Service path tracking and fault tracing method and device for video monitoring system

Abstract

The invention provides a business path tracking and fault tracing method and device for a video monitoring system, and relates to the technical field of video monitoring fault diagnos.The video monitoring system is divided into a plurality of layers, and multiple types of data acquisition probes are deployed at key nodes of each layer; when a user initiates a video request, generating a Trace ID to be spread along with a video streaming signaling, and reconstructing an end-to-end service path in combination with node information captured by a probe; associating and fusing the multi-dimensional performance data by taking the TraceID as a key, and establishing an end-to-end total delay decomposition and dynamic health baseline; and positioning a root cause node based on the weighted directed graph, the node anomaly confidence score and the fault propagation consistency, displaying a path, data and a traceability result through a visual interface, and generating a precise alarm. According to the method, service link visualization is realized, traditional equipment alarm is upgraded to service fault root cause positioning, the troubleshooting efficiency is improved, and the method is suitable for an isomerized large-scale video monitoring system.

Description

Technical Field

[0001] This invention relates to the field of video surveillance fault diagnosis technology, and in particular to a method, apparatus, and electronic device for business path tracing and fault source tracing in video surveillance systems. Background Technology

[0002] Video surveillance systems have evolved into massive, complex Internet of Things (IoT) systems. A typical video surveillance system usually includes tens of thousands of front-end cameras (IPCs), network video recorders (NVRs), video transmission networks (including access, aggregation, core switches, and routers), video cloud platforms, storage clusters, and various application clients. Its service flow (video stream) transmission path spans wide area networks (WANs), local area networks (LANs), virtualization platforms, and multiple hardware and software components.

[0003] When service quality (QoE) issues such as video stuttering, screen tearing, latency, black screen, and video recording loss occur, traditional operation and maintenance methods face significant challenges:

[0004] Troubleshooting is difficult: problems can occur at any point from the camera to the user's screen (camera encoding, network jitter, excessive device load, platform service anomalies, insufficient decoding capabilities, etc.), and maintenance personnel need to log in to each device one by one to troubleshoot, which is extremely inefficient.

[0005] Troubleshooting relies on experience: Lacking objective and quantitative data support, the troubleshooting process heavily depends on the personal experience of maintenance personnel, making it difficult to guarantee accuracy;

[0006] Lack of panoramic view: It is impossible to intuitively see the complete end-to-end transmission path of a video stream, as well as the real-time health status of each node on the path.

[0007] Existing technical solutions, such as SNMP monitoring, NetFlow analysis, and device log analysis, mostly focus on network or device metrics and are not strongly correlated with the user's video service experience. Distributed tracing technologies (such as OpenTracing and Zipkin) are mature in microservice call chains, but their models and sampling strategies are not directly applicable to continuous, high-bandwidth video streaming services.

[0008] Therefore, there is an urgent need to provide a method for business path tracing and fault tracing in video surveillance systems. Summary of the Invention

[0009] To address the shortcomings of existing technologies, this invention provides a method and apparatus for business path tracing and fault source identification in video surveillance systems. It enables automatic discovery and visualization of end-to-end video stream paths, accurate location of fault root causes, improved operation and maintenance efficiency, and proactive operation and maintenance.

[0010] In a first aspect, the present invention provides a method for business path tracing and fault source tracing in a video surveillance system, comprising the following steps:

[0011] S1) Logically abstract the video surveillance system, divide it into multiple layers, and deploy lightweight data acquisition probes at key nodes of each layer;

[0012] S2) When a user initiates a video viewing request, dynamically construct the end-to-end business path for that video;

[0013] S3) Once the end-to-end service path of the video is constructed, the central analyzer uses the global tracking identifier TraceID as the key and collects and merges multi-dimensional performance data of each node of the path in a fixed time window.

[0014] S4) Decompose the total latency of video services into the sum of the latency of each segment on the path and establish a dynamic health baseline for each key indicator of each node on the path;

[0015] S5) When the client's QoE probe detects a fault or an abnormal total latency of the video service, the fault tracing process is triggered, and the end-to-end path is modeled as a weighted directed graph.

[0016] S6) The constructed business path, real-time performance data of each node, health baseline and fault tracing results are displayed through a visual interface. Accurate alarms are automatically generated based on the root cause, indicating the faulty equipment, abnormal indicators and suggested handling measures.

[0017] Preferably, in step S1), the four layers are: front-end device layer, network transmission layer, platform service layer, and application client layer.

[0018] Preferably, in step S1), the data acquisition probe types include device indicator probes, service flow probes, application service probes, and client QoE probes.

[0019] Preferably, in step S2), the end-to-end service path of the video is dynamically constructed, including the following steps:

[0020] S21) The client or the initial service node generates a globally unique global tracking identifier, TraceID, for this video request;

[0021] S22) TraceID is transmitted along the video stream or signaling throughout the transmission path. When each probe on the path captures a data packet with TraceID, it identifies itself as a node Span of the path and records its own node identifier SpanID, node type and parent node identifier ParentSpanID.

[0022] S23) The central analysis server collects the Span information of all nodes carrying the same TraceID; and reconstructs the complete end-to-end directed acyclic graph path based on the relationship between the node identifier SpanID and the parent node identifier ParentSpanID.

[0023] Preferably, in step S4), the total latency of the video service is... The expression for decomposing the path into the sum of the delays of each segment is:

[0024] ;

[0025] In the formula, Source encoding latency occurs at the front-end device layer; This refers to network transmission latency, which occurs at the network transport layer. The latency handling by the platform occurs at the platform service layer; The client-side decoding and rendering latency occurs at the application client layer.

[0026] Preferably, in step S4), each key indicator for each node on the path is... Establish dynamic health baseline ,Right now:

[0027] ;

[0028] In the formula, This is a trend item; For periodic terms; subscript Indicates the node index; Indicates time; This is the residual.

[0029] Preferably, in step S4), the current index value With health baseline degree of deviation Represented as:

[0030] ;

[0031] In the formula, The standard deviation of historical data; when If the threshold is exceeded, the node is considered to have an abnormality in that indicator.

[0032] Preferably, in step S5), the weighted directed graph... Represented as:

[0033] ;

[0034] In the formula, For vertices; For the edge; As weight;

[0035] Among them, weight As vertices Calculate an anomaly confidence score ,Right now:

[0036] ;

[0037] In the formula, As a contribution weight; Indicates the first position on the path One node; Indicates key indicators All elements are executed The product operation is performed, and then all the results are added together to obtain the final anomaly confidence score. .

[0038] Preferably, in step S5), the root cause node satisfies: Highest; its abnormal timestamp An abnormal time point that is earlier than or equal to that of its downstream node.

[0039] Preferably, in step S5), the determination of the root cause node specifically includes the following steps:

[0040] S51) For weighted directed graphs For all vertices in the array, calculate their... ;

[0041] S52) Obtain the timestamps of when key indicators of each vertex become abnormal. ;

[0042] S53) Select from all vertices The vertices form a candidate set. ;

[0043] S54), on the candidate set Each vertex in Check whether the nodes on its downstream path also subsequently experience anomalies;

[0044] S55), based on the candidate set Each vertex in The root cause node is obtained by filtering the maximum value of the anomaly score and time consistency. .

[0045] Secondly, the present invention provides a service path tracing and fault source tracing device for video surveillance systems, comprising:

[0046] The video surveillance system partitioning module is used to logically abstract the video surveillance system and divide it into multiple layers, and deploy lightweight data acquisition probes at key nodes of each layer.

[0047] The business path construction module is used to dynamically construct the end-to-end business path for a user to initiate a video viewing request;

[0048] The multi-dimensional performance data fusion module is used to perform spatiotemporal alignment and correlation of multi-dimensional performance data of each node in the path collected by the central analyzer;

[0049] The end-to-end latency decomposition module is used to decompose the total latency of video services into the sum of the latencies of each segment on the path;

[0050] The health baseline calculation module is used to establish a dynamic health baseline for each key indicator at each node on the path, and to calculate the degree of deviation between the current indicator value and the health baseline.

[0051] The fault tracing module triggers the fault tracing process when the client's QoE probe detects a fault or an abnormal total latency of the video service, and models the end-to-end path as a weighted directed graph.

[0052] The visualization and alarm module displays the constructed business path, real-time performance data of each node, health baseline, and fault tracing results through a visual interface; and automatically generates accurate alarms based on the root cause.

[0053] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described business path tracing and fault tracing method.

[0054] The beneficial effects of this invention are as follows:

[0055] 1. This invention enables automatic discovery and visualization of end-to-end video stream paths, making invisible business links visible and manageable; through multi-source data fusion and fault tracing algorithms, it elevates traditional device alarms to root cause localization of business faults, greatly improving troubleshooting efficiency; through dynamic baseline learning, it can detect performance degradation trends before users perceive problems, enabling proactive operation and maintenance, and is not dependent on equipment from specific vendors. Through open probes and data models, it is suitable for heterogeneous large-scale video surveillance systems.

[0056] 2. This invention integrates four types of data: device, network, service flow, and application QoE. It uses TraceID to ensure the correlation and consistency of the data and avoids the problem of misjudgment based on a single indicator.

[0057] 3. The data acquisition probe of this invention adopts a lightweight design, reducing the cost of upgrading and deploying large-scale video surveillance systems;

[0058] 4. This invention upgrades the traditional equipment alarm stacking to business fault root cause localization through multi-dimensional data association fusion and weighted directed graph algorithm, avoiding the blindness of maintenance personnel checking equipment one by one;

[0059] 5. The dynamic health baseline constructed based on the STL / LSTM time series algorithm of this invention can adapt to the periodic patterns and trend changes of the system. Through Z-Score anomaly detection, it can detect the performance degradation trend before users perceive business problems, realize the transformation from passively responding to faults to actively preventing problems, and reduce the occurrence rate of business failures. Attached Figure Description

[0060] Figure 1 This is a flowchart illustrating the method of an embodiment of the present invention;

[0061] Figure 2 This is a flowchart illustrating the system of an embodiment of the present invention;

[0062] Figure 3 This is a schematic diagram illustrating the process of constructing the end-to-end service path of the video in an embodiment of the present invention;

[0063] Figure 4 This is a schematic diagram of the process for screening root cause nodes in an embodiment of the present invention. Detailed Implementation

[0064] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0065] Example 1

[0066] like Figure 1 As shown in the figure, this embodiment provides a method for business path tracing and fault source tracing in a video surveillance system, including the following steps:

[0067] S1) Logically abstract the video surveillance system, divide it into four layers, and deploy lightweight data acquisition probes at key nodes in each layer;

[0068] In this embodiment, the four levels are as follows:

[0069] Front-end device layer, such as IPC or encoder;

[0070] Network transport layer, such as switches, routers, and firewalls;

[0071] Platform service layer, such as streaming media service, video recording service, signaling service, and analytics service;

[0072] Application client layer, such as PC client, mobile APP, large screen decoder.

[0073] S2) When a user initiates a video viewing request, dynamically construct the end-to-end business path P for that video;

[0074] S3) Once the end-to-end service path of the video is constructed, the central analyzer uses the global trace identifier TraceID as the key and collects multi-dimensional performance data of each node in the path within a fixed time window. And then merge them; that is:

[0075] ;

[0076] In the formula, For equipment specifications; For network metrics; For business flow metrics; Metrics for application services.

[0077] S4) Decompose the total latency of video services into the sum of the latency of each segment on the path and establish a dynamic health baseline for each key indicator of each node on the path;

[0078] In this embodiment, the total latency of the video service It is a key metric for measuring QoE, which measures the total latency of video services. The expression for decomposing the path into the sum of the delays of each segment is:

[0079] ;

[0080] In the formula, Source encoding latency occurs at the front-end device layer; Network transmission latency occurs at the network transport layer; it can be further decomposed into the latency of each network hop. ; The latency handling for the platform occurs at the platform service layer, such as the time spent on stream forwarding, transcoding, and analysis. The client-side decoding and rendering latency occurs at the application client layer.

[0081] The central analyzer learns from historical data to analyze each key indicator for each node along the path. Establish dynamic health baseline Health baseline in this embodiment Periodic patterns, such as day or night patterns, can be considered, and time series forecasting algorithms such as STL decomposition or LSTM can be used to calculate the results.

[0082] ;

[0083] In the formula, This is a trend item; For periodic terms; subscript Indicates the node index; Indicates time; This is the residual.

[0084] In this embodiment, the current indicator value With health baseline degree of deviation Represented as:

[0085] ;

[0086] In the formula, The standard deviation of historical data; when If the threshold is exceeded, the node is considered to have an abnormality in that indicator.

[0087] S5) When the client's QoE probe detects a fault or an abnormal total latency in the video service, the fault tracing process is triggered, and the end-to-end path is modeled as a weighted directed graph. ;

[0088] The weighted directed graph Represented as:

[0089] ;

[0090] In the formula, For vertices; For the edge; As weight;

[0091] Among them, weight As vertices Calculate an anomaly confidence score ,Right now:

[0092] ;

[0093] In the formula, As a contribution weight; Indicates the first position on the path One node; Indicates key indicators All elements are executed The product operation is performed, and then all the results are added together to obtain the final anomaly confidence score. .

[0094] The aforementioned anomaly confidence score It takes into account the degree of abnormal deviation of all its indicators. and its contribution weight to the final business failure In this embodiment, the contribution weight is... Pre-set parameters through expert systems or machine learning, such as network jitter. The contribution weight of lag is greater than that of CPU utilization; troubleshooting is then transformed into the problem of finding the root cause node in this directed graph;

[0095] S6) The constructed business path, real-time performance data of each node, health baseline and fault tracing results are displayed through a visual interface. Accurate alarms are automatically generated based on the root cause, indicating the faulty equipment, abnormal indicators and suggested handling measures.

[0096] This embodiment realizes automatic discovery and visualization of the end-to-end path of video streams, making invisible business links visible and manageable. Through multi-source data fusion and fault tracing algorithms, traditional equipment alarms are upgraded to "business fault root cause localization", which greatly improves troubleshooting efficiency. Through dynamic baseline learning, performance degradation trends can be detected before users perceive the problem, realizing proactive operation and maintenance. Moreover, this method does not depend on equipment from specific manufacturers. Through open probes and data models, it is suitable for heterogeneous large-scale video surveillance systems.

[0097] Example 2

[0098] like Figure 2 As shown, this embodiment provides a service path tracing and fault source tracing device for video surveillance systems, including the following modules:

[0099] The video surveillance system partitioning module is used to logically abstract the video surveillance system and divide it into four layers. In this embodiment, lightweight data acquisition probes are deployed at key nodes of each layer to achieve data acquisition at different layers. The four layers are as follows:

[0100] Front-end device layer, such as IPC or encoder;

[0101] Network transport layer, such as switches, routers, and firewalls;

[0102] Platform service layer, such as streaming media service, video recording service, signaling service, and analytics service;

[0103] Application client layer, such as PC client, mobile APP, large screen decoder.

[0104] The business path construction module is used to dynamically construct the end-to-end business path for a user to initiate a video viewing request; when a user initiates a video viewing request, the business path construction module dynamically constructs the end-to-end business path P for that video.

[0105] The multi-dimensional performance data fusion module is used to perform spatiotemporal alignment and correlation of multi-dimensional performance data of each node in the path collected by the central analyzer;

[0106] In this embodiment, after the end-to-end service path of the video path is constructed, the central analyzer uses the global tracking identifier TraceID as the key and collects multi-dimensional performance data of each node of the path in a fixed time window. And then they are integrated, specifically as follows:

[0107] ;

[0108] In the formula, For equipment specifications; For network metrics; For business flow metrics; Metrics for application services.

[0109] The end-to-end latency decomposition module is used to decompose the total latency of video services into the sum of the latencies of each segment on the path;

[0110] In this embodiment, the total latency of the video service It is a key metric for measuring QoE, which measures the total latency of video services. The expression for decomposing the path into the sum of the delays of each segment is:

[0111] ;

[0112] In the formula, Source encoding latency occurs at the front-end device layer; Network transmission latency occurs at the network transport layer; it can be further decomposed into the latency of each network hop. ; The latency handling for the platform occurs at the platform service layer, such as the time spent on stream forwarding, transcoding, and analysis. The client-side decoding and rendering latency occurs at the application client layer.

[0113] The health baseline calculation module is used to establish a dynamic health baseline for each key indicator at each node on the path, and to calculate the degree of deviation between the current indicator value and the health baseline.

[0114] In this embodiment, the central analyzer learns from historical data to analyze each key indicator of each node on the path. Establish dynamic health baseline Health baseline in this embodiment Periodic patterns, such as day or night patterns, can be considered, and time series forecasting algorithms such as STL decomposition or LSTM can be used to calculate the results.

[0115] ;

[0116] In the formula, This is a trend item; For periodic terms; subscript Indicates the node index; Indicates time; This is the residual.

[0117] In this embodiment, the current indicator value With health baseline degree of deviation Represented as:

[0118] ;

[0119] In the formula, The standard deviation of historical data; when If the threshold is exceeded, the node is considered to have an abnormality in that indicator.

[0120] The fault tracing module triggers the fault tracing process when the client's QoE probe detects a fault or an abnormal total latency of the video service, and models the end-to-end path as a weighted directed graph.

[0121] The weighted directed graph Represented as:

[0122] ;

[0123] In the formula, For vertices; For the edge; As weight;

[0124] Among them, weight As vertices Calculate an anomaly confidence score ,Right now:

[0125] ;

[0126] In the formula, As a contribution weight; Indicates the first position on the path One node; Indicates key indicators All elements are executed The product operation is performed, and then all the results are added together to obtain the final anomaly confidence score. .

[0127] The aforementioned anomaly confidence score It takes into account the degree of abnormal deviation of all its indicators. and its contribution weight to the final business failure In this embodiment, the contribution weight is... Pre-set parameters through expert systems or machine learning, such as network jitter. The contribution weight of lag is greater than that of CPU utilization; troubleshooting is then transformed into the problem of finding the root cause node in this directed graph;

[0128] The visualization and alarm module displays the constructed business path, real-time performance data of each node, health baseline, and fault tracing results through a visual interface; and automatically generates accurate alarms based on root causes, indicating faulty equipment, abnormal indicators, and suggested handling measures.

[0129] Example 3

[0130] The data acquisition probe types include device indicator probes, service flow probes, application service probes, and client QoE probes;

[0131] The device indicator probes are used to collect data on CPU, memory, disk I / O, network port traffic, packet error rate, etc.

[0132] The aforementioned service flow probe is based on technologies such as DPDK or AF_PACKET to bypass and capture and analyze network packets, extracting key metrics of the video stream, such as: bitrate, frame rate (FPS), key frame interval (GOP), network jitter, packet loss, and transmission latency.

[0133] The application service probe is used to collect information such as service process resource usage, service response time, service queue length, and log error information.

[0134] The aforementioned client-side QoE probe is integrated into the client and is used to collect user experience metrics such as video rendering frame rate, first frame opening time, number of stutters, and subjective rating (MOS).

[0135] Example 4

[0136] like Figure 3 As shown, dynamically constructing the end-to-end service path for this video stream involves the following steps:

[0137] S21) The client or the initial service node generates a unique global tracking identifier, TraceID, for this video request;

[0138] S22) TraceID is transmitted along the video stream (such as through RTP or RTSP extension headers or custom RTCP packets) or signaling (such as SIP, ONVIF) throughout the transmission path. When each probe on the path captures a data packet with TraceID, it identifies itself as a node Span of the path and records its own node identifier SpanID, node type and parent node identifier ParentSpanID.

[0139] S23) The central analysis server collects the Span information of all nodes carrying the same TraceID; and reconstructs the complete end-to-end directed acyclic graph path based on the relationship between the node identifier SpanID and the parent node identifier ParentSpanID.

[0140] The relationships between nodes are represented as follows:

[0141] In the formula, Indicates the first position on the path 1 node For the first Each node represents the destination, indicating the flow originating from... Flow direction ;

[0142] The complete end-to-end directed acyclic graph path Represented as:

[0143] ;

[0144] In the formula, Source.

[0145] Example 5

[0146] In this embodiment, fault tracing is transformed into the problem of finding the root cause node in the directed graph. The root cause node satisfies the following condition: its anomaly confidence score... Highest; its abnormal timestamp An abnormal time point that is earlier than or equal to its downstream node, such as Figure 4 As shown, the specific steps include the following:

[0147] S51) Calculate the weighted directed graph. Anomaly confidence scores of all vertices ;

[0148] S52) Obtain the timestamps of when key indicators of each vertex become abnormal. ;

[0149] S53) Filter out the anomaly confidence scores from all vertices. The vertices form a candidate set. ;

[0150] S54), on the candidate set Each vertex in Check whether the nodes on its downstream path also subsequently experience anomalies;

[0151] This embodiment defines a propagation consistency index. for:

[0152] in, This indicates the number of elements that meet a specific condition. Indicates and All those with downstream relationships The set that constitutes;

[0153] S55), based on the candidate set Each vertex in The root cause nodes were obtained by filtering the maximum values ​​of the anomaly confidence score and the propagation consistency index. ,Right now:

[0154] ;

[0155] In the formula, Represents vertices The abnormal confidence score; Represents vertices Indicators of consistency in transmission; and It is a harmonic parameter used to balance the anomaly confidence score and the propagation consistency index.

[0156] Example 6

[0157] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the end-to-end service path tracing and fault tracing method described above.

[0158] In this embodiment, the memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. A processor, coupled to the memory, is used to execute computer programs stored in the memory.

[0159] The computer program includes computer program code, which may be in the form of source code, object code, executable file, or some intermediate form.

[0160] Example 7

[0161] This embodiment uses the method described in the above embodiment to solve the problem of severe video stuttering at an intersection in a large-scale urban video surveillance system with 100,000 IPCs, as detailed below:

[0162] Data is continuously collected using various probes deployed on front-end IPCs, transmission network switches, streaming media servers, storage servers, and clients;

[0163] A user initiates a request to view the video of this intersection, generating... When the streaming media server processes this stream, it will Injected into RTP packets; network probes and service probes along the path are identified. And report its own node information;

[0164] The central analyzer constructs the complete path: ;

[0165] The client-side QoE probe detected a sudden drop in frame rate from 25 FPS to 5 FPS, with a surge in stuttering incidents. It immediately reported the anomaly and triggered source tracing. The time point was... ;

[0166] Central Analyzer Pull Within the time window, the path Data for all nodes:

[0167] Bitrate and frame rate are normal, CPU is normal;

[0168] Port error rate The baseline was 0.01%, which has now surged to 15%. Extremely high, abnormal time ;

[0169] Because of its upstream Packet loss caused the received streaming service metrics to drop. The packet loss rate is abnormal. ;

[0170] Due to the receipt of network packets with packet loss, processing efficiency has decreased, resulting in service delays. Slightly increased, ;

[0171] The incomplete video stream caused decoding difficulties, resulting in stuttering. ;

[0172] Calculate the anomaly score for each node: of Highest (due to its extremely high port error rate and high weighting); Consistency of inspection time: The abnormal time (14:04:50) is earlier than all its downstream nodes. , Approaching version 1.0, therefore, the core switch is determined to be... This is the root cause node.

[0173] On the large screen, maintenance personnel saw that in this video path The node is highlighted in red, and the alarm message is: Core Switch The error rate of port GigabitEthernet1 / 0 / 1 is abnormally high at 15%, causing downstream video streaming to stutter.

[0174] The maintenance personnel immediately checked the switch port and found that the optical module was faulty. After replacement, the service was restored.

[0175] The embodiments and descriptions above are merely illustrative of the principles and preferred embodiments of the present invention. Various changes and modifications may be made to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed.

Claims

1. A method for business path tracing and fault source identification in video surveillance systems, characterized in that, Includes the following steps: S1) Divide the video surveillance system into multiple layers and deploy data acquisition probes at key nodes in each layer; S2) When a user initiates a video viewing request, dynamically construct the end-to-end business path for that video; S3) Use the central analyzer to collect and merge multi-dimensional performance data of each node in the business path using the global trace identifier TraceID as the key. S4) Decompose the total latency of video services into the sum of the latency of each segment on the path and establish a dynamic health baseline for each key indicator of each node on the path; S5) When the client's QoE probe detects a fault or an abnormal total latency in the video service, the fault tracing process is triggered, and the end-to-end service path is modeled as a weighted directed graph, transforming fault tracing into finding the root cause node in the weighted directed graph.

2. The service path tracing and fault source tracing method for video surveillance systems according to claim 1, characterized in that: In step S2), the end-to-end service path of the video is dynamically constructed, including the following steps: S21) The client or the initial service node generates a globally unique global tracking identifier, TraceID, for this video request; S22) The global tracking identifier TraceID is transmitted along the entire transmission path with the video stream or signaling. When each probe on the path captures a data packet with TraceID, it identifies itself as a node Span of the path and records its own node identifier SpanID, node type and parent node identifier ParentSpanID. S23) The central analysis server collects Span information from all nodes carrying the same TraceID; The complete end-to-end directed acyclic graph path is reconstructed based on the relationship between the node identifier SpanID and the parent node identifier ParentSpanID.

3. The service path tracing and fault source tracing method for video surveillance systems according to claim 1, characterized in that: Step S4), the total latency of the video service The expression for decomposing the path into the sum of the delays of each segment is: ； In the formula, Source encoding latency occurs at the front-end device layer; This refers to network transmission latency, which occurs at the network transport layer. The latency handling by the platform occurs at the platform service layer; The client-side decoding and rendering latency occurs at the application client layer.

4. The service path tracing and fault source tracing method for video surveillance systems according to claim 3, characterized in that: In step S4), the central analyzer learns from historical data for each key indicator of each node on the path. Establish dynamic health baseline ,Right now: ； In the formula, This is a trend item; For periodic terms; subscript Indicates the node index; Indicates time; This is the residual.

5. The service path tracing and fault source tracing method for video surveillance systems according to claim 4, characterized in that: In step S4), the current indicator value With health baseline degree of deviation Represented as: ； In the formula, The standard deviation of historical data; when If the threshold is exceeded, the node is considered to have an abnormality in that indicator.

6. The service path tracing and fault source tracing method for video surveillance systems according to claim 5, characterized in that: In step S5), the weighted directed graph Represented as: ； In the formula, For vertices; For the edge; As weight; Among them, weight As vertices Calculate an anomaly confidence score ,Right now: ； In the formula, As a contribution weight; Indicates the first position on the path One node; Indicates key indicators All elements are executed The product operation is performed, and then all the results are added together to obtain the final anomaly confidence score. .

7. The service path tracing and fault source tracing method for video surveillance systems according to claim 6, characterized in that: In step S5), the root cause node satisfies the following condition: Its abnormal confidence score Highest; Its abnormal timestamp An abnormal time point that is earlier than or equal to that of its downstream node.

8. The service path tracing and fault source tracing method for video surveillance systems according to claim 7, characterized in that: Step S5), the selection of root cause nodes, specifically includes the following steps: S51) Calculate the weighted directed graph. Anomaly confidence scores of all vertices ; S52) Obtain key indicators for each vertex The timestamp of the exception ; S53) Filter out the anomaly confidence scores from all vertices. The vertices form a candidate set. ; S54), for candidate set Each vertex in Check whether the nodes on its downstream path also subsequently experience anomalies; S55), based on the candidate set Each vertex in The root cause node was obtained by filtering the maximum values ​​of the abnormal confidence scores and time consistency. .

9. A service path tracing and fault source tracing device for video surveillance systems, characterized in that, The device utilizes the method described in any one of claims 1-8 to achieve business path tracing and fault source tracing, and the device comprises: The video surveillance system partitioning module is used to divide the video surveillance system into multiple layers and deploy lightweight data acquisition probes at key nodes in each layer. The business path construction module is used to dynamically construct the end-to-end business path for a user to initiate a video viewing request; The multi-dimensional performance data fusion module is used to perform spatiotemporal alignment and correlation of multi-dimensional performance data of each node in the business path collected by the central analyzer; The end-to-end latency decomposition module is used to decompose the total latency of video services into the sum of the latencies of each segment on the path; The health baseline calculation module is used to learn historical data through the central analyzer to establish a dynamic health baseline for each key indicator of each node on the path, and to calculate the degree of deviation of the current indicator value from the health baseline. The fault tracing module triggers the fault tracing process when the client's QoE probe detects a fault or an abnormal total latency in the video service. It models the end-to-end path as a weighted directed graph, transforming fault tracing into finding the root cause node in the weighted directed graph. The visualization and alarm module displays the constructed business path, real-time performance data of each node, health baseline, and fault tracing results through a visual interface; and automatically generates accurate alarms based on the root cause.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, it implements the business path tracing and fault tracing method according to any one of claims 1-8.

Images