Intelligent diagnosis method and device for SRv6 network fault

Abstract

The invention provides an SRv6 network fault intelligent diagnosis method and device, and the method comprises the steps: obtaining a flow feature vector between a head node and a tail node in a lower segment list set in a flow engineering strategy mode in an SRv6 network, and the flow feature vector comprises end-to-end time delay, network time delay jitter, a packet loss rate and a link bandwidth utilization rate; inputting the traffic feature vector into a pre-trained data dimension reduction clustering model to obtain a plurality of network state tags output by the pre-trained data dimension reduction clustering model, each network state tag in the plurality of network state tags being used for representing the network state of each traffic engineering strategy mode; according to the invention, the fault positioning efficiency of the SRv6 network can be improved.

Description

Technical Field

[0001] The present invention relates to the field of computer network technology, and in particular to an intelligent diagnosis method and device for SRv6 network failures. Background Art

[0002] The SRv6 TE Policy (Segment Routing IPv6 Traffic Engineering Policy) network is an application of IPv6-based segment routing technology. It is mainly used to implement traffic engineering, improve network quality, and meet end-to-end service requirements.

[0003] When SRv6 TE Policy fails, it has a significant impact on the traffic and transmission quality of the service. Since most of the current operation and maintenance activities for SRv6 TE Policy still rely on manual participation, the common maintenance mode involves understanding a large amount of device-related information for each network element. The daily maintenance service and operation and maintenance labor costs account for a high proportion, the degree of automation is low, and the efficiency is low.

[0004] Therefore, it can be seen that the fault diagnosis method of the SRv6 TE Policy network in the related art has technical problems of low automation and low efficiency. Summary of the invention

[0005] The present invention provides an intelligent diagnosis method and device for SRv6 network faults, which are used to solve the defects of low automation and low efficiency in the fault diagnosis method of SRv6 TE Policy network in the prior art, realize regular detection of SRv6 network status, reduce the fault time of SRv6 TE Policy, and improve the fault location efficiency of SRv6 TE Policy.

[0006] The present invention provides an intelligent diagnosis method for SRv6 network faults, comprising the following steps: obtaining a traffic feature vector between a head node and a tail node in a segment list set in a traffic engineering strategy mode in an SRv6 network, wherein the traffic feature vector includes end-to-end delay, network delay jitter, packet loss rate, and link bandwidth utilization; The traffic feature vector is input into a pre-trained data dimensionality reduction clustering model to obtain multiple network status labels output by the pre-trained data dimensionality reduction clustering model, wherein each of the multiple network status labels is used to represent the network status of each of the traffic engineering strategy modes.

[0007] According to an intelligent diagnosis method for SRv6 network faults provided by the present invention, before obtaining the traffic feature vector between the head node and the tail node in the segment list set under the traffic engineering policy mode in the SRv6 network, the method also includes: determining a first timestamp of a single data packet passing through the head node in the segment list set under the traffic engineering policy mode in the SRv6 network; determining a second timestamp of the single data packet passing through the tail node in the segment list set under the traffic engineering policy mode in the SRv6 network; taking the difference between the second timestamp and the first timestamp as the single list delay of the single data packet under the traffic engineering policy mode in the SRv6 network; determining the sum of the single list delays of all data packets within the target time, and taking the ratio of the sum to the number of all data packets as the end-to-end delay.

[0008] According to an intelligent diagnosis method for SRv6 network faults provided by the present invention, after determining the sum of the single list delays of all data packets within the target time and taking the ratio of the sum to the number of all data packets as the end-to-end delay, the method further includes: determining the single list delay of each data packet among all the data packets within the target time; taking the ratio of the sum of the squares of the difference between the single list delay of each data packet and the end-to-end delay to the number of all data packets as the network delay jitter.

[0009] According to an intelligent diagnosis method for SRv6 network faults provided by the present invention, before obtaining the traffic feature vector between the head node and the tail node in the segment list set under the traffic engineering policy mode in the SRv6 network, the method also includes: determining the total number of sent data packets of the head node in the segment list set under the traffic engineering policy mode in the SRv6 network within the target time; determining the total number of received data packets of the tail node in the segment list set under the traffic engineering policy mode in the SRv6 network within the target time; and taking the percentage of the ratio of the total number of received data packets to the total number of sent data packets as the packet loss rate.

[0010] According to an intelligent diagnosis method for SRv6 network faults provided by the present invention, before obtaining the traffic feature vector between the head node and the tail node in the segment list set under the traffic engineering policy mode in the SRv6 network, the method also includes: determining the end-to-end actually used total bandwidth of the target link under the traffic engineering policy mode in the SRv6 network at the target time; determining the end-to-end available total bandwidth of the target link under the traffic engineering policy mode in the SRv6 network at the target time; and taking the percentage of the ratio of the end-to-end actually used total bandwidth to the end-to-end available total bandwidth as the link bandwidth utilization.

[0011] According to an intelligent diagnosis method for SRv6 network faults provided by the present invention, before the traffic feature vector is input into a pre-trained data dimensionality reduction clustering model to obtain multiple network status labels output by the pre-trained data dimensionality reduction clustering model, the method also includes: pre-training the data dimensionality reduction clustering model for a target number of iterations based on a training sample set: setting the number of neuron nodes in the competition layer of the data dimensionality reduction clustering model according to the number of training samples in the training sample set, and initializing the node weights of the neuron nodes in the competition layer; inputting the pre-processed training sample set into the data dimensionality reduction clustering model, traversing each of the neuron nodes in the competition layer to determine the similarity between the training sample and the neuron node; taking the neuron node with the highest similarity as the winning node; determining the neighboring neuron nodes included in the winning neighborhood based on the winning node according to a preset domain radius; updating the node weights of the neighboring neuron nodes; and obtaining the pre-trained data dimensionality reduction clustering model when the current number of iterations is equal to the target number of iterations.

[0012] The present invention also provides an intelligent diagnostic device for SRv6 network faults, comprising the following modules: an acquisition module, used to obtain the traffic feature vector between the head node and the tail node in the segment list set under the traffic engineering strategy mode in the SRv6 network, wherein the traffic feature vector includes end-to-end delay, network delay jitter, packet loss rate and link bandwidth utilization; an output module, used to input the traffic feature vector into a pre-trained data dimensionality reduction clustering model to obtain multiple network status labels output by the pre-trained data dimensionality reduction clustering model, wherein each of the multiple network status labels is used to represent the network status of each of the traffic engineering strategy modes.

[0013] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, an intelligent diagnosis method for SRv6 network failure as described above is implemented.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described intelligent diagnosis methods for SRv6 network failures.

[0015] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the intelligent diagnosis method for SRv6 network failure as described in any one of the above is implemented.

[0016] The intelligent diagnosis method and device for SRv6 network faults provided by the present invention use traffic feature vectors such as end-to-end delay, delay jitter, packet loss rate and bandwidth utilization rate based on SRv6 network characteristics, and call a pre-trained data dimensionality reduction clustering model to detect the SRv6 network based on the traffic feature vectors. Therefore, it is possible to perform fault analysis on the traffic feature data according to the data dimensionality reduction clustering model, thereby improving the fault location efficiency of SRv6 TE Policy. This solves the technical problem of low automation and low efficiency in the fault diagnosis method of SRv6 TE Policy network in the related art. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced one by one below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 It is a flow chart of the intelligent diagnosis method for SRv6 network failure provided by the present invention.

[0019] Figure 2 It is a schematic diagram of the SRv6 TE Policy network fault diagnosis simulation process provided by the present invention.

[0020] Figure 3 This is a KPI data dimensionality reduction clustering model diagram provided by the present invention.

[0021] Figure 4 This is a historical KPI data training flow chart provided by the present invention.

[0022] Figure 5 It is a schematic diagram of the structure of the SRv6 TE Policy network topology provided by the present invention.

[0023] Figure 6 It is a structural schematic diagram of the intelligent diagnosis device for SRv6 network failure provided by the present invention.

[0024] Figure 7 It is a schematic diagram of the physical structure of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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 creative work are within the scope of protection of the present invention.

[0026] SRv6 is a key technology in the IPv6+ technology system. It is a technical innovation based on IPv6 and Segment Routing (SR) technology. It is implemented by defining the Segment Routing Header in the IPv6 Routing Header extension header. SRv6 TE Policy is a working mode in SRv6 technology. SRv6 TE Policy uses the source routing mechanism of Segment Routing to guide packets to traverse the network by encapsulating an ordered instruction list (path information) in the head node. SRv6 TE Policy is used to implement traffic engineering, improve network quality, and meet the end-to-end requirements of services.

[0027] The service traffic carried in SRv6 TE Policy has high requirements for the network's service-level agreement (SLA) and refined scheduling. When SRv6 TE Policy fails, the service traffic and transmission quality are greatly affected. Since most of the current operation and maintenance activities for SRv6 TE Policy still rely on manual participation, the common maintenance mode involves understanding a large amount of device-related information for each network element device. The daily maintenance service and operation and maintenance labor costs account for a high proportion, the degree of automation is low, and the efficiency is low.

[0028] Based on the above analysis, an embodiment of the present invention proposes an intelligent diagnosis method for SRv6 TE Policy network faults. In this method, the end-to-end data collection, fault introduction, algorithm execution, and simulation result output processes of fault diagnosis are first described, and a dimensionality reduction clustering algorithm based on multi-dimensional KPI data of the SRv6 network is proposed in particular.

[0029] In the related art, the patent publication number CN115766552A "Network Measurement Method and Device Based on SRv6 and INT" discloses a network measurement method and device based on SRv6 and INT, which includes a telemetry demand generation module, a path generation module, a data packet generation module, a source node, a switch, a destination node, a telemetry monitoring module and a telemetry data analysis module, which are connected in sequence. The main process is that the path generation module generates a telemetry path, the telemetry data packet generation module generates a telemetry data packet, the telemetry monitoring module monitors the network status information, and the telemetry data analysis module analyzes the network status. It can realize the collection of specified link status information, and parse the telemetry indicators of the link into indicators of two nodes for measurement; make full use of the space of the Segment List field in the SRH to guide the forwarding of telemetry data packets and realize the collection of telemetry data at the same time; the telemetry data of the node or link can be specified without collecting all the information on the data packet forwarding path.

[0030] The above-mentioned related technical solutions can realize the collection of specified link status information. However, for large-scale network-level information collection, the overhead is relatively large, such as the message extension header field is too long and the number of telemetry data packets is too large. At the same time, there is no clear description and analysis of how to process the collected information.

[0031] In the related technology, the patent publication number CN113328915A, "A Method for Measuring a Network on SRv6", relates to a method for measuring a network on SRv6, which solves the problem that the encapsulation mode of the method for measuring a network on SRv6 will increase the processing time of the node and reduce the processing efficiency of the node. A new OAM I SID is designed, which is suitable for measuring a network on SRv6 in ordinary SRv6 and SRv6L3VPN network scenarios. New SID additional behaviors USD+ and USP+ are designed. When Segment Left in SRH is 1 and the second bit of Flags is 1, the collected data is uploaded and the outer IPv6 header and all extended headers are removed. A new format for uploading collected data is designed, and the message includes OAM I SID and collected data. The Postcard mode of network on SRv6 measurement is adopted, and the instruction header in the network on SRv6 measurement is encapsulated in the Segment List[0] of SRH, which improves the processing efficiency of the node.

[0032] The above-mentioned related technical solutions solve the problem that the encapsulation mode of the SRv6-based network measurement method will increase the processing time of the node and reduce the node processing efficiency. However, each message needs to expand the message header, which has a large overhead; the message length will also expand infinitely with the increase of the number of hops.

[0033] In the related technology, the patent publication number is CN113709000A "A method and system for detecting the bidirectional connectivity of SRv6 tunnels by ping", which discloses a method for detecting the bidirectional connectivity of SRv6 tunnels by ping, including: the ping initiator sends an ICMPv6Request message carrying preset information to the destination, and the preset information is a color value or a Binding SID value; the destination performs a matching search for the return path according to the preset information in the ICMPv6Request message to find a suitable return tunnel; the destination uses the found return tunnel to construct an SRH header and respond to the ICMPv6Reply message. Through this invention method: operation and maintenance personnel can use a simple ping to detect the bidirectional connectivity of the SRv6 tunnel path at a single point, thereby improving operation and maintenance efficiency; operation and maintenance personnel can customize the path when pinging the SRv6 tunnel, thereby improving the flexibility of connectivity detection; this detection method simulates the forwarding method of real business traffic and can better reflect the SRv6 network status.

[0034] The above-mentioned related technical solutions can detect the bidirectional connectivity of the specified path of the SRv6 tunnel at a single point. However, the path needs to be specified manually, the degree of automation is low, and it can only be used for the detection of the specified path, which is not suitable for fault diagnosis at the SRv6 network level.

[0035] In summary, the above-mentioned related technical solutions are not suitable for fault diagnosis of SRv6 networks, mainly due to the defects of low automation and high overhead. This patent mainly focuses on how to complete the intelligent location of List faults by only analyzing the data of the nodes at both ends of the Policy when a fault occurs in an individual List in the SRv6 TE Policy, thereby reducing the fault diagnosis time and ensuring the reliability of the SRv6 network.

[0036] refer to Figure 1 , Figure 1 FIG. 1 is a flow chart of the intelligent diagnosis method for SRv6 network failure provided by the present invention, such as Figure 1 As shown, the method includes the following steps.

[0037] Step 101, obtain the traffic feature vector between the head node and the tail node in the segment list set in the traffic engineering policy mode in the SRv6 network, wherein the traffic feature vector includes end-to-end delay, network delay jitter, packet loss rate and link bandwidth utilization.

[0038] TE Policy in SRv6 TE Policy refers to the Traffic Engineering policy based on SRv6 technology. SRv6 (Segment Routing over IPv6) is a segment routing technology based on IPv6. It implements fine-grained control of the packet path by embedding the segment identifier (SID) in the extended header field of the IPv6 packet header. SRv6 TE Policy is a new tunnel diversion technology developed based on SRv6 technology. It is used to implement traffic engineering to meet the network's needs for flexible scheduling and optimization of service traffic.

[0039] SRv6 TE Policy allows network administrators to customize the path of data packets as needed. By specifying a Segment ID List (SID List), the end-to-end path from source to destination is defined to implement flexible traffic engineering.

[0040] An SRv6 TE Policy can contain multiple candidate paths, each of which carries a priority attribute. The valid candidate path with the highest priority is used as the primary path. In addition, a candidate path can contain multiple SID lists, which can form load balancing to improve the utilization of network resources.

[0041] In the embodiment of the present invention, in the end-to-end data collection phase, the network status and key performance indicators in the current SRv6 TE Policy mode are mainly acquired periodically. The data acquisition source can be the head and tail nodes configured with SRv6 TE Policy in the SRv6 network, or offline data files. The data and definitions to be acquired include: G:SRv6 network topology is an abstraction of the physical structure of the network.

[0042] N: A collection of network nodes, which is an abstraction of routers that support SRv6 features in the network.

[0043] L: The set of Segment Lists in the SRv6 network.

[0044] Step 102, input the traffic feature vector into a pre-trained data dimensionality reduction clustering model to obtain multiple network status labels output by the pre-trained data dimensionality reduction clustering model, wherein each of the multiple network status labels is used to represent the network status of each traffic engineering strategy mode.

[0045] refer to Figure 2 , Figure 2The figure is a schematic diagram of the SRv6 TE Policy network fault diagnosis simulation process provided by the present invention, wherein the SRv6 TE Policy network fault diagnosis simulation process mainly includes end-to-end data collection, fault introduction, algorithm execution, and diagnosis result output.

[0046] In the embodiment of the present invention, the end-to-end data collection phase mainly periodically obtains the network status and key performance indicators under the current SRv6 TE Policy mode. The data acquisition source can be the head and tail nodes configured with SRv6 TEPolicy in the SRv6 network, or offline data files. The fault perception phase refers to the phase in which the system is operating normally, and perceives the second-level or even millisecond-level network fault conditions through the end-to-end traffic data in the SRv6 network collected in real time in the previous phase; after the fault is introduced, the algorithm execution phase determines the network fault status based on the fault intelligent diagnosis model based on KPI data and the end-to-end data collected in the current period; after the algorithm is executed, the result output phase displays the status of each Policy service in a table or other manner. If a service fails, the Segment List-level fault location in the SRv6 network is achieved.

[0047] In the embodiment of the present invention, the location of network faults is time-related. The end-to-end data within 1 second at the moment is statistically analyzed as the traffic characteristics at the current moment. To simplify the problem, we consider the node Only one policy is configured, in which different Segment Lists are assigned different weights, and List-level fault location is achieved through different end-to-end traffic characteristics.

[0048] In summary, the following fault diagnosis mathematical model is established (i.e., the data dimension reduction clustering model mentioned above).

[0049] To input the feature vector, you can refer to the following formula (1): (1)

[0050] in, Represents a fault diagnosis model based on SRv6 network key performance indicator (KPI, Key Performance Indicator) data, and the input feature vector Respectively represent the current time Configure SRv6 TE policy on both ends of the node The delay, delay jitter, packet loss rate and bandwidth utilization between SRv6 and SRv6 networks are closely related to the status of SRv6 networks. Understand the status of the network and detect abnormal faults in time.

[0051] The output function can refer to the following formula (2): (2)

[0052] in, Represents an output function. The output is a label of different network states, representing the fault diagnosis result. For example, 0 represents normal Policy performance, 1 represents List 1 fault, etc.

[0053] refer to Figure 3 , Figure 3 It is a KPI data dimension reduction clustering model diagram provided by the present invention, which includes a competition layer and an input layer; The model inputs include: delay, delay jitter, packet loss rate, and bandwidth utilization; The output of the model is Output function.

[0054] This model maps multi-dimensional KPI data into a two-dimensional neural network in the competition layer. The topological structure of the neurons represents the distribution of the original data: the closer the neurons are, the closer the original data they map is. This achieves the two-dimensional representation of high-dimensional data under different traffic characteristics and the preliminary clustering of corresponding network faults, thereby realizing the diagnosis of SRv6 network faults.

[0055] Through the above steps of the embodiment of the present invention, a traffic feature vector between the head node and the tail node in the segment list set in the traffic engineering policy mode in the SRv6 network is obtained, wherein the traffic feature vector includes end-to-end delay, network delay jitter, packet loss rate and link bandwidth utilization; the traffic feature vector is input into a pre-trained data dimensionality reduction clustering model to obtain a plurality of network status labels output by the pre-trained data dimensionality reduction clustering model, wherein each of the plurality of network status labels is used to represent the network status of each traffic engineering policy mode, thereby improving the fault location efficiency of the SRv6 TE Policy; thereby solving the technical problem that the fault diagnosis method of the SRv6 TE Policy network in the related art has a low degree of automation and low efficiency.

[0056] According to an intelligent diagnosis method for SRv6 network faults provided by the present invention, before obtaining a traffic feature vector between a head node and a tail node in a segment list set in a traffic engineering strategy mode in the SRv6 network, the method further includes: Determine the first timestamp of a single data packet passing through the head node in the segment list set in the traffic engineering policy mode in the SRv6 network; Determine a second timestamp of a single data packet passing through a tail node in a segment list set in a traffic engineering policy mode in an SRv6 network; The difference between the second timestamp and the first timestamp is used as the single list delay of a single data packet in the traffic engineering policy mode in the SRv6 network; The sum of the single list delays of all data packets within the target time is determined, and the ratio of the sum to the number of all data packets is taken as the end-to-end delay.

[0057] In the embodiment of the present invention, the network status and key performance indicators in the current SRv6 TE Policy mode are periodically obtained. The data acquisition source can be the head and tail nodes configured with SRv6 TE Policy in the SRv6 network, or offline data files. The data and definitions to be obtained also include: :Starting from the beginning To the end node A Segment List between , .

[0058] : Single list latency refers to the total latency experienced by a single data packet from the head node of an SRv6 TE Policy to its tail node.

[0059] There may be large differences between different lists in the same policy. The difference between the second timestamp and the first timestamp is used as the single list delay of a single data packet in the traffic engineering policy mode in the SRv6 network. The following formula (3) can be used as a reference: (3)

[0060] in, Represents a single data packet The single list delay of Is a single data packet The timestamp when passing through the SRv6 network tail node (i.e. the second timestamp mentioned above), It is the timestamp when the data packet passes through the SRv6 network head node (that is, the first timestamp mentioned above), which is obtained by capturing the packet through the network management system.

[0061] End-to-end delay is an important indicator that affects network communication quality and user experience. The end-to-end delay at each moment is the average value of the delays of all data packets in the second before that moment. The sum of the single list delays of all data packets at the target moment is determined, and the ratio of the sum to the number of all data packets is taken as the end-to-end delay. The following formula (4) can be used as a reference: (4)

[0062] in, Indicates the target time The end-to-end delay of Is a single data packet The total delay from the head node of the SRv6 TEPolicy link to its tail node, It is the total number of packets transmitted in the SRv6 network within one second.

[0063] According to an intelligent diagnosis method for SRv6 network faults provided by the present invention, after determining the sum of the single list delays of all data packets within the target time, and taking the ratio of the sum to the number of all data packets as the end-to-end delay, the method further includes: Determine the single list delay of each data packet among all data packets within the target time; The ratio of the sum of the squares of the difference between the single list delay of each data packet and the end-to-end delay to the number of all data packets is taken as the network delay jitter.

[0064] In the embodiment of the present invention, the network delay jitter It refers to the change or instability of end-to-end delay in network transmission. For details, please refer to the following formula (5): (5)

[0065] in, Indicates that at the target time The network delay jitter, is the total number of packets transmitted in the SRv6 network within one second, Is a single data packet The total delay from the head node of the SRv6 TE Policy link to its tail node, Indicates the target time end-to-end delay.

[0066] According to an intelligent diagnosis method for SRv6 network faults provided by the present invention, before obtaining a traffic feature vector between a head node and a tail node in a segment list set in a traffic engineering strategy mode in the SRv6 network, the method further includes: Determine the total number of packets sent by the head node in the segment list set in the traffic engineering policy mode in the SRv6 network within the target time; Determine the total number of received data packets of the tail node in the segment list set in the traffic engineering policy mode in the SRv6 network within the target time; The percentage of the ratio of the total number of received data packets to the total number of sent data packets is taken as the packet loss rate.

[0067] In the embodiment of the present invention, the packet loss rate It refers to the proportion of data packets lost due to various reasons during data transmission. For details, please refer to the following formula (6): (6)

[0068] in, Indicates that at the target time The packet loss rate, Refers to the number of data packets received by the egress node of the SRv6TE Policy within the time interval. is the number of packets sent by the head node.

[0069] According to an intelligent diagnosis method for SRv6 network faults provided by the present invention, before obtaining a traffic feature vector between a head node and a tail node in a segment list set in a traffic engineering strategy mode in the SRv6 network, the method further includes: Determine the end-to-end actual total bandwidth used by the target link in the traffic engineering policy mode in the SRv6 network at the target time; Determine the total end-to-end available bandwidth of the target link in the traffic engineering policy mode in the SRv6 network at the target time; The link bandwidth utilization is the percentage of the ratio of the end-to-end total bandwidth actually used to the end-to-end total available bandwidth.

[0070] In the embodiment of the present invention, the link bandwidth utilization It refers to the ratio between the actual bandwidth used and the available bandwidth, usually expressed as a percentage. Bandwidth utilization is an indicator to measure the current load level of a network link or device. For details, please refer to the following formula (7): (7)

[0071] in, The current period (i.e. the target time mentioned above), the link in the Policy (i.e. the target link mentioned above) actual bandwidth used, The current period link The end-to-end actual total bandwidth used, The current period link The bandwidth available, For the current period link The total end-to-end available bandwidth.

[0072] According to an intelligent diagnosis method for SRv6 network faults provided by the present invention, before inputting the traffic feature vector into a pre-trained data dimensionality reduction clustering model to obtain a plurality of network status labels output by the pre-trained data dimensionality reduction clustering model, the method further includes: Based on the training sample set, pre-train the data dimension reduction clustering model for the target number of iterations: According to the number of training samples in the training sample set, the number of neuron nodes in the competition layer of the data dimensionality reduction clustering model is set, and the node weights of the neuron nodes in the competition layer are initialized; The preprocessed training sample set is input into the data dimension reduction clustering model, and each neuron node in the competition layer is traversed to determine the similarity between the training sample and the neuron node; The neuron node with the highest similarity is regarded as the winning node; Determine the neighboring neuron nodes included in the winning neighborhood based on the winning node according to the preset domain radius; Update the node weights of the neighboring neuron nodes; When the current number of iterations is equal to the target number of iterations, the pre-trained data dimension reduction clustering model is obtained.

[0073] refer to Figure 4 , Figure 4 The historical KPI data training flow chart provided by the present invention specifically includes the following steps.

[0074] In an embodiment of the present invention, a dimensionality reduction clustering algorithm (model) based on multi-dimensional KPI data of the SRv6 network is proposed with the self-organizing map algorithm as the core and multi-dimensional KPI data as the driver. The specific pre-training process of the above-mentioned data dimensionality reduction clustering model can refer to the following steps.

[0075] Step 1: Set the number of neurons in the competitive layer according to the number of training samples. After the size of the competitive layer is determined, initialize the weights of each node in the competitive layer. , the dimension is the same as the feature vector dimension, and is usually assigned a small random value.

[0076] Step 2: Before model training, check the end-to-end latency , Network delay jitter , Packet loss rate and link bandwidth utilization The feature vector data of the four dimensions are preprocessed and can be standardized (mean is 0, standard deviation is 1), which helps to make each feature contribute equally to the calculation of similarity.

[0077] Step 3: After the data is processed, input the sample.

[0078] Step 4, traverse each node of the competition layer: calculate the similarity between the input sample and the neuron node (common ones include Euclidean distance, Manhattan distance, Minkowski distance, etc.). The embodiment of the present invention exemplarily uses Euclidean distance to calculate the similarity. , please refer to the following formula (8) for details.

[0079] (8)

[0080] in, Indicates The value of a KPI (key performance indicator); Indicates that at the target time For KPI, in a specific row , and specific columns The weight value below.

[0081] Step 5: Select the node with the smallest distance as the winning node (BMU, Best Matching Unit).

[0082] Step 6: Determine the nodes that the winning neighborhood will include based on the neighborhood radius σ (sigma); and calculate the amplitude of their respective updates through the neighborhood function (the basic idea of ​​the neighborhood function is: the closer to the winning node, the larger the update amplitude; the farther away from the winning node, the smaller the update amplitude). The specific neighborhood function can be referred to the following formula (9).

[0083] (9)

[0084] in, represents the neighborhood function, is the distance from the current node to the winning node, when Beyond the field radius Range A value of zero means that the node weight is not affected.

[0085] Step 7: Update the weights of nodes in the winning area , please refer to the following formula (10) for details.

[0086] 10)

[0087] in, Indicates that at the target time For KPI, in a specific row , and specific columns The weight value under is the learning rate, represents the neighborhood function, Indicates The value of a KPI.

[0088] In an embodiment of the present invention, is the learning rate, Affected by the distance from the current node to the winning node, both jointly affect the update amplitude of the weight.

[0089] Step 8: After completing one round of iteration, determine whether the set number of iterations is met. If not, return to step 3. If so, the historical KPI data training is completed and the data is divided into several categories. In the embodiment of the present invention, after the historical KPI data training is completed, experts are introduced to perform fault analysis on the classified data, and fault labels are attached to the neuron nodes in a supervised manner. After the new data generated is input into the system, it is mapped to the closest neuron, and the category label of the neuron is directly output to determine whether it is faulty and perform fault diagnosis. After diagnosing a certain amount of new data, the historical KPI data can be executed again to verify and update the system.

[0090] Through the above embodiments of the present invention, first, a fault diagnosis model is established based on end-to-end delay, delay jitter, packet loss rate, and bandwidth utilization; second, an unsupervised learning algorithm is used to reduce the dimension and cluster the pre-processed data, and finally experts are introduced to perform fault analysis on the classified data and supervised fault labels are attached. The closest category label is used to determine whether it is a fault and perform root cause analysis, which reduces the time for fault diagnosis and improves the reliability of the SRv6 network TEPolicy mode.

[0091] Among them, a fault intelligent diagnosis model was established based on SRv6 multi-dimensional KPI data, which can periodically diagnose the link status of the list in the policy. Based on the SRv6 network characteristics, comprehensive end-to-end delay, delay jitter, packet loss rate, bandwidth utilization and other factors, the SRv6 network status is regularly detected, the SRv6 TE Policy failure time is reduced, and the SRv6 TE Policy fault location efficiency is improved.

[0092] In the following, an intelligent diagnosis method for SRv6 network faults provided by an embodiment of the present invention is described in combination with actual application scenarios.

[0093] refer to Figure 5 , Figure 5 It is a schematic diagram of the structure of the SRv6 TE Policy network topology provided by the present invention.

[0094] In order to verify the effectiveness of the method proposed in this patent, a simulation was performed under a topology close to the existing network. Figure 4It is the topology diagram referenced by the simulation. Among them, the Policy has a total of 5 Segment Lists including 9 nodes. CE1 and CE2 (CustomerEdge) usually refer to customer edge devices, PE1 is the head node of the Policy, PE2 is the tail node of the Policy, and the failure probability of the intermediate nodes (P1, P2, P3, P4 and P5) is randomly generated within [0.001, 0.01]. The SegmentList information in the SRv6 network is shown in Table 1 (the service distribution in the network at a certain moment is shown in Table 1 below), which includes List number, path information and List weight.

[0095] Table 1

[0096] When the SRv6 TE Policy has five segment lists, the network status is as follows from the list perspective: In this case, in order to make the label more intuitive to display the network fault situation, this design uses the order of the fault list number as the label. The labels corresponding to various network status situations are shown in Table 2 below: Table 2

[0097] In general, the system periodically uses the Policy end-to-end performance indicators as the input of the diagnostic model, and the output is the label "0", indicating that the SRv6 network status is normal; assuming that the L4 and L5 links fail at a certain moment, the system inputs the data collected during this time period into the diagnostic model, and the output is the label "45", indicating that the L4 and L5 links fail. Based on this, the operation and maintenance personnel can further analyze that it may be that the public router P4 of the L4 and L5 links has failed, and realize intelligent diagnosis of SRv6 network failures.

[0098] The intelligent diagnosis device for SRv6 network failures provided by the present invention is described below. The intelligent diagnosis device for SRv6 network failures described below and the intelligent diagnosis method for SRv6 network failures described above can be referenced to each other.

[0099] refer to Figure 6 , Figure 6 It is a structural diagram of the intelligent diagnosis device for SRv6 network faults provided by the present invention, which includes an acquisition module 601 and an output module 602.

[0100] An acquisition module 601 is used to acquire a traffic feature vector between a head node and a tail node in a segment list set in a traffic engineering policy mode in an SRv6 network, wherein the traffic feature vector includes end-to-end delay, network delay jitter, packet loss rate, and link bandwidth utilization; The output module 602 is used to input the traffic feature vector into the pre-trained data dimensionality reduction clustering model to obtain multiple network status labels output by the pre-trained data dimensionality reduction clustering model, wherein each of the multiple network status labels is used to represent the network status of each traffic engineering strategy mode.

[0101] Specifically, the above-mentioned intelligent diagnosis device for SRv6 network faults provided by the present invention can implement all the method steps implemented by the above-mentioned intelligent diagnosis method embodiment for SRv6 network faults, and can achieve the same technical effect. The parts and beneficial effects that are the same as the method embodiment in this embodiment will not be described in detail here.

[0102] Figure 7 is a schematic diagram of the physical structure of the electronic device provided by the present invention, such as Figure 7 As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communication interface 720 and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute the intelligent diagnosis method of the SRv6 network fault, the method comprising: obtaining the traffic feature vector between the head node and the tail node in the segment list set under the traffic engineering strategy mode in the SRv6 network, wherein the traffic feature vector includes end-to-end delay, network delay jitter, packet loss rate and link bandwidth utilization; inputting the traffic feature vector into the pre-trained data dimensionality reduction clustering model to obtain multiple network status labels output by the pre-trained data dimensionality reduction clustering model, wherein each network status label in the multiple network status labels is used to represent the network status of each traffic engineering strategy mode.

[0103] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0104] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the intelligent diagnosis method of SRv6 network faults provided by the above methods, and the method includes: obtaining a traffic feature vector between a head node and a tail node in a segment list set under a traffic engineering strategy mode in the SRv6 network, wherein the traffic feature vector includes end-to-end delay, network delay jitter, packet loss rate, and link bandwidth utilization; inputting the traffic feature vector into a pre-trained data dimensionality reduction clustering model to obtain multiple network status labels output by the pre-trained data dimensionality reduction clustering model, wherein each of the multiple network status labels is used to represent the network status of each traffic engineering strategy mode.

[0105] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the intelligent diagnosis method for SRv6 network faults provided by the above-mentioned methods, the method comprising: obtaining a traffic feature vector between a head node and a tail node in a segment list set under a traffic engineering strategy mode in the SRv6 network, wherein the traffic feature vector comprises end-to-end delay, network delay jitter, packet loss rate, and link bandwidth utilization; inputting the traffic feature vector into a pre-trained data dimensionality reduction clustering model to obtain a plurality of network status labels output by the pre-trained data dimensionality reduction clustering model, wherein each of the plurality of network status labels is used to represent the network status of each traffic engineering strategy mode.

[0106] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0107] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0108] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent diagnosis method for SRv6 network failures, characterized in that: include: Obtain a traffic feature vector between a head node and a tail node in a segment list set in a traffic engineering policy mode in an SRv6 network, wherein the traffic feature vector includes end-to-end delay, network delay jitter, packet loss rate, and link bandwidth utilization; The traffic feature vector is input into a pre-trained data dimensionality reduction clustering model to obtain multiple network status labels output by the pre-trained data dimensionality reduction clustering model, wherein each of the multiple network status labels is used to represent the network status of each of the traffic engineering strategy modes.

2. The intelligent diagnosis method for SRv6 network failure according to claim 1 is characterized in that: Before obtaining the traffic feature vector between the head node and the tail node in the segment list set in the traffic engineering policy mode in the SRv6 network, the method further includes: Determine the first timestamp of a single data packet passing through the head node in the segment list set in the traffic engineering policy mode in the SRv6 network; Determine a second timestamp of the single data packet passing through the tail node in the segment list set in the traffic engineering policy mode in the SRv6 network; Using the difference between the second timestamp and the first timestamp as the single list delay of the single data packet in the traffic engineering policy mode in the SRv6 network; The sum of the single list delays of all data packets within the target time is determined, and the ratio of the sum to the number of all data packets is used as the end-to-end delay.

3. The intelligent diagnosis method for SRv6 network failure according to claim 2 is characterized in that: After determining the sum of the single list delays of all data packets within the target time, and taking the ratio of the sum to the number of all data packets as the end-to-end delay, the method further includes: Determine a single list delay of each data packet in all the data packets within the target time; The ratio of the sum of the squares of the difference between the single list delay of each data packet and the end-to-end delay to the number of all data packets is taken as the network delay jitter.

4. The intelligent diagnosis method for SRv6 network failure according to claim 1, characterized in that: Before obtaining the traffic feature vector between the head node and the tail node in the segment list set in the traffic engineering policy mode in the SRv6 network, the method further includes: Determine the total number of packets sent by the head node in the segment list set in the traffic engineering policy mode in the SRv6 network within the target time; Determine the total number of received data packets of the tail node in the segment list set in the traffic engineering policy mode in the SRv6 network within the target time; The percentage of the ratio of the total number of received data packets to the total number of sent data packets is taken as the packet loss rate.

5. The intelligent diagnosis method for SRv6 network failure according to claim 1, characterized in that: Before obtaining the traffic feature vector between the head node and the tail node in the segment list set in the traffic engineering policy mode in the SRv6 network, the method further includes: Determine the end-to-end actual total bandwidth used by the target link in the traffic engineering policy mode in the SRv6 network at the target time; Determine the end-to-end total available bandwidth of the target link in the traffic engineering policy mode in the SRv6 network at the target time; The percentage of the ratio of the end-to-end actually used total bandwidth to the end-to-end available total bandwidth is taken as the link bandwidth utilization.

6. The intelligent diagnosis method for SRv6 network failure according to claim 1, characterized in that: Before inputting the traffic feature vector into a pre-trained data dimensionality reduction clustering model to obtain a plurality of network status labels output by the pre-trained data dimensionality reduction clustering model, the method further includes: Based on the training sample set, pre-train the data dimension reduction clustering model for the target number of iterations: According to the number of training samples in the training sample set, the number of neuron nodes in the competition layer of the data dimensionality reduction clustering model is set, and the node weights of the neuron nodes in the competition layer are initialized; Inputting the preprocessed training sample set into the data dimension reduction clustering model, traversing each of the neuron nodes in the competition layer to determine the similarity between the training sample and the neuron node; The neuron node with the highest similarity is taken as the winning node; Determine the neighborhood neuron nodes included in the winning neighborhood based on the winning node according to a preset domain radius; Updating the node weights of the neighborhood neuron nodes; When the current number of iterations is equal to the target number of iterations, a pre-trained data dimension reduction clustering model is obtained.

7. An intelligent diagnostic device for SRv6 network failure, characterized in that: include: An acquisition module is used to obtain a traffic feature vector between a head node and a tail node in a segment list set in a traffic engineering policy mode in an SRv6 network, wherein the traffic feature vector includes end-to-end delay, network delay jitter, packet loss rate, and link bandwidth utilization; An output module is used to input the traffic feature vector into a pre-trained data dimensionality reduction clustering model to obtain multiple network status labels output by the pre-trained data dimensionality reduction clustering model, wherein each of the multiple network status labels is used to represent the network status of each of the traffic engineering strategy modes.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the intelligent diagnosis method for SRv6 network failure according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the intelligent diagnosis method for SRv6 network failures as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the intelligent diagnosis method for SRv6 network failures as described in any one of claims 1 to 6 is implemented.

Images