inherent overheads constitute the basic monitoring overhead. During actual runtime, the basic monitoring overhead must also be adjusted for network environment factors, including dynamic parameters such as protocol parsing complexity and encrypted traffic processing overhead. This overhead data is stored in a configuration file for direct access in different scenarios.

[0045] The calculation of the theoretical monitoring load combines the total data traffic load and the delay detection capability. The delay detection capability parameter reflects the efficiency of the probe model in processing unit traffic, which has been obtained in the performance test of Example 3. The calculation process decomposes the total data traffic load by protocol type, and the different types are matched with corresponding processing coefficients. For example, TCP traffic is weighted by the number of connections, and UDP traffic is weighted by the packet rate. The weighted traffic value is added to the basic monitoring overhead and then multiplied by the delay detection capability factor to output the theoretical monitoring load value. The load value is continuously updated according to the time window to form a load change curve.

[0046] Determining the optimal sampling frequency requires balancing monitoring accuracy and resource consumption. Using the theoretical monitoring load as input, a matching interval is searched in a preset load-frequency mapping table. The mapping table is predefined based on the probe model and contains discrete load intervals and corresponding recommended sampling frequencies. When the theoretical monitoring load falls at the intersection of two intervals, linear interpolation is used to calculate the intermediate value. After determining the initial sampling frequency, the packet loss rate detection accuracy constraint is superimposed. The higher the accuracy requirement, the greater the upward adjustment of the sampling frequency, but it does not exceed the maximum sampling rate supported by the probe hardware. The final optimal sampling frequency is the maximum value that satisfies all constraints.

[0047] Data sampling frequency control instructions are generated using a structured data format. These instructions contain fields such as the time window number, target sampling frequency, effective timestamp, and checksum. The instruction queue is arranged in order of time windows, with each window corresponding to a separate instruction. When switching windows, new control instructions are sent to the probe model via a secure channel. The probe's configuration management module parses and executes the frequency switch. The instruction transmission process utilizes a transaction mechanism, with automatic retransmission if no confirmation response is received, ensuring reliable completion of configuration changes.

[0048] Table 1: Aggregation of performance node traffic data within a typical time window.

[0049] The table data is derived from an anonymized sample of real-world network environments and displays traffic statistics for three performance nodes within a 15-minute window. The number of inbound and outbound packets reflects the traffic volume processed by the node, the number of bytes used to calculate bandwidth usage, and the distribution of protocol types influences probe processing strategies. The summary row displays the total data traffic load for the window, serving as an input baseline for subsequent calculations.

[0050] The probe model's dynamic frequency modulation mechanism enables adaptive resource allocation. When the theoretical monitoring load indicates that a higher sampling frequency is required for a certain time window, the probe automatically allocates more computing resources to the data collection module, potentially temporarily reducing resource quotas for non-critical functions. The frequency modulation process is smooth, avoiding data discontinuities caused by sudden changes in the sampling interval. A snapshot of the current sampling state is saved during frequency switching, ensuring that business monitoring is not disrupted by configuration changes. Under extremely high load conditions, the probe enters a degraded mode to prioritize the collection integrity of core metrics.

[0051] The time window division strategy affects monitoring timeliness and resource consumption. Shorter windows reflect network status changes more quickly but increase system overhead; longer windows conserve resources but may miss important events. Actual deployments employ a multi-level time window system: the basic level uses fixed-length windows to ensure routine monitoring, and the event-triggered level uses variable-length windows to address emergencies. Window switching is aligned with the statistical cycle of network devices to avoid data inconsistencies caused by time skew.

[0052] Protocol type distribution data guides sampling strategy optimization. TCP traffic sampling focuses on tracking connection establishment and closing processes, retransmission events, and window size changes; UDP traffic monitoring focuses on identifying packet interval jitter and loss patterns. These differentiated sampling requirements are implemented through additional parameters in the control instructions, enabling the probe to adopt the most effective monitoring method for different protocols. Changes in protocol distribution exceeding a threshold trigger sampling strategy recalculation to ensure that the monitoring method is aligned with current traffic characteristics.

[0053] Analyzing the discrepancies between theoretical monitored load and actual operational load allows for continuous optimization of parameter settings. A larger load margin is allowed during initial deployment, and as operational data accumulates, the error range for load forecasting is gradually tightened. The load calculation model is regularly retrained with the latest network data to adapt to changes in network architecture and evolving business models. A library of load patterns developed over long periods of operation supports predictive resource allocation, enabling preemptive adjustments to probe configurations in anticipation of expected high-load windows.

[0054] The effectiveness of sampling frequency control instructions is evaluated through a closed-loop feedback mechanism. After each frequency switch, the probe reports actual resource utilization and data collection quality indicators. The control center compares the deviation between the expected and actual results and adjusts the parameters for generating subsequent instructions. This adaptive adjustment gradually approaches the optimal operating state of the system, reducing the need for manual intervention. Feedback data is also used to construct a performance profile of the probe and identify the optimal operating mode under specific load conditions.

[0055] A dynamic balance between network status changes and sampling frequency is maintained through an event-driven mechanism. When network topology changes, traffic patterns mutate, or equipment failures occur, the event processor immediately adjusts the remaining duration of the current time window and recalculates the theoretical monitoring load. Major events directly trigger emergency sampling mode, bypassing conventional frequency limits to obtain detailed diagnostic data. Event processing priority is tied to business criticality, ensuring monitoring continuity for critical services regardless of regular scheduling.

[0056] The probe model implements layered monitoring of resource usage. The basic resource layer tracks hardware metrics such as CPU and memory to ensure that frequency adjustments do not cause system overload. The business indicator layer monitors the integrity and timeliness of data collection, verifying that the sampling frequency meets monitoring requirements. The network impact layer assesses the impact of probe behavior on the measured network to prevent monitored traffic from becoming a new source of interference. These three layers of monitoring data collectively guide the refined adjustment of the sampling frequency.

[0057] Metadata management of time windows supports monitoring, backtracking, and analysis. Each window records complete configuration parameters, network status, and execution results, forming a traceable operation log. Historical data supports multi-dimensional queries by time range, node group, or service type, facilitating analysis of long-term network performance trends. Log storage utilizes a ring buffer structure, balancing storage overhead with historical depth requirements. Key window data is permanently archived for in-depth root cause analysis.

[0058] Versioning of sampling frequency control strategies ensures smooth upgrades. When introducing a new frequency algorithm or optimizing parameters, the old and new strategy versions are run in parallel to compare their effectiveness. Once the new strategy is confirmed to be more stable than the old one, its application is gradually expanded. A version rollback mechanism ensures rapid recovery to a known good working state in the event of an anomaly. Strategy version information is embedded in control instructions, ensuring that the probe model always uses the appropriate processing logic.

[0059] Spatiotemporal correlation analysis of node traffic data enhances load forecasting accuracy. This approach not only considers absolute traffic values ​​within a single window but also analyzes correlation features such as traffic shift patterns between adjacent nodes and historical traffic patterns over the same time period. These results refine the base load calculation to more accurately reflect actual monitoring needs. The spatiotemporal analysis algorithm utilizes sliding window technology to balance computational complexity and forecast timeliness.

[0060] Distributed collaboration within the probe model improves the efficiency of large-scale network monitoring. In multi-probe deployments, control command generation considers load balancing between probes to prevent overloading some nodes while leaving others idle. Probes share network status data and coordinate the adjustment of key sampling areas. The collaborative mechanism utilizes a hierarchical control architecture, combining autonomous local decision-making with global coordinated optimization.

[0061] Example 5: The data throughput of the network monitoring probe model at a unit sampling frequency is obtained through a standardized test process. The test environment is configured as a dedicated isolated network to avoid external traffic interference. The test uses a set of data packet samples of uniform specifications, including typical data packets of different sizes and protocol types. During the test, the sampling frequency is fixed as the benchmark value, and the number of data packets successfully processed by the probe model per unit time is recorded, while the computing resource consumption during the processing process is monitored. The throughput test is performed continuously for multiple rounds, and the data packet distribution ratio is changed in each round to output the average throughput and resource consumption range. The test results form a benchmark parameter table, which associates the corresponding relationship between the sampling frequency and the processing power.

[0062] The frequency switching time plan for the data acquisition module is generated based on the optimal sampling frequency and throughput test data. The plan content contains a list of frequency sequence change points within the complete time window, and each change point specifies the target frequency value and the switching timestamp. The plan generation algorithm first calculates the difference between the optimal sampling frequency and the current frequency, and combines the time cost required for a unit frequency change in the throughput test to derive the safe switching time range. Under the principle of uniform distribution of switching points, the entire preset time window is divided into discrete time periods of equal length, and the switching operation is scheduled to be performed at the dividing point of adjacent time periods. The number of time periods is set to take into account the stability requirements of frequency conversion. Too dense time periods increase system overhead, while too sparse time periods reduce the adjustment accuracy.

[0063] The division of the preset time window into discrete time segments is strictly aligned with the system clock cycle. The start time of the time segment is calibrated according to the time synchronization protocol of the network equipment, ensuring that all nodes use the same time slice count basis. The length of a single time segment is determined by dividing the total window duration by the number of planned switches. The lower limit of the length is limited by the probe hardware state retention period and the upper limit is constrained by the rate of change of network state. Each time segment is assigned a unique number, which is associated with the corresponding frequency control parameter set. The time window segmentation results are broadcast to relevant network components to maintain global consistency of monitoring timing.

[0064] After a single discrete time period, the data acquisition module's actual data throughput is verified. This verification operation reads the total number of processed packets recorded by the probe's hardware counters and compares it to the theoretically expected throughput for that time period. The expected value is calculated based on the sampling frequency and duration in effect for that time period. The deviation between the actual throughput and the expected value is statistically analyzed, including both the absolute difference and the relative difference ratio. The verification process also monitors the probe's internal resource status, including cache queue depth, peak memory usage, and processor interrupt frequency. A verification report is automatically generated, noting the severity of any deviations.

[0065] Based on the verification results, compensation corrections are made to the frequency switching time plan. When the deviation is below the tolerance threshold, the original plan for the subsequent time period is maintained; when the deviation exceeds the threshold, the compensation algorithm is activated. The compensation algorithm calculates the impact of the current deviation on the overall monitoring target and derives the required frequency compensation coefficient. This coefficient is applied to the planned frequency value for the subsequent time period. Positive deviations correct the frequency downward, while negative deviations correct the frequency upward. The correction range is limited to the maximum and minimum frequency boundaries supported by the probe. A cumulative compensation strategy is used for continuous deviations in multiple time periods. The correction operation updates the list of demarcation points for the subsequent time period and generates a new switching time sequence.

[0066] The frequency switching schedule dynamically evolves throughout the preset time window. At the end of each time slice, the plan parameters for the remaining time slots are iteratively adjusted based on the latest validation results. This adjustment process maintains the overall planned load and only redistributes the frequency distribution across time slots. The plan version number is incremented after compensation and the change record is stored in association with the time window number. Historical data on plan evolution is used to analyze the accuracy of the network status prediction model and support subsequent algorithm optimization.

[0067] The data acquisition module's operating frequency updates strictly adhere to the switching schedule. Upon reaching the scheduled switching point, the probe executes a pre-defined state transition process: pausing the current sampling thread → saving the sampling context → loading the new frequency parameters → initializing hardware registers → restarting the sampling thread. The switching process is completed within milliseconds, and any packet loss during the process is factored into quality control metrics. In high-frequency switching scenarios, an incremental configuration update strategy is employed, modifying only the differential parameters to reduce switching time.

[0068] Verification of actual data throughput utilizes multiple auxiliary metrics. In addition to core throughput values, analysis includes packet processing latency distribution, abnormal packet ratio, and protocol parsing error rate. These metrics help identify the source of throughput deviations: insufficient system processing capacity leading to overall degradation, or specific protocol processing defects causing localized performance degradation. These multi-dimensional metrics assist in locating the root cause of the problem and guide the development of targeted remediation strategies. Throughput verification data is integrated with the probe log system, supporting retrospective analysis over any time period.

[0069] The time slice expiration trigger mechanism utilizes two conditions. The primary condition is the expiration of the preset duration, implemented using a high-precision timer; the secondary condition is a sudden network event signal, such as a traffic surge alarm or device failure notification. The secondary condition takes precedence, immediately interrupting the current time slice and entering the verification process. This design ensures a rapid response to significant network changes, avoiding the need to wait for the time slice to expire.

[0070] Compensation correction strategies are differentiated based on the type of deviation. Systematic deviations are addressed using a linear compensation coefficient, random deviations are addressed using a smoothing filter algorithm, and persistent deviations trigger a plan reconstruction process. Correction parameters are normalized to eliminate the impact of network scale differences. The final determination of compensation values ​​is based on historical correction effect data to avoid oscillation caused by overcorrection. The correction strategy's decision logic incorporates a built-in anti-error mechanism that prohibits frequency adjustments outside of a safe range.

[0071] Discrete time slices are managed through a state machine. Each time slice consists of four phases: preparation, operation, verification, and transition. The preparation phase loads configuration parameters and initializes the hardware; the operation phase performs routine data collection; the verification phase performs throughput comparisons; and the transition phase completes compensation calculations and schedule updates. State transitions are coordinated by a central scheduler, and the time spent in each phase is recorded for performance optimization. The state machine ensures that the system maintains deterministic behavior even when complex compensation corrections are required.

[0072] The closed-loop control of the frequency switching schedule relies on feedback loop strength adjustment. The feedback coefficient is dynamically adjusted based on network stability: in stable environments, the feedback strength is reduced to prevent sensitive oscillations; in turbulent environments, the feedback strength is increased to accelerate response. The feedback strength parameter is linked to the standard deviation of the verification results for each time slice, and the coefficient is automatically reduced when low deviation is observed for multiple consecutive time slices. The feedback adjustment range has upper and lower limits to prevent the system from entering a state of positive feedback runaway.

[0073] The entire dynamic adjustment process generates an operational audit log. This log contains the switchover time for each time slice, the set frequency value, the actual throughput, the deviation value, and compensation correction details. Audit data is stored in a tamper-proof format, supporting third-party verification. Long-term audit data forms a frequency optimization knowledge base, refining the optimal control mode for typical network scenarios. Audit log generation triggers an integrity check, and any data anomalies trigger an alarm and freeze the adjustment operation.

[0074] After the time window completes all discrete time period adjustments, it generates an overall execution report. This report summarizes key metrics within the window, such as average frequency, total throughput achieved, and the number of compensation actions. The report also includes an analysis of deviations from theoretical expectations, noting the time periods where major deviations occurred and categorizing the root causes. The final report is uploaded to the network management center and serves as a key input for evaluating system operational status. When a new time window is launched, all parameters are initialized, inheriting the optimized configuration from the previous window as the default baseline.

[0075] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0076] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A communication network performance monitoring and optimization method, characterized in that: include: Collecting performance nodes that affect transmission quality in the communication network, and building a network transmission quality model based on the performance nodes; Configuring a network monitoring probe model, performing a performance test operation on the network monitoring probe model, and obtaining the delay detection capability and packet loss rate detection accuracy of the network monitoring probe model; Integrating the network monitoring probe model into the network transmission quality model; Calculate the total data traffic load of the communication network within a preset time window, and calculate the theoretical monitoring load of the network monitoring probe model within the preset time window in combination with the delay detection capability of the network monitoring probe model; Based on the theoretical monitoring load and packet loss rate detection accuracy, generating a data sampling frequency control instruction for the network monitoring probe model; Dynamically adjust the operating frequency of the data acquisition module of the network monitoring probe model according to the data sampling frequency control instruction; Evaluate the transmission quality optimization effect of the network monitoring probe model on the communication network.

2. The communication network performance monitoring and optimization method according to claim 1, characterized in that: The collecting performance nodes that affect transmission quality in the communication network and constructing a network transmission quality model based on the performance nodes includes: Extracting topological structure information of a communication network and dividing the communication network into a plurality of logical subnets; Identify key performance nodes in each logical subnet and generate a set of performance nodes for the communication network; Measuring the bandwidth carrying capacity of each performance node and analyzing external interference factors acting on the performance node; Based on the bandwidth carrying capacity and external interference factors, a data transmission balance equation of the communication network is established.

3. The communication network performance monitoring and optimization method according to claim 2, characterized in that: The data transmission balance equation of the communication network established based on the bandwidth carrying capacity and external interference factors includes: Analyze the data transmission coupling effect between each performance node and quantify the impact of the external interference factors on each performance node; The data transmission balance equation is constructed by combining the bandwidth carrying capacity of the performance node, the data transmission coupling effect, and the influence intensity of external interference factors.

4. The communication network performance monitoring and optimization method according to claim 1, wherein: The configuring of the network monitoring probe model and performing the performance test operation of the network monitoring probe model include: The delay response characteristics and packet loss detection accuracy of the network monitoring probe model under high load, low load and fluctuating load conditions are tested respectively to obtain the delay detection capability and packet loss rate detection accuracy.

5. The communication network performance monitoring and optimization method according to claim 1, wherein: Integrating the network monitoring probe model into the network transmission quality model includes: Analyze the impact weight of each performance node in the network transmission quality model and locate the target performance node with the largest impact weight; The network monitoring probe model is deployed in the logical subnet corresponding to the target performance node.

6. The communication network performance monitoring and optimization method according to claim 1, wherein: The calculating the total data traffic load of the communication network within a preset time window includes: Aggregate the data traffic generated by each performance node within the preset time window to obtain the total data traffic load; Calculating the theoretical monitoring load of the network monitoring probe model includes: Obtaining the basic monitoring overhead of the network monitoring probe model within a preset time window, and calculating the theoretical monitoring load in combination with the total data traffic load; Generate data sampling frequency control instructions including: Based on the theoretical monitoring load and packet loss rate detection accuracy, the optimal sampling frequency of the network monitoring probe model within a preset time window is determined.

7. The communication network performance monitoring and optimization method according to claim 6, characterized in that: The generating of data sampling frequency control instructions comprises: The data throughput of the network monitoring probe model at a unit sampling frequency is tested, and a frequency switching time plan of the data acquisition module is generated in combination with the optimal sampling frequency.

8. The communication network performance monitoring and optimization method according to claim 7, characterized in that: The dynamically adjusting the operating frequency of the data acquisition module of the network monitoring probe model includes: Dividing the preset time window into discrete time periods, and evenly allocating the frequency switching time plan to each discrete time period; After a single discrete time period ends, verifying whether the actual data throughput of the data acquisition module meets expectations; The frequency switching time plan of subsequent discrete time periods is compensated and corrected according to the verification results until the adjustment operation of the entire preset time window is completed.

9. The communication network performance monitoring and optimization method according to claim 1, wherein: The evaluating the transmission quality optimization effect of the network monitoring probe model on the communication network includes: Obtain the change in delay and packet loss rate of the communication network before and after optimization; Calculating an actual monitoring load of the communication network based on the delay variation and the packet loss rate variation; Comparing the deviation between the actual monitoring load and the theoretical monitoring load to generate a frequency control authority index of the data acquisition module; The network transmission quality optimization effect is evaluated according to the frequency control authority index.

10. A communication network performance monitoring and optimization system, used to implement the communication network performance monitoring and optimization method according to any one of claims 1 to 9, characterized in that: The system comprises: A transmission quality modeling module is used to collect performance nodes that affect transmission quality in the communication network and build a network transmission quality model based on the performance nodes; The probe performance test module is used to configure the network monitoring probe model, perform performance test operations on the network monitoring probe model, and obtain the delay detection capability and packet loss rate detection accuracy; A model fusion module, used to integrate the network monitoring probe model into the network transmission quality model; A monitoring load calculation module, configured to calculate the total data traffic load of the communication network within a preset time window and calculate the theoretical monitoring load in combination with the delay detection capability; A sampling control module, configured to generate a data sampling frequency control instruction for the network monitoring probe model based on the theoretical monitoring load and packet loss rate detection accuracy; A dynamic adjustment module, configured to dynamically adjust the operating frequency of the data acquisition module according to the data sampling frequency control instruction; The optimization evaluation module is used to evaluate the transmission quality optimization effect of the network monitoring probe model on the communication network.

Images