Network quality diagnosis method and device

Abstract

According to the network quality diagnosis method and device provided by the invention, perception comprehensiveness is remarkably improved through multi-dimensional data collaborative acquisition, key levels such as terminal states, AP loads, neighbor interference, equipment parameters and network configuration are covered, and the limitation that only isolated indexes such as bandwidth and delay are monitored in a traditional scheme is broken through; on the basis, a machine learning algorithm is fused, accurate root cause positioning of complex network faults is achieved, and the diagnosis accuracy and the automation level are greatly improved; furthermore, according to a real-time stream processing architecture, an event triggering type response mechanism is established, sudden anomalies can be recognized, and the timeliness defect of a periodic polling scheme is overcome; then a closed-loop optimization system is constructed, through linkage of a diagnosis result and an execution strategy, generation of user operable suggestions is supported, and a complete closed loop of'detection-analysis-optimization 'is formed; and finally, various home network environments are compatible by virtue of modular design, and the system can be quickly deployed on an operator platform or an intelligent routing system, so that low-cost large-scale application is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication networks, in particular to a network quality diagnosis method and device. BACKGROUND

[0002] With the popularization of smart home devices and the normalization of remote office and online education, the home network has evolved from a single Internet access channel to a key infrastructure supporting smart security, 4K / 8K ultra-high-definition video, cloud gaming, and VR / AR and other diversified services. Users have extremely high requirements for the stability, low latency, and high throughput capacity of home networks, and network quality directly determines the quality of digital life experience.

[0003] Current home network quality evaluation mainly relies on operator broadband speed test platforms or terminal built-in diagnostic tools, and its technical path is concentrated in two typical methods of passive polling detection and active probe measurement. Passive polling periodically collects basic indicators such as router CPU load and port traffic through SNMP protocol, while active probe relies on ICMP Ping or HTTP download test to evaluate end-to-end delay and bandwidth. Although these solutions can obtain some network performance parameters, their perspective is limited to isolated indicators at the device layer or transmission layer, neither correlating the differences in wireless communication capabilities of terminal devices nor integrating the characteristics of business needs in application scenarios, nor lacking collaborative analysis of hidden factors such as Wi-Fi environmental interference, resulting in a serious disconnection between evaluation results and actual user experience.

[0004] The existing solutions have systemic defects due to narrow data collection dimensions and rigid diagnostic mechanisms: at the perception level, traditional tools can only capture surface indicators such as bandwidth and delay, and cannot diagnose deep problems such as Wi-Fi co-channel interference, neighbor AP competition, or terminal protocol compatibility. For example, when an old terminal limits the overall network throughput because it only supports the 802.11n protocol, the system only feedbacks "insufficient bandwidth" and cannot locate the root cause; at the response level, the fixed periodic polling mechanism cannot capture transient faults, making problem backtracking rely on user complaints and manual log grabbing; at the disposal level, the diagnostic results are presented in abstract numerical values or general alarms, neither correlating the differences in business sensitivity of specific application scenarios nor generating executable optimization instructions, resulting in high user operation and maintenance costs. SUMMARY

[0005] In view of the above problems, a network quality diagnosis method and device are provided to overcome the above problems or at least partially solve the above problems, specifically: A network quality diagnosis method, comprising: Collecting network operation data, the network operation data including terminal connection information, Wi-Fi configuration information, a neighbor wireless access point list, WAN port connection state information, and local abnormal event information; uploading the network running data to a cloud platform and distributing the network running data to a message queue; identifying the network running data in the message queue by using a terminal identification subsystem and an application identification subsystem; performing real-time aggregation and feature extraction on the identified network running data by using a stream processing framework, inputting the extracted features into a machine learning model, and performing network quality diagnosis; in a case where the diagnosis result is network quality anomaly, storing the diagnosis result in a data warehouse, and generating an optimization suggestion according to the diagnosis result and displaying the optimization suggestion on a front-end platform.

[0006] Optionally, the terminal connection information includes a MAC address, signal strength, and connection state; the Wi-Fi configuration information includes a channel, frequency band, bandwidth, and security type; the neighbor wireless access point list includes RSSI, channel conflict, and BSSID quantity; WAN port connection state information includes IP state, jitter, and delay; and local abnormal event information includes flow interruption frequency and roaming events.

[0007] Optionally, collecting network running data includes: periodically collecting local network running data; and / or dynamically collecting local network running data according to diagnosis needs.

[0008] Optionally, in a case where the local network running data is periodically collected, a default collection period is set to 5 minutes.

[0009] Optionally, uploading the network running data to a cloud platform and distributing the network running data to a message queue includes: locally encoding and packaging the collected network running data, uploading the network running data to an access layer of the cloud platform by using MQTT and / or HTTPS protocols, uniformly receiving the network running data by a message middleware, and distributing the network running data to the message queue according to data types.

[0010] Optionally, identifying the network running data in the message queue by using a terminal identification subsystem and an application identification subsystem includes: inputting the network running data into the terminal identification subsystem, identifying a terminal MAC address, and extracting operating system and network capability information; and inputting the network running data into the application identification subsystem, and identifying active applications of the terminal by using the application identification subsystem based on DPI or port protocol analysis.

[0011] Optionally, the stream processing framework includes Flink.

[0012] Optionally, in a case where any one of the following conditions is met, the network quality is determined to be abnormal: a network experience score of a terminal is lower than a threshold value, a rule engine detects an abnormal event, or a lightweight machine learning model identifies an implicit problem.

[0013] Optionally, the abnormal event at least includes one of the following: weak signal area, frequent disconnection, channel conflict; the implicit problem at least includes one of the following: abnormal load of the wireless access point, existence of an interference source.

[0014] A network quality diagnosis device comprises: The plug-in collection module is configured to collect network operation data, which includes terminal connection information, Wi-Fi configuration information, a neighbor wireless access point list, WAN port connection state information, and local abnormal event information. The connection and transmission module is configured to upload the network operation data to a cloud platform and distribute it to a message queue. The data analysis and identification module is configured to identify the network operation data in the message queue by using a terminal identification subsystem and an application identification subsystem. The real-time calculation and diagnosis module is configured to use a stream processing framework to perform real-time aggregation and feature extraction on the identified network operation data, and input the extracted features into a machine learning model to perform network quality diagnosis. The data storage and display module is configured to store the diagnosis result in a data warehouse when the diagnosis result is a network quality abnormality, and generate an optimization suggestion according to the diagnosis result and display it on a front-end platform.

[0015] The embodiment of the present application significantly improves comprehensive perception through multi-dimensional data collaborative collection, covering key aspects such as terminal state, AP load, neighbor interference, device parameters, and network configuration, breaking through the limitations of traditional solutions that only monitor isolated indicators such as bandwidth and delay; on this basis, machine learning algorithms are integrated to achieve accurate root cause positioning of complex network faults, significantly improving diagnosis accuracy and automation level; further, according to the real-time stream processing architecture, an event-triggered response mechanism is established, which can identify sudden abnormalities and solve the timeliness defects of periodic polling solutions; further, a closed-loop optimization system is built, which supports the generation of user-operable suggestions through the linkage of diagnosis results and execution strategies, forming a complete closed loop of "detection-analysis-optimization"; finally, thanks to the modular design, it is compatible with various home network environments, and can be quickly deployed on operator platforms or intelligent routing systems, realizing low-cost large-scale application. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the present application, the following will briefly introduce the drawings needed to be used in the description of the present application. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0017] Figure 1is a network quality diagnosis method flowchart provided by an embodiment of the present application; Figure 2 is another network quality diagnosis method flowchart provided by an embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to make the above objectives, characteristics and advantages of the present application more apparent, comprehensible and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without making any creative efforts fall within the scope of protection of the present application.

[0019] With reference to Figure 1 , Figure 2 , an embodiment of the present application provides a network quality diagnosis method, which can specifically include the following steps. Step 101: Collecting network operation data, the network operation data including terminal connection information, Wi-Fi configuration information, neighbor wireless access point list, WAN port connection state information and local abnormal event information.

[0020] In an embodiment of the present application, in order to realize comprehensive perception and state monitoring of a home network environment, a plug-in collection module deployed and running on a user home gateway or AP (wireless access point) device can be used to continuously collect network operation data in multiple dimensions. Specifically, the plug-in collection module can collect information of all terminal devices connected to the home network through wired or wireless ways; obtain detailed configuration parameters of the current Wi-Fi network; actively scan and record the neighbor wireless access point list information that can be detected in the surrounding environment; and monitor the connection state information of the WAN port (wide area network interface) of the gateway device in real time. Meanwhile, the module also has the function of detecting various abnormal event information occurring locally in the gateway or AP. Through such distributed and fine-grained information collection at the edge device end, a data foundation is laid for subsequent network quality analysis, intelligent optimization and fault troubleshooting.

[0021] Among them, the terminal connection information can include a MAC address, signal strength and connection state; the Wi-Fi configuration information can include a channel, frequency band, bandwidth and security type; the neighbor wireless access point list can include RSSI, channel conflict and BSSID quantity; the WAN port connection state information can include IP state, jitter and delay; and the local abnormal event information can include flow breakage times and Roaming events.

[0022] It can be understood that the terminal connection information, Wi-Fi configuration, neighbor AP list, WAN port state and local abnormal event collected by the embodiment of the present application are used to build a full-factor coverage system of terminal-device-environment-external network-event: the root cause of user-side experience difference is located through the terminal connection state, the device layer setting defects such as channel conflict or frequency band mismatch are identified through the Wi-Fi configuration parameters, the spatial radio frequency interference map is drawn through the neighbor AP list to crack the invisible environmental problems such as same frequency competition, the operator side and the home network responsibility boundary are isolated through the WAN port state, such as the external network fault indicated by the jitter surge, and finally the transient pulse type fault is captured by means of the local abnormal event. The multi-dimensional cooperative collection mechanism solves the one-sidedness of the traditional scheme which only focuses on isolated indicators such as bandwidth and delay, and realizes the depth perspective and root cause chain tracing of the home network quality problem.

[0023] In one or more embodiments of the present application, the collection of network operation data can be performed in two ways: one is periodic collection according to a preset time interval; the other is dynamic triggered collection according to actual network diagnosis needs. When the periodic collection mode is used, the system automatically collects the local network operation data at a fixed time frequency. As the default configuration, the length of the collection period is set to be executed once every 5 minutes, which can provide a regular and continuous network health monitoring basic data stream. In actual application, the user or system administrator can adjust the default period parameter according to the specific monitoring granularity and resource consumption. The dynamic collection provides higher flexibility, which allows the data collection action to be initiated immediately when a specific diagnosis task is started or potential abnormalities are detected, to meet the immediacy needs of in-depth analysis or troubleshooting. That is, the user can manually initiate the "one-key diagnosis" through the platform, or the platform can automatically trigger the diagnosis task according to the periodic plan. The control module issues specific collection tasks (such as instantaneous neighbor AP list grabbing, PING delay test, etc.) according to the strategy, dynamically coordinates the collection frequency and data items, and ensures the diagnosis coverage. The two collection methods can be used independently or in combination to serve the network operation state monitoring and diagnosis purposes, and the person skilled in the art can freely choose according to actual needs.

[0024] Step 102: uploading the network operation data to the cloud platform and distributing to the message queue.

[0025] In the embodiment of the present application, the collected network operation data is locally encoded and packaged, uploaded to the access layer of the cloud platform through the MQTT and / or HTTPS protocol, and uniformly received by the message middleware and distributed to the message queue according to the data type.

[0026] Specifically, when the network operation data is collected, first, preprocessing and encoding and packaging operations are performed on the local device or edge node to optimize the data format and ensure transmission safety. Subsequently, the packaged data units are transmitted to the access layer of the cloud platform through a network transmission protocol (specifically, MQTT protocol, HTTPS protocol, or a combination of both protocols according to actual scene requirements). The cloud access layer is deployed with a message middleware (such as Kafka, RabbitMQ, Pulsar, etc.) as a unified entrance, responsible for receiving all uploaded data streams. The message middleware will intelligently classify and route the received data according to preset rules or specific identifiers contained in the data packet, and distribute them to the corresponding message queues of different categories in the backend, realizing asynchronous decoupling of the data source, so as to facilitate subsequent data processing.

[0027] Step 103: identifying the network operation data in the message queue using the terminal identification subsystem and the application identification subsystem.

[0028] In the embodiments of the present application, the network operation data can be input into the terminal identification subsystem to identify the terminal MAC address brand and model, and extract the operating system and network capability information; and the network operation data is input into the application identification subsystem to identify the active application of the terminal based on the DPI or port protocol analysis method.

[0029] Specifically, the network operation data is first input into the terminal identification subsystem, and by analyzing the terminal MAC address feature library, the brand and model of the terminal device are accurately identified, for example, identified as "iPhone 14 Pro", and the operating system version and network capability information of the terminal are extracted at the same time; then, the same network operation data is input into the application identification subsystem, which analyzes the network traffic characteristics of the terminal device in real time based on the deep packet inspection technology or the port protocol analysis method, thereby identifying the current active application type, such as "high-definition video streaming media", "online game", "real-time voice call", providing behavior context support for subsequent network quality diagnosis. The network capability information includes but is not limited to supported Wi-Fi protocol, maximum transmission rate, etc.

[0030] It can be understood that the present application extracts the terminal operating system version and network capability information to solve the problem of network misdiagnosis caused by terminal capability differences: by identifying terminal hardware limitations and protocol compatibility defects, it avoids mistakenly attributing terminal performance bottlenecks to network failures; at the same time, it provides dynamic calibration basis for subsequent network experience scoring model, ensuring the accuracy of the diagnosis results.

[0031] In addition, the terminal identification subsystem maps the MAC address to the device capability profile, which can eliminate the interference of terminal differences on diagnosis; and the application identification subsystem assigns traffic service semantic labels through deep packet inspection (DPI) or port protocol analysis, which can establish the association rules of service types and network indicators. Both of them convert raw data into diagnostic context, so that the subsequent model can dynamically adapt to the scene requirements, and fundamentally avoid the misjudgment caused by semantic blind area in the traditional scheme.

[0032] Step 104: Real-time aggregation and feature extraction of the identified network operation data are performed by using a stream processing framework, and the extracted features are input into a machine learning model for network quality diagnosis. The stream processing framework can include Flink.

[0033] In the embodiments of the present application, the network operation data processed by the terminal identification and application identification subsystems can be real-time aggregated and feature extracted by using a stream processing framework: first, the data is stream-aggregated according to a preset time window to generate statistical features including signal strength mean, packet loss rate standard deviation, and uplink traffic peak; then, time sequence features are extracted based on a sliding window mechanism. The extracted multi-dimensional feature vector is input into a machine learning model (XGBoost or Isolation Forest algorithm can be used) to perform network quality diagnosis: on the one hand, the classification model identifies the explicit failure mode, and on the other hand, the anomaly detection model finds the implicit risk, finally outputs the diagnosis event with confidence and root cause positioning result, and realizes real-time intelligent identification of complex network problems.

[0034] Step 105: In the case of network quality anomaly diagnosis result, the diagnosis result is stored in the data warehouse, and optimization suggestions are generated according to the diagnosis result and displayed on the front-end platform.

[0035] In the embodiments of the present application, the diagnosis result after structured processing can be stored in a high-performance data warehouse such as ClickHouse, and the multi-dimensional persistence strategy is used to guarantee the traceability and query response of historical data: the storage content can cover diagnosis event type, root cause positioning details, network experience score trend, and optimization operation record. The platform front-end dynamically renders the above data into an interactive diagnosis report based on a visualization engine, which can include: Space-time situation chart: shows the network state evolution process in the form of time axis, and associates with diagnosis event marker points; Root cause topology graph: visualizes the physical location of the problem device / AP and the affected link; Optimization effect comparison panel: automatically generates key indicator comparison before and after parameter adjustment.

[0036] Meanwhile, an open standardized API interface is provided for the operation and maintenance system to call, supporting: event subscription, diagnosis retriggering, and remote optimization instruction issuing.

[0037] In the embodiment of the application, when any of the following conditions is met, it can be determined that the network quality is abnormal: the network experience score of the terminal is lower than a threshold value, the rule engine detects an abnormal event, and the light machine learning model identifies an implicit problem.

[0038] The abnormal event at least includes one of the following: a weak signal area, frequent disconnection, and channel conflict; and the implicit problem at least includes one of the following: abnormal load of a wireless access point and existence of an interference source.

[0039] In actual application, the network experience score of the terminal can be dynamically calculated based on signal strength, uplink / downlink rate, packet loss rate, and handover frequency, and when the network experience score is continuously lower than a preset threshold value, for example, the score is lower than 60 for three consecutive periods, it can be determined that the network quality is abnormal; in addition, whether an abnormal event occurs can be detected based on a preset rule library, for example, one of the following conditions exists: a weak signal area (received signal strength indication RSSI≤-75dBm and lasting for more than 30 seconds), frequent disconnection (Wi-Fi disconnection number of a single terminal in 1 hour≥5 times), and channel conflict (number of neighbor APs on the same channel≥3 and channel utilization rate>80%), and then it can be determined that the network quality is abnormal; in addition, a light machine learning model (such as XGBoost or TinyML) can be introduced to further identify an implicit problem, for example, one of the following conditions exists: abnormal load of an AP or existence of an interference source, and then it can be determined that the network quality is abnormal.

[0040] The embodiment of the application significantly improves comprehensive perception by multi-dimensional data collaborative collection, covering key aspects such as terminal state, AP load, neighbor interference, device parameters, and network configuration, and breaks through the limitation of traditional schemes that only monitor isolated indicators such as bandwidth and delay; on this basis, a machine learning algorithm is fused to realize accurate root cause positioning of complex network faults, greatly improving diagnosis accuracy and automation level; further, according to a real-time stream processing architecture, an event-triggered response mechanism is established, which can identify sudden abnormalities and solve the timeliness defect of the periodic polling scheme; then, a closed-loop optimization system is built, which supports generating user-operable suggestions through linkage of diagnosis results and execution strategies, forming a complete closed loop of “detection-analysis-optimization”; finally, the modular design is compatible with various home network environments, and can be quickly deployed on an operator platform or an intelligent routing system, realizing low-cost large-scale application.

[0041] The embodiment of the application also provides a network quality diagnosis device, which comprises: The plug-in collection module is configured to collect network operation data, which includes terminal connection information, Wi-Fi configuration information, a neighbor wireless access point list, WAN port connection state information, and local abnormal event information. The connection and transmission module is configured to upload the network operation data to a cloud platform and distribute the network operation data to a message queue. The data analysis and identification module is configured to identify the network operation data in the message queue by using a terminal identification subsystem and an application identification subsystem. The real-time calculation and diagnosis module is configured to perform real-time aggregation and feature extraction on the identified network operation data by using a stream processing framework, and input the extracted features into a machine learning model to diagnose network quality. The diagnosis control module supports manual initiation of one-key diagnosis by a user, and also supports automatic periodic triggering of automatic diagnosis by the platform, and issuing instructions to the plug-in end to collect specified data. The data storage and display module is configured to store the diagnosis result in a data warehouse when the diagnosis result is a network quality abnormality, and generate optimization suggestions according to the diagnosis result and display the optimization suggestions on a front-end platform.

[0042] The above describes a network quality diagnosis method and device in detail. The principles and implementation manners of the present application are described by using specific examples. The above examples are used to help understand the method and core idea of the present application. Meanwhile, for those skilled in the art, the specific implementation manners and application range can be changed according to the idea of the present application. In summary, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A network quality diagnosis method characterized by comprising: The method comprises: collecting network operation data, wherein the network operation data comprises terminal connection information, Wi-Fi configuration information, a neighbor wireless access point list, WAN port connection state information, and local abnormal event information; uploading the network operation data to a cloud platform and distributing the network operation data to a message queue; identifying the network operation data in the message queue by using a terminal identification subsystem and an application identification subsystem; performing real-time aggregation and feature extraction on the identified network operation data by using a stream processing framework, inputting the extracted features into a machine learning model, and performing network quality diagnosis; in a case where the diagnosis result is network quality abnormality, storing the diagnosis result in a data warehouse, generating an optimization suggestion according to the diagnosis result, and displaying the optimization suggestion on a front-end platform.

2. The method of claim 1, wherein, The terminal connection information comprises a MAC address, signal strength, and a connection state; the Wi-Fi configuration information comprises a channel, a frequency band, a bandwidth, and a security type; the neighbor wireless access point list comprises RSSI, channel conflict, and a BSSID number; the WAN port connection state information comprises an IP state, jitter, and delay; and the local abnormal event information comprises a number of flow interruptions and roaming events.

3. The method of claim 2, wherein, The collecting of the network operation data comprises: periodically collecting local network operation data; and / or dynamically collecting local network operation data according to diagnosis needs.

4. The method of claim 3, wherein, In a case where the local network operation data is periodically collected, a default setting of a collection period is 5 minutes.

5. The method of claim 4, wherein, The uploading of the network operation data to the cloud platform and the distribution of the network operation data to the message queue comprise: locally encoding and packaging the collected network operation data, uploading the network operation data to an access layer of the cloud platform by using MQTT and / or HTTPS protocols, uniformly receiving the network operation data by using a message middleware, and distributing the network operation data to the message queue according to data types.

6. The method of claim 5, wherein, The identification of the network operation data in the message queue by using the terminal identification subsystem and the application identification subsystem comprises: inputting the network operation data into the terminal identification subsystem, identifying a terminal MAC address, and extracting operating system and network capability information; and inputting the network operation data into the application identification subsystem, and identifying active terminal applications by using the application identification subsystem based on a DPI or port protocol analysis mode.

7. The method of claim 6, wherein, The stream processing framework comprises Flink.

8. The method according to claim 1 or 7, characterized in that, In a case where any one of the following conditions is met, it is determined that the network quality is abnormal: a network experience score of a terminal is lower than a threshold value, a rule engine detects an abnormal event, or a lightweight machine learning model identifies an implicit problem.

9. The method of claim 8, wherein, The abnormal event at least comprises one of the following: a weak signal area, frequent disconnection, and channel conflict; and the implicit problem at least comprises one of the following: abnormal wireless access point load and existence of an interference source.

10. A network quality diagnosis apparatus characterized by comprising: The apparatus comprises: a plug-in collection module configured to collect network operation data, wherein the network operation data comprises terminal connection information, Wi-Fi configuration information, a neighbor wireless access point list, WAN port connection state information, and local abnormal event information; a connection and transmission module configured to upload the network operation data to a cloud platform and distribute the network operation data to a message queue; and a terminal identification subsystem and an application identification subsystem. A data analysis and recognition module is configured to recognize the network operation data in the message queue by using a terminal recognition subsystem and an application recognition subsystem; A real-time calculation and diagnosis module is configured to perform real-time aggregation and feature extraction on the recognized network operation data by using a stream processing framework, and input the extracted features into a machine learning model to diagnose the network quality; A data storage and display module is configured to store the diagnosis result in a data warehouse in the case of network quality anomaly, and generate an optimization suggestion according to the diagnosis result and display the optimization suggestion on a front-end platform.

Images