Network cluster data acquisition method and device

Abstract

The embodiments of this specification provide a network cluster data acquisition method and device, wherein the network cluster data acquisition method includes: acquiring a question text, wherein the question text is for at least one network cluster; processing the question text based on a target data processing model to generate an answer text, wherein the answer text includes an answer text and a step analysis text, wherein the answer text is generated by the target data processing model based on the step analysis text, and the step analysis text is generated by the target data processing model according to the question text and the network cluster data corresponding to each network cluster. This solution is based on data acquisition and processing of the target data processing model, which solves the problems of complex operation and low efficiency of traditional methods, thereby realizing flexible response of multiple dimensions and multiple indicators, and significantly improving the speed and accuracy of data acquisition.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of natural language processing technology, and in particular to a method for acquiring network cluster data. Background Art

[0002] With the rapid development of cloud network technology, time series indicator data analysis plays an important role in network operation and maintenance. Especially when analyzing network cluster data, this type of scenario is very common in daily operation and maintenance, and can help make analysis decisions quickly. For example, by analyzing the performance of devices in different directions and indicators in the cluster, you can find out the devices that have performed well in the recent period of time, providing data support for operation and maintenance and optimization.

[0003] At present, the traditional network cluster data acquisition method mainly relies on the SQL query capability of the database, which can solve the problem of obtaining certain network cluster data within a specific time period. However, this method has the problems of complex operation and low efficiency, especially in the cross-acquisition and analysis of multiple dimensions and multiple indicators, the flexibility and response speed of SQL query are limited. Therefore, in order to solve these shortcomings, a network cluster data acquisition method is proposed. Summary of the invention

[0004] In view of this, an embodiment of this specification provides a network cluster data acquisition method. One or more embodiments of this specification also relate to a network cluster data acquisition device, a computing device, a computer-readable storage medium and a computer program product to solve the technical defects existing in the prior art.

[0005] According to a first aspect of an embodiment of this specification, a network cluster data acquisition method is provided, including:

[0006] Obtaining a question text, wherein the question text is for at least one network cluster;

[0007] The question text is processed based on the target data processing model to generate an answer text, wherein the answer text includes an answer text and a step analysis text, the answer text is generated by the target data processing model based on the step analysis text, and the step analysis text is generated by the target data processing model according to the question text and the network cluster data corresponding to each network cluster.

[0008] According to a second aspect of an embodiment of this specification, a network cluster data acquisition device is provided, including:

[0009] An acquisition module is configured to acquire a question text, wherein the question text is for at least one network cluster;

[0010] A generation module is configured to process the question text based on a target data processing model to generate an answer text, wherein the answer text includes an answer text and a step analysis text, the answer text is generated by the target data processing model based on the step analysis text, and the step analysis text is generated by the target data processing model according to the question text and the network cluster data corresponding to each network cluster.

[0011] According to a third aspect of an embodiment of this specification, a computing device is provided, including:

[0012] Memory and processor;

[0013] The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the above-mentioned network cluster data acquisition method are implemented.

[0014] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores computer-executable instructions, and when the instructions are executed by a processor, the steps of the above-mentioned network cluster data acquisition method are implemented.

[0015] According to a fifth aspect of the embodiments of this specification, a computer program product is provided, including a computer program / instruction, which implements the steps of the above-mentioned network cluster data acquisition method when executed by a processor.

[0016] An embodiment of the present specification implements: obtaining a question text, wherein the question text is for at least one network cluster; processing the question text based on a target data processing model to generate an answer text, wherein the target data processing model is trained based on network cluster data feature information, the network cluster data feature information includes at least one network cluster indicator feature information, the network cluster indicator feature information represents the actual meaning corresponding to the indicator data in the network cluster data; and determining the target answer information corresponding to the question text based on the answer text.

[0017] The solution of the embodiment of this specification is applied to solve the operational complexity and inefficiency of traditional reliance on database queries through a new network cluster data acquisition process. Especially when it comes to cross-analysis of multiple dimensions and indicators, traditional manual query methods are often limited by flexibility and response speed. The new process not only simplifies the operation, but also greatly improves the processing speed, and can respond quickly and generate accurate results. This automated data analysis method adapts to more complex network environments, ensuring that while efficiently acquiring data, the flexibility and adaptability of the system are maintained. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1is a flow chart of a network cluster data acquisition method provided by an embodiment of this specification;

[0019] Figure 2 It is an architecture diagram of a network cluster data acquisition system provided by an embodiment of this specification;

[0020] Figure 3 It is a process flow chart of a method for obtaining the top N device information of a network cluster indicator provided by an embodiment of this specification;

[0021] Figure 4 It is a structural diagram of a network cluster data acquisition device provided by an embodiment of this specification;

[0022] Figure 5 It is a structural block diagram of a computing device provided by an embodiment of this specification. DETAILED DESCRIPTION

[0023] Many specific details are described in the following description to facilitate a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of this specification, so this specification is not limited to the specific implementation disclosed below.

[0024] The terms used in one or more embodiments of this specification are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of this specification. The singular forms of "a", "said" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0025] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0026] In addition, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0027] In one or more embodiments of this specification, a large model refers to a deep learning model with large-scale model parameters, which usually contains hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than 10 trillion model parameters. A large model can also be called a foundation model / foundation model. The large model is pre-trained with large-scale unlabeled corpus to produce a pre-trained model with more than 100 million parameters. This model can adapt to a wide range of downstream tasks, and the model has good generalization ability, such as a large-scale language model (LLM), a multi-modal pre-training model, etc.

[0028] When the big model is used in practice, only a small number of samples are needed to fine-tune the pre-trained model and it can be applied to different tasks. The big model can be widely used in natural language processing (NLP), computer vision and other fields. Specifically, it can be applied to computer vision tasks such as visual question answering (VQA), image caption (IC), image generation, as well as natural language processing tasks such as text-based sentiment classification, text summary generation, and machine translation. The main application scenarios of the big model include digital assistants, intelligent robots, search, online education, office software, e-commerce, intelligent design, etc.

[0029] First, the terms involved in one or more embodiments of this specification are explained.

[0030] Structured Query Language (SQL): is a standardized language used to perform query, insert, update, delete and other operations in relational database management systems. SQL is mainly used to process structured data, that is, data sets with clear patterns. Through SQL, users can accurately query the data in the table according to specific grammatical structures, and implement complex data filtering and classification. SQL queries can extract the required information from the database according to the conditions provided by the user, and return the result set that meets these conditions. It is widely used in data analysis, system management and application development.

[0031] SQL Query: is a method of sending instructions to the database through specific syntax and returning results. SQL query statements usually include operations such as selection (SELECT), filtering (WHERE), sorting (ORDER BY), etc., which are used to extract data that meets the requirements from the table. SQL query can handle the association query between a single table or multiple tables, and filter data by combining multiple conditions.

[0032] Cloud Data Transfer (CDT): is a service for efficient and secure transmission of large-scale data between different cloud environments or data centers. CDT can support cross-regional and cross-network data transmission needs, ensuring stable data exchange between different data nodes with low latency. This service is suitable for a variety of scenarios, such as data backup, distributed data processing, and cross-regional data synchronization. By optimizing the transmission path and bandwidth usage, it improves the overall performance and reliability of data transmission.

[0033] Cloud Enterprise Network (CEN): is an architecture used to connect different network environments, helping enterprises establish global private network connections. CEN can support seamless connections across different regions and network environments, and optimize the transmission path of network traffic through intelligent routing mechanisms to ensure fast and secure data transmission. This network architecture is often used in scenarios such as enterprise data center interconnection and network interconnection of multiple branches, providing a flexible and efficient network connection solution.

[0034] Data Transfer Plan (DTP): is a bandwidth or traffic service for sharing multiple network resources. It is designed to help users allocate bandwidth resources between multiple network nodes or clusters. DTP can dynamically adjust bandwidth usage according to the needs of different network clusters to ensure efficient allocation of network traffic. This service is widely used in bandwidth sharing, traffic optimization and other scenarios in multiple network environments. It can help users flexibly manage network resources and avoid bandwidth waste or overload.

[0035] In this specification, a network cluster data acquisition method is provided. This specification also relates to a network cluster data acquisition device, a computing device, a computer-readable storage medium, and a computer program product, which are described in detail one by one in the following embodiments.

[0036] With the rapid development of cloud network technology, time series indicator data analysis has become an indispensable tool in network operation and maintenance. Time series data helps managers monitor system performance in real time and detect potential problems in a timely manner by continuously recording changes in various indicators in the network. In network cluster management, time series indicators can conduct in-depth analysis of multiple key parameters such as network traffic, device performance, latency, and error rate to ensure the normal operation of the system. For example, in the daily operation and maintenance of the data center, by analyzing device indicators in different directions (such as data inflow and outflow), devices with abnormal performance or excessive load can be quickly identified, thereby effectively allocating resources and optimizing overall network performance. These analysis results not only provide a reliable decision-making basis for daily operation and maintenance, but also predict potential network bottlenecks or failure points for early prevention.

[0037] However, the data acquisition and analysis of many current network clusters rely on traditional SQL (Structured Query Language) queries. Although SQL queries perform well when processing static data, their limitations gradually become apparent when faced with complex multi-dimensional, multi-indicator data cross-analysis. Especially in a network environment, the cluster contains a large amount of real-time dynamically changing data, and SQL queries are often difficult to respond quickly, resulting in system processing delays and low query efficiency. In addition, as the scale of network clusters continues to expand, manually writing SQL queries to cope with the complexity of multi-dimensional analysis also increases the operational burden. This method not only lacks flexibility, but also easily leads to untimely analysis results when dealing with complex scenarios, which in turn affects decision-making efficiency.

[0038] See also Figure 1 , Figure 1 A flow chart of a network cluster data acquisition method provided according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0039] Step 102: Obtain question text, wherein the question text is for at least one network cluster.

[0040] In practical applications, the question text is the query or data acquisition requirement raised by the user, which can include a specific time range, indicator name or device information. A network cluster is a collection of multiple computing or network resources that work together to process or manage data. These clusters can be systems composed of physical servers, virtual machines or cloud computing nodes.

[0041] Exemplarily, the question text can be understood as a query or request entered by a user when he wants to obtain specific data from the system or solve a specific problem. It appears in natural language and involves the acquisition, analysis or operation of certain data. For example, a user may ask a question text like this: "What are the top two devices with the largest number of active sessions and input bytes per second in network cluster A in the past 10 minutes?" The question text not only specifies the specific indicators of the data (the number of active sessions and the number of input bytes per second), but also limits the time range of the query (the past 10 minutes), with the goal of finding devices that meet the conditions. The question text is the core input for analysis and processing in the automation system. The system determines the data range and query target to be processed by parsing the text.

[0042] A network cluster is a resource collection consisting of multiple nodes, usually used for high-performance computing, data storage or network management. The nodes in each cluster can be physical servers, virtual machines, or even containerized applications, which work together to process large amounts of data or tasks. For example, in a cloud computing platform, a network cluster is usually composed of multiple distributed computing nodes to handle tasks such as high-concurrency access requests, data analysis, or machine learning model training. A typical application scenario of a network cluster is a content distribution network (CDN), in which multiple server nodes are distributed around the world to improve data access speed and reliability. Through distributed computing, network clusters can significantly improve data processing capabilities and system stability.

[0043] It should be noted that obtaining question text can be understood as the process of extracting or generating data query instructions that can be processed by the system from user input. The specific method of obtaining question text can be through natural language questions directly input by the user, such as "What is the number of active sessions in the network cluster in the past 10 minutes?"; question text can also be generated by an automated system, dynamically generating text based on the user's operation history or predefined query templates; and voice recognition technology can also be used to convert the user's voice instructions into processable text questions for subsequent processing by the system, etc. This manual does not impose any restrictions on this.

[0044] In one embodiment provided in this specification, the question text is "What are the top two device IPs with larger traffic in clusters S1 and S2 in the past 10 minutes?", which is for cluster 1 corresponding to cluster C1 and cluster 2 corresponding to cluster C2.

[0045] By obtaining the question text, it is possible to accurately understand the user's query requirements by receiving the natural language written by the user, and automatically parse the key elements therein, thereby generating efficient and targeted data query tasks. This not only simplifies the interaction process between the user and the system, reduces the difficulty of manually entering complex queries, but also improves the system's response speed and processing capabilities.

[0046] Further, obtain the question text, including:

[0047] Receiving question request information input by a user, wherein the question request information includes a question request text;

[0048] Inputting the request question text into a question judgment model to obtain a question judgment result generated by the question judgment model;

[0049] When the model judgment result is related to the network cluster, the request question text is determined to be a question text.

[0050] In actual applications, the question request information is the entire content of the query or operation request submitted by the user to the system, including the specific question text of the query. The request question text is the core content of the question request information, that is, the question actually asked by the user. The question judgment model is a model used by the system to analyze and identify the type of question, and determine whether the question is related to the network cluster. The question judgment result is the output after the problem judgment model analysis, which clarifies the classification or characteristics of the problem for further processing by the system.

[0051] Question request information can be understood as a complete information package transmitted when a user interacts with the system, which includes the specific problem that the user wants the system to solve and related auxiliary information. For example, in a network monitoring scenario, a user may submit a question request information, including "What is the traffic situation of the S1 cluster in the last 10 minutes?" This question request information not only contains the query text, but also may include parameters such as priority and time range, providing a complete background for subsequent processing by the system.

[0052] The request question text can be understood as the main question content in the question request information, that is, the query target directly expressed by the user. The request question text is the core part of the question received by the system and determines the actual content to be processed by the system. For example, the "CPU usage of the S1 cluster in the past 10 minutes" submitted by the user is the request question text. The system will judge and analyze this text to determine the required processing steps.

[0053] The problem judgment model can be understood as a machine learning or rule model that the system uses to analyze the request question text and identify its type. The model determines whether the request is related to the network cluster or whether it involves other specific topics based on pre-trained data. For example, when the system receives a question text "Abnormal traffic at a certain network node", the problem judgment model can identify that the problem is related to network cluster monitoring and use this result for further processing and analysis.

[0054] The problem judgment result can be understood as the result generated by the problem judgment model after analyzing the request question text. This result is used to determine the type or topic of the question and provide guidance for subsequent steps. For example, when the problem judgment model analyzes that the question text is related to network cluster performance, the problem judgment result will indicate "network cluster" as the classification of the question, allowing the system to use a dedicated network cluster analysis module to further process the request.

[0055] It should be noted that when the model determines that the result is irrelevant to the network cluster, since the request question text input by the user is irrelevant to the network cluster, the request question text can be directly input into the large model to obtain the answer of the large model.

[0056] The problem judgment model determines whether the question text entered by the user is related to the network cluster, and determines whether to perform data processing related to the network cluster based on the judgment result, which can effectively improve the processing efficiency and response accuracy of the system. It not only avoids the waste of resources on irrelevant questions, but also ensures that the system can focus on processing queries related to the network cluster, improving the pertinence and execution efficiency of problem analysis.

[0057] Step 104: Process the question text based on the target data processing model to generate an answer text, wherein the answer text includes an answer text and a step analysis text, the answer text is generated by the target data processing model based on the step analysis text, and the step analysis text is generated by the target data processing model according to the question text and the network cluster data corresponding to each network cluster.

[0058] In practical applications, the target data processing model is a model trained based on the network cluster data feature information, which is used to process input data and generate corresponding output. The answer text is a response generated based on the input question and the model processing results, providing specific analysis steps and conclusions. The network cluster data feature information is a key parameter used to describe the overall performance or status of the network cluster, including at least one network cluster indicator feature information. The network cluster indicator feature information is specific indicator data, which characterizes the actual meaning of the indicator data in the network cluster data and reflects the specific performance of different nodes or devices in the network cluster.

[0059] The target data processing model can be understood as a processing tool trained with a large amount of historical data, which is used to parse user input and output results that meet the requirements. In network operation and maintenance, the target data processing model can analyze information such as time series data and performance indicators in the network cluster, and provide reasonable analysis and decision-making based on the user's questions. For example, the model can cross-process multi-dimensional data, identify bottleneck devices or high-load nodes in the cluster, and automatically generate optimization suggestions or problem answers. This model has a wide range of applications in modern big data and automated operation and maintenance scenarios.

[0060] The answer text can be understood as a detailed description document generated based on the questions entered by the user and the analysis of the target data processing model. The answer text not only provides the results, but also shows the entire data analysis process, allowing users to understand how the data is processed and to easily obtain the answer information later.

[0061] The network cluster data feature information can be understood as a data set that describes the overall operation status of the network cluster. It contains a variety of key indicators for monitoring cluster performance. These data feature information include network traffic, device latency, error rate, etc., and play a fundamental role in the system analysis and optimization process. For example, in a cloud computing environment, network cluster data feature information can help operation and maintenance personnel understand the load of each node in the cluster in real time, ensure the smooth operation of the system, and provide data support for optimizing resource allocation.

[0062] The network cluster indicator characteristic information can be understood as the performance indicator data of specific devices or nodes in the network cluster, which is used to evaluate the specific performance of the network cluster. For example, network inflow / outflow bandwidth, number of bytes transmitted per second, packet loss rate, etc. are all typical manifestations of network cluster indicator characteristic information. These indicators provide a quantitative basis for the specific behavior of each device in the cluster, so that the system can identify nodes or devices with abnormal performance and make corresponding adjustments. This type of information is widely used in network monitoring, traffic optimization and troubleshooting to ensure the efficient operation of the cluster as a whole.

[0063] In an embodiment provided in this specification, the question text input by the user follows the above example, and this specification will not repeat it. The IP addresses of two machines with large traffic in network cluster S1 and network cluster S2 are found, and the execution status of each step is listed in detail. The answer text output by the target data processing model is:

[0064] "Based on the data you provided, we need to find the two IP addresses with the largest sum of the output bytes per second (OutBps) and the input bytes per second (InBps) in each cluster. Since we cannot obtain the traffic directly, we will calculate and find the target device through the following steps:

[0065] ## First, we extract the OutBps and InBps data of each device from the S1 cluster and S2 cluster respectively:

[0066] S1 cluster device data:

[0067] Device 1: OutBps = 1000, InBps = 500;

[0068] Device 2: OutBps = 2000, InBps = 1000;

[0069] Device 3: OutBps = 1500, InBps = 800.

[0070] S2 cluster device data:

[0071] Device 4: OutBps = 3000, InBps = 1500;

[0072] Device 5: OutBps = 4000, InBps = 2500;

[0073] Device 6: OutBps = 1000, InBps = 300.

[0074] ## Next, we calculate the sum of OutBps and InBps for each device:

[0075] S1 cluster traffic calculation results:

[0076] Device 1 flow = 1000 + 500 = 1500;

[0077] Device 2 traffic = 2000 + 1000 = 3000;

[0078] Device 3 traffic = 1500 + 800 = 1300.

[0079] S2 cluster traffic calculation results:

[0080] Device 4 flow = 3000 + 1500 = 4500;

[0081] Device 5 flow = 4000 + 2500 = 6500;

[0082] Device 6 flow = 1000 + 300 = 1300.

[0083] ## Then, we sort the traffic in each cluster and find the top two devices with the most traffic:

[0084] S1 cluster sorting results:

[0085] Device 2: flow = 3000;

[0086] Device 1: Flow rate = 1500.

[0087] S2 cluster sorting results:

[0088] Device 4: flow = 4500;

[0089] Device 5: Flow rate = 6500.

[0090] ##result:

[0091] The two machine IPs with the largest traffic in the current S1 cluster are:

[0092] Device 2 IP: IP2;

[0093] Device 1 IP: IP1.

[0094] The two machine IPs with the largest traffic in the current S2 cluster are:

[0095] Device 4 IP: IP4;

[0096] Device 5 IP: IP5.

[0097] I hope this analysis was helpful! Please let me know if further information or additional analysis is required.”

[0098] Since the target data processing model is trained based on the network cluster data feature information including the network cluster indicator feature information, the target data processing model can identify the natural language description corresponding to the ID or name of the actual indicator to meet the user's needs. Compared with using SQL statements to meet user needs, it can more intuitively parse complex queries, reduce the burden of users manually writing query statements, and greatly improve the flexibility and efficiency of queries. By directly using natural language input, users can obtain multi-dimensional data analysis results more quickly, especially when dealing with multi-indicator and multi-dimensional cross-analysis. The system can automatically identify relevant data and generate detailed analysis steps and results, avoiding the cumbersome operation and inefficiency of SQL statements in complex data queries, and providing a more intelligent and convenient user experience.

[0099] Furthermore, the target data processing model is obtained by training the following steps, including:

[0100] Acquire network cluster data feature information and an initial data processing model, wherein the network cluster data feature information includes at least one network cluster indicator feature information and network cluster question and answer information;

[0101] The initial data processing model is trained based on the network cluster data feature information to obtain a target data processing model.

[0102] In practical applications, the initial data processing model is a model that is pre-trained or configured when there is no specific task; the network cluster question and answer information is a related data structure of historical queries and feedback in the network cluster, which helps the system understand and generate answers that meet user needs.

[0103] Exemplarily, the initial data processing model can be understood as a basic model that has not been trained in a targeted manner, a large model that has the ability to process general data but has not been optimized for a specific field or task, or a randomly initialized model, etc. This specification does not impose any restrictions on this. Network cluster question and answer information can be understood as user interaction data generated in the direction of the network cluster, including query text entered by the user and response information generated by the system based on these queries. For example, when a user asks the number of active devices in a network cluster, the network cluster question and answer information will include the user's natural language query and relevant information such as device status and performance data extracted from the cluster by the system. This type of question and answer information is not only a record of the query results, but can also be used by the system to further optimize answers to similar questions. By accumulating a large amount of question and answer information, a better target data processing model can be trained through the accumulated large amount of question and answer information, so that the target data processing model can better understand the needs of the user, thereby improving the responsiveness and accuracy of the model.

[0104] It should be noted that training the initial data processing model based on the network cluster data feature information and obtaining the target data processing model can be understood as further optimizing the initial model through the network cluster data feature information so that it can better adapt to specific tasks for the network cluster. Specifically, the target data processing model can be obtained by setting a fine-tuning layer and adjusting the parameters of the fine-tuning layer using the network cluster data feature information; the parameters of the initial model can also be adjusted by cross-validating the initial model using the network cluster data feature information to improve the performance of the model in different scenarios; and weighted training can also be performed on specific network cluster indicators in the network cluster data feature information so that the model can make more accurate predictions and analyses of key performance indicators, etc. This manual does not impose any restrictions on this.

[0105] In one embodiment provided in this specification, the network cluster data feature information includes:

[0106] Network Cluster:

[0107] cluster1: {

[0108] name: S1 cluster

[0109] Abstract: Cloud Data Transfer (CDT) is used for efficient and secure data transmission between different cloud environments or data centers. S1 cluster is suitable for cross-region data synchronization, data backup and distributed data processing scenarios. It achieves stable and low-latency data exchange by optimizing transmission paths and bandwidth usage.

[0110] }, cluster2: {

[0111] name: S2 cluster

[0112] abstract: Cloud Enterprise Network (CEN) is used to connect different network environments and realize cross-regional and cross-network dedicated network interconnection. S2 cluster can optimize data transmission paths through intelligent routing, which is suitable for enterprise data center interconnection and network interconnection between branches in multiple locations.

[0113] }, cluster3: {

[0114] name: S3 cluster

[0115] Abstract: Shared traffic package DTP (Data Transfer Plan) is used to dynamically allocate bandwidth resources between multiple network nodes or clusters. S3 cluster is suitable for scenarios that require flexible management of network bandwidth to ensure efficient distribution of network traffic and avoid bandwidth waste or overload.

[0116] },…]; Network cluster indicators: [

[0117] in_bps: Inbound bps, the number of bits transmitted per second in the inbound direction, often used to express the inbound bandwidth of the cloud network.

[0118] out_bps: outbound bps, bits transmitted per second in the outbound direction, often used to express the outbound bandwidth of the cloud network.

[0119] out_drop_bps: outbound dropped bps, bits dropped per second in the outbound direction, often used to indicate packet loss in the outbound direction of the cloud network. This is an abnormal situation.

[0120] in_slowpath_bps: Inbound slow path bps, the number of bits transmitted per second on the slow path in the inbound direction. It is used to express the bandwidth of the cloud network in the inbound direction on the slow path. The performance is lower than that of the fast path.

[0121] out_slowpath_bps: Incoming slow path bps, outgoing direction, bits transmitted per second on the slow path, used to express the bandwidth of the cloud network in the outgoing direction on the slow path, the performance is lower than that of the fast path.

[0122] …]; Question and Answer: [

[0123] Question 1:

[0124] Q: What are the top two device IPs with the highest OutBps and ActiveSessionCount in network cluster XX-cluster-2 in the past 15 minutes?

[0125] A: We first obtain the OutBps (output bytes per second) and ActiveSessionCount (active session count) data of the XX-cluster-2 cluster in the last 15 minutes, sort them and find the top two devices. The sorting results show that the IP addresses of devices with higher OutBps are 123.456.789.0 and 123.456.789.1, and the IP addresses of devices with higher ActiveSessionCount are 123.456.789.2 and 123.456.789.3. By comparing these two indicators, we found that OutBps and ActiveSessionCount are not high at the same time, which means that although some devices have a large number of sessions, their data output is not necessarily high.

[0126] Q: So, what are the device IPs with high OutBps and ActiveSessionCount?

[0127] A: Next, we analyze the devices with both high OutBps and ActiveSessionCount. By screening the devices with both high OutBps and ActiveSessionCount, we find that the IP addresses of the devices with high OutBps and ActiveSessionCount are 123.456.789.0 and 123.456.789.2. These two devices not only have a large number of sessions, but also a high data output volume, and may be key nodes in the system.

[0128] }, Question 2{

[0129] Question: What are the IP addresses of the top two devices with the largest traffic in network cluster XX-cluster-3 in the past 10 minutes?

[0130] A: We obtained the InBps and OutBps data from XX-cluster-3 for the last 10 minutes, and then calculated the traffic of each device (OutBps + InBps). The results show that the IP addresses of devices with larger traffic are 123.456.789.4 and 123.456.789.5, which indicates that these devices have higher data throughput.

[0131] Q: Do these devices with higher traffic also have higher numbers of active sessions?

[0132] A: By further analyzing the ActiveSessionCount of these devices, we found that the number of active sessions of device 123.456.789.4 is also at a high level, indicating that this device is not only responsible for a large amount of data transmission, but also has a high number of concurrent sessions. The number of active sessions of 123.456.789.5 is relatively low, which means that its data transmission volume mainly comes from fewer sessions, which may be related to its task type.

[0133] },…], wherein the data in the question and answer: [...] is the network cluster question and answer information, and the data in the network cluster indicator: [...] is the network cluster indicator characteristic information. In this embodiment, the information corresponding to the description of each network cluster is also included, which is the data in the network cluster: [...].

[0134] By training the initial data processing model with network cluster data feature information including network cluster question and answer information and network cluster indicator feature information, the target data processing model obtained can more accurately understand and process complex network cluster data query requirements. Since the training process combines actual question and answer information and various indicators in the network cluster, the model can better capture the relationship between network operation status and user needs, so as to quickly generate accurate answers when facing queries in different scenarios.

[0135] Further, training the initial data processing model based on the network cluster data feature information to obtain a target data processing model includes:

[0136] Obtaining an initial model training layer and an initial model training layer, wherein the initial model training layer is connected to a self-attention layer in the initial data processing model;

[0137] Adjust the parameters of the initial model training layer based on the network cluster data feature information to obtain a target model training layer;

[0138] The target model training layer and the initial data processing model are combined to obtain a target data processing model.

[0139] In practical applications, the initial model training layer is a structured layer used for preliminary training of the data processing model, which is responsible for improving the basic performance of the model; the self-attention layer is the neural network layer in the initial data processing model; the target model training layer is a model layer that is further optimized after the initial training is completed, which can accurately handle complex task requirements.

[0140] Exemplarily, the initial model training layer can be understood as a structured layer used by the model to capture the overall characteristics of the data at the beginning of training, which is usually represented in the form of a matrix. The main function of this layer is to help the model learn the basic patterns of the data, such as the performance indicators of devices in the network cluster and user behavior data. In network cluster data processing, the initial model training layer will identify and process the broad characteristics of the data.

[0141] Since the self-attention layer of the initial data processing model focuses on extracting more relevant information from the input data when processing data, rather than processing all data dimensions, it is preferred to make the rank of the self-attention layer smaller than the rank of the initial data processing model so that the self-attention layer can capture important features more efficiently and reduce computational complexity. This structure not only reduces the resource consumption of the model, but also improves the performance of the model in specific tasks, especially when processing high-dimensional, noisy data. A lower rank means that the model can focus on key data dimensions, thereby reducing the interference of redundant information, allowing the model to more accurately analyze core indicators in network cluster data processing, and improve the efficiency and accuracy of the overall analysis.

[0142] The target model training layer can be understood as a high-performance layer formed by further optimization after the model completes the initial training. Unlike the initial model training layer, the target model training layer is more carefully trained and can handle more complex and specific tasks. For example, in a network cluster, the target model training layer can not only identify basic performance issues of the device, but also cross-analyze multi-dimensional performance indicators to find out the key factors that affect the overall efficiency of the network. This layer makes the model more accurate and adaptable, and can provide more insightful analysis results in complex network environments.

[0143] It should be noted that the initial data processing model is a neural network model including a self-attention layer, and the target data processing model is a neural network model with a target model training layer connected to the self-attention layer.

[0144] Adjusting the parameters of the initial model training layer based on the network cluster data feature information to obtain the target model training layer can be understood as adjusting the parameters of the initial training model to make the model more suitable for specific network cluster data processing tasks. The specific method can be to freeze the parameters of the initial model and only adjust the parameters of the initial model training layer to improve the generalization ability of the model without increasing a large amount of computing resources, etc. This manual does not impose any restrictions on this.

[0145] The target model training layer obtained by adjusting the parameters of the initial model training layer generates a target data processing model, which can make the model more suitable for specific network cluster data processing tasks while reducing the model training consumption.

[0146] Further, processing the question text based on the target data processing model to generate an answer text includes:

[0147] Acquire network cluster data corresponding to at least one network cluster according to the question text;

[0148] The question text and each network cluster data are processed based on the target data processing model to generate an answer text.

[0149] In practical applications, obtaining network cluster data corresponding to at least one network cluster according to the question text can be understood as extracting network cluster data related to the question by parsing the question input by the user. The specific method can be to analyze the question text using the target data processing model, automatically identify the relevant network clusters and extract their data; it can also be done by manually or automatically mapping the keywords in the question text to the network clusters through the preset keyword matching rules, and then extracting the corresponding data; it can also be done by directly calling the data of the relevant cluster from the database or API interface through the network cluster ID or name specified by the user, etc. This manual does not impose any restrictions on this.

[0150] By automatically obtaining the network cluster data corresponding to the problem text, the complexity and error rate of manual operations can be reduced, and the efficiency of data query and analysis can be improved. The system can quickly parse the text entered by the user, automatically identify the relevant network clusters, and extract the corresponding real-time data for processing, ensuring that the analysis results are more timely and accurate. This not only speeds up problem solving, but also provides operation and maintenance personnel with more intelligent tools, allowing them to focus on high-value decision-making and optimization without manually finding and entering tedious network data.

[0151] Further, obtaining network cluster data corresponding to at least one network cluster according to the question text includes:

[0152] Acquire a cluster determination prompt word template, and based on the cluster determination prompt word template and the question text, acquire a cluster determination prompt word corresponding to the question text, wherein the cluster determination prompt word template is a formatting template for enabling a data processing model to generate interface call information;

[0153] Inputting the cluster determination prompt word into the target data processing model, and acquiring data acquisition interface call information output by the target data processing model, wherein the data acquisition interface call information corresponds to at least one network cluster;

[0154] The network cluster data corresponding to each network cluster is acquired based on the data acquisition interface call information.

[0155] In actual applications, the data acquisition interface call information is the instruction obtained from the target data processing model, which contains the interface parameters required to call the network cluster related data; the cluster determination prompt word template is a predefined template for generating a specific cluster query request, which helps the target data processing model generate valid prompt words according to the question text; the cluster determination prompt word is a specific request generated based on the question text and the prompt word template, which is used to obtain the data of a specific network cluster.

[0156] Exemplarily, the data acquisition interface call information can be understood as the call instructions for extracting network cluster data generated by the target data processing model. This information includes the required interface, query parameters, and corresponding cluster ID, etc., to ensure that the real-time data of the corresponding cluster can be accurately called. For example, when a user asks "the number of active sessions in a cluster", the interface call information will be generated according to the model, specifying the cluster ID, query indicator, and time range, and calling the relevant API to obtain data. This information is a key link in the data acquisition process to ensure the accuracy and efficiency of the query.

[0157] The cluster determination prompt word template can be understood as a formatted template used by the system to generate cluster data query requests. The template contains a set of fixed query formats and variable positions. When the system receives a user question, it will fill the question text into the corresponding position of the template to generate a specific cluster query prompt. For example, when a user asks "the traffic situation of cluster B in the last 10 minutes", the template will embed "the traffic situation of cluster B in the last 10 minutes" as a variable into the template, thus forming a complete query prompt for subsequent system processing.

[0158] The cluster determination prompt word can be understood as a specific query request generated based on the question text and the prompt word template, which is used to guide the system to obtain the data of the relevant cluster from the target data processing model. For example, when a user asks the question "Device with a large number of active sessions in cluster A in the last 10 minutes", the system will generate a prompt word, such as "Please tell me how to obtain the data of the cluster described in 'Device with a large number of active sessions in cluster A in the last 10 minutes' in the corresponding time period by calling the interface 'Metric API and device information API'." This prompt word can accurately guide the data processing model to execute the query, thereby efficiently obtaining the required cluster data.

[0159] In one embodiment provided in this specification, the cluster determination prompt word template is: Please call the interface "Metric API and device information API" to obtain the data call information of the cluster described in "<question text>" in the corresponding time period. Continuing with the above example, the generated cluster determination prompt word is: Please call the interface "Metric API and device information API" to obtain the data call information of the cluster described in "What are the top 2 device IPs with larger traffic in the S1 cluster and the S2 cluster in the past 10 minutes?" in the corresponding time period.

[0160] Then, the cluster determination prompt word is input into the target data processing model. Since the data processing model is trained based on the cluster data feature information, it can identify the network cluster targeted in the question text and output the interface call information. The interface call information output by the target data processing model is: "cluster": [

[0161] {

[0162] "id": "cluster1",

[0163] "begin_ts ": "1729561389000 ",

[0164] "end_ts": "1729561989000"

[0165] }, {

[0166] "id": "cluster2",

[0167] "begin_ts": "1729561389000",

[0168] "end_ts": "1729561989000"

[0169] }]” Cluster1 is the ID of the S1 cluster in the system, and cluster2 is the ID of the S2 cluster in the system.

[0170] Automatically obtaining the network cluster data corresponding to the question text through the target data processing model can significantly improve the efficiency and accuracy of data processing. The system can quickly parse the questions raised by users, automatically identify the relevant clusters and corresponding indicators, and directly retrieve the required data from the interface, avoiding the errors and delays that may be caused by manual queries. This not only simplifies the operation process, but also ensures the real-time and accuracy of the query results, providing users with more intuitive and rapid feedback, and facilitating more efficient monitoring and optimization decisions in complex network environments.

[0171] Further, the question text and each network cluster data are processed based on the target data processing model to generate an answer text, including:

[0172] Acquire a question answer prompt word template, wherein the question answer prompt word template is used to enable the target data processing model to output an answer text based on the question text;

[0173] Generate question answer prompt words based on the question answer prompt word template, the question text and each network cluster data;

[0174] The question answer prompt words are input into the target data processing model to obtain the answer text generated by the target data processing model.

[0175] In practical applications, the question answer prompt word template is a predefined framework that guides the system to process questions raised by users and conduct detailed analysis of network cluster data. The question answer prompt word is a detailed analysis instruction generated based on the question answer prompt word template and the specific question text, which is used to guide the system to perform multi-step data processing and result output.

[0176] Exemplarily, the question answer prompt word template can be understood as a structured template used by the system to generate detailed data analysis steps. The template provides clear operating steps for the target data processing model to ensure that the system can fully process and analyze the network cluster data based on the question text. The template usually contains steps such as "extract data", "sort analysis", and "interpret results" to help the system handle complex problems in stages. For example, the question answer prompt word template may require the system to first extract specific indicators from the network cluster, then calculate these indicators, and output analysis results with explanations. The question answer prompt words can be understood as specific analysis instructions generated based on the question text and the question answer prompt word template. The system fills the question text into the relevant parts of the template to form a complete set of analysis steps.

[0177] In one embodiment provided in this specification, the question answer prompt word template is:

[0178] Based on the performance data of the network cluster and the input question '<question text>', please follow the steps below to conduct a comprehensive analysis:

[0179] 1. Extract performance-related data from the network cluster, such as CPU usage, memory consumption, and other indicators.

[0180] 2. Perform statistical and analytical operations on these performance data as required by the problem.

[0181] 3. Classify or sort the analyzed data to find the devices or nodes that meet the requirements.

[0182] 4. Describe in detail the analysis process of each step and the relevant intermediate results.

[0183] 5. Output the analysis results and explain the relationship between the results and cluster performance. "

[0184] Network cluster: <network cluster data>”, continue with the above example, and then fill the question text and network cluster data of the previous example into the corresponding part to obtain the question answer prompt word, which will not be repeated here.

[0185] The question answer prompt word template enables the target data processing model to directly process the question text and obtain the answer text, which can improve the system's automated processing capabilities and response speed. The template provides a structured analysis path for the model, enabling it to systematically parse the question text and perform multi-step data processing, making the entire process from extracting relevant data to generating results more accurate and efficient.

[0186] Further, the question text and each network cluster data are processed based on the target data processing model to generate an answer text, including:

[0187] According to the question text, obtaining answer step information corresponding to the question text;

[0188] The question text, the answer step information and each network cluster data are input into the target data processing model to obtain the answer text generated by the target data processing model.

[0189] In practical applications, answer step information is structured steps generated when the system processes the question text to guide the model to complete the answer step by step.

[0190] Exemplarily, the answer step information can be understood as a collection of various operations and analysis steps that the model needs to perform when generating the answer text. It provides a clear path for the target data processing model, and the entire process from problem parsing, data extraction, processing to result generation relies on the guidance of this information. For example, when a user asks "the device with the highest traffic volume in the network cluster in the last 10 minutes", the answer step information includes operations such as extracting relevant cluster data, calculating traffic peaks, and sorting devices. This information ensures that the system can perform the necessary analysis steps in sequence to generate accurate answers that meet user needs.

[0191] It should be noted that obtaining answer step information based on the question text can be understood as analyzing the question text to generate the steps required for the system to perform data processing and analysis. The specific method can be to analyze the question text using the target data processing model to determine the corresponding answer step information; or to match the predefined step template according to the keywords of the question to quickly generate the processing path; or to extract common steps from the answers to similar historical questions and adjust them to apply to the current question, etc. This manual does not impose any restrictions on this.

[0192] In addition, inputting the question text, answer step information and each network cluster data into the target data processing model, and obtaining the answer text generated by the target data processing model can be understood as automatically generating analysis results and answers by combining the text input by the user, the step information generated by the system and the real-time network cluster data. The specific method can be to connect the question text, the answer step information and each network cluster data in series through a simple prompt word template and input it into the target data processing model, so that the target data processing model generates the answer text, etc. This specification does not impose any restrictions on this.

[0193] In one embodiment provided in this specification, the question text, the answer step information and each network cluster data are connected in series through a question answer prompt word template containing an answer step reminder. The answer prompt word template here is: please process the following network cluster data according to the steps described below to answer <question text>;

[0194] Steps to follow: <answer step information>;

[0195] Network cluster: <network cluster data>, then fill the question text, the obtained answer step information and the network cluster data into the corresponding parts to obtain the question answer prompt words containing the answer step reminders, so that the target data processing model can output the answer text, which will not be repeated here.

[0196] By obtaining the answer step information and inputting the answer prompt words of the answer step reminder into the target data processing model to obtain the answer text, the model's processing efficiency for the answer text and the accuracy of the answer can be improved. In this way, the system can process the question text and related data in an orderly manner according to the preset analysis steps, ensure the execution order and logic of each step, and reduce confusion and uncertainty in the data processing process. This not only ensures that the answer text generated by the system is more in line with user needs, but also provides a more transparent and traceable analysis path, which facilitates the generation of more targeted results in complex query tasks, while reducing the need for manual intervention.

[0197] Further, according to the question text, obtaining answer step information corresponding to the question text includes:

[0198] Obtain a step parsing prompt word template corresponding to the question text;

[0199] Generate step parsing prompt words according to the step parsing prompt word template and the question text;

[0200] The step analysis prompt words are input into the target data processing model to obtain the answer step information generated by the target data processing model.

[0201] In actual applications, the step parsing prompt word template is a framework for guiding the system to generate processing steps, which is used to parse user questions and generate analysis steps. The step parsing prompt words are specific instructions generated based on the step parsing prompt word template, which help the system to gradually perform problem processing and analysis until the results are generated.

[0202] Exemplarily, the step parsing prompt word template can be understood as a preset template used to guide the system to parse the problem and generate execution steps. The structured content in the template ensures that the system can gradually decompose the problem, instructing the target data processing model how to extract key information from the problem text and generate appropriate analysis steps. For example, when a user asks for network performance data, the template requires the system to first extract traffic indicators, then sort and analyze, and then generate results. Through this template, the system can better convert problems into executable tasks and ensure that the processing path is clear and unambiguous.

[0203] The step parsing prompt words can be understood as instructions generated based on specific questions and step parsing prompt word templates, which clarify the various steps that the system should perform. The system fills in the template according to the question text and generates prompt words for operations such as data extraction, sorting, and calculation. For example, if a user asks "What is the network traffic situation of cluster S1 in the past 10 minutes", the step parsing prompt words may generate the following content: "Use the network traffic situation to sort the cluster data, find the devices with higher traffic, and generate analysis results." This prompt word provides the system with a step-by-step execution path to ensure that the model can accurately and efficiently answer complex questions.

[0204] It should be noted that obtaining the step parsing prompt word template corresponding to the question text can be understood as generating a preset template suitable for question parsing by analyzing the question text content, helping the system to gradually solve the user's problem. The specific method can be to determine the preset step parsing prompt word template as the step parsing prompt word template corresponding to the question text; or to parse the keywords and structures in the question text through the target data processing model, and select the most matching question parsing template from the existing template library; or to automatically match and adjust the template content to optimize the question processing process by combining historical data and the answer steps of similar questions, etc. This manual does not impose any restrictions on this.

[0205] In one embodiment provided in this specification, the question text follows the above example: What are the IP addresses of the first two devices with the largest traffic in the S1 cluster and the S2 cluster in the last 10 minutes? The step parsing prompt word template is: Please follow the steps below to determine the steps required to solve <question text>, 1. Analyze the question text and extract the main analysis tasks. 2. Determine the key data processing steps that need to be performed. 3. List the various operation steps required to solve the problem, and ensure that each step is clear and logical; then fill the question text into the above step parsing prompt word template, obtain the step parsing prompt word, and then input the prompt word into the target data processing model to obtain the answer step information output by the target data processing model: 1. Extract the OutBps and InBps data of the last 10 minutes from the S1 cluster and the S2 cluster. 2. For each device in each cluster, calculate the sum of OutBps and InBps (traffic = OutBps + InBps). 3. Sort the devices in each cluster by traffic and find the first two devices with the largest traffic. 4. Record the IP addresses of these devices and output the results, showing the IP addresses of the devices with the largest traffic in the S1 and S2 clusters.

[0206] Obtaining the steps required to solve the problem text through step analysis prompts can effectively simplify the problem handling process and ensure that the system can perform each step logically and orderly. This method can help the system quickly identify the key steps required and avoid missing or repeating operations, thereby improving the efficiency and accuracy of data processing. In addition, it can make the analysis of problems more transparent and controllable, provide users with a clear execution path and result explanation, and help to perform more accurate analysis and problem solving in complex tasks.

[0207] Considering the efficiency of subsequent question analysis, after processing the question text based on the target data processing model to generate the answer text, the method further includes:

[0208] According to the answer text, target answer information corresponding to the question text is determined.

[0209] In actual applications, the target answer information is the information actually displayed to the user. It can be understood as the actual output result generated by the system after processing the question text for the user's reference or use. Depending on the user's query content, the target answer information may include direct numerical results, chart displays, or further analysis and explanation. For example, in network cluster management, users may query the performance data of a cluster for the past 10 minutes, and the target answer information may display the traffic, latency, and load of related devices. If the user wants to view it in the form of a chart, the target answer information may generate a visual chart to show the data trend, while in simpler query scenarios, the system may only return a numerical value or a short result. This flexibility ensures that the target answer information can meet the needs of users in different scenarios.

[0210] It should be noted that the display method of the target answer information can be diversified, depending on the usage scenario and user needs, that is, the target answer information is determined according to the answer display information sent together with the question text. For example, for question texts sent by technical operation and maintenance personnel, the target answer information may be displayed through the interface of the operation and maintenance dialogue assistant to provide real-time feedback on the problem; for question texts sent by managers, the target answer information may be displayed in the form of a data dashboard to provide a summary of the operating status of the overall network cluster; in specific scenarios, the target answer information can also only display the value or status of the analysis results to avoid excessive information interference. This flexible display method not only enables the target answer information to adapt to a variety of usage scenarios, but also allows different users to obtain feedback that is more in line with their usage habits according to their needs.

[0211] Optionally, determining target answer information corresponding to the question text according to the answer text includes:

[0212] Based on the answer text, target answer information corresponding to the question text is generated.

[0213] In actual applications, the analysis step text is a detailed description of the steps performed by the system when generating the final answer, which is used to show the specific actions taken by the system in the problem processing process. The answer text is the final answer result generated by the system through these steps, which directly responds to the user's query.

[0214] Exemplarily, the analysis step text can be understood as a record of the execution process of the system when processing the problem text. This text shows how the system solves the problem step by step, and is usually used to illustrate the logic of problem solving. For example, when a user asks about the performance of a network cluster, the system may first extract the corresponding data, then perform calculations and sort, and finally draw conclusions. The analysis step text records these operation processes in detail, allowing users to understand the resolution path of the problem. This is especially important for complex technical problems, which can help users understand how the system draws conclusions and ensure the transparency of the process.

[0215] The answer text can be understood as the final result obtained by the system based on the executed steps, which is used to directly respond to the user's query. For example, when a user asks "What are the active devices in the network cluster in the last 10 minutes?", the answer text will contain the specific IP addresses and performance indicators of these active devices. The answer text is usually concise and direct, with the purpose of providing users with clear and unambiguous answers and reducing unnecessary additional information. It is the core result required by the user, which can be a value, status, or execution summary.

[0216] It should be noted that generating target answer information based on answer text can be understood as converting the answer text generated by the target data processing model into a display form that is easy for users to understand. The specific method can be to visualize the data in the answer text, such as generating charts or trend analysis, to help users intuitively understand complex data relationships; or to display the key information in the answer text in a concise summary form, which is suitable for scenarios that require quick decision-making; or to create an interactive data dashboard so that users can view and analyze different dimensional data in the results in real time, etc. This manual does not impose any restrictions on this.

[0217] By converting answer text into diversified target answer information, it can not only adapt to a variety of usage scenarios, but also allow different users to obtain feedback that is more in line with their usage habits according to their needs.

[0218] The solution of the embodiment of this specification solves the efficiency problem of traditional methods in cross-processing of multi-dimensional data, especially when facing complex problems, it can automatically generate analysis steps and answer texts, simplifying the interaction between users and the system. Different from the previous method of manually writing queries, the new data processing model can automatically call the corresponding data interface to obtain real-time data in the network cluster according to the question text and generate corresponding prompt words, and quickly parse the data based on the generated answer prompt words, which not only improves the accuracy and speed of data acquisition, but also realizes the solution process of complex data analysis in an automated way, reduces human intervention, and makes the system more suitable for a variety of application scenarios. This method provides users with an efficient and flexible solution and improves the ability to acquire and analyze network cluster data.

[0219] See also Figure 2 , Figure 2 The diagram shows an architecture diagram of a network cluster data acquisition system provided by an embodiment of the present specification. The network cluster data acquisition system may include a client 100 and a server 200;

[0220] The client 100 is used to send a question text for at least one network cluster to the server 200;

[0221] The server 200 is used to obtain a question text, and process the question text based on the target data processing model to generate an answer text, wherein the answer text includes an answer text and a step analysis text, wherein the answer text is generated by the target data processing model based on the step analysis text, and the step analysis text is generated by the target data processing model according to the question text and the network cluster data corresponding to each network cluster; and send the answer text to the client 100;

[0222] The client 100 is also used to receive the answer text sent by the server 200.

[0223] By applying the solution of the embodiments of this specification, the efficient collaboration between the client and the server greatly optimizes the process of network cluster data processing. The client can send a question text to the server, and the server automatically processes and generates the corresponding answer text or target answer based on the trained target data processing model. This design eliminates the burden of complex manual query and data processing in traditional methods. The server can not only quickly obtain key information based on the multi-dimensional data characteristics in the network cluster, but also automatically parse the meaning of the data and generate detailed answers. The client further simplifies the data interaction process and improves the efficiency of users in obtaining accurate data by receiving the answer or answer text fed back by the server. The entire solution significantly enhances the processing capability of complex data through efficient client-server interaction, making the system suitable for a wider range of network cluster application scenarios.

[0224] The network cluster data acquisition system may include multiple clients 100 and a server 200, wherein the client 100 may be referred to as a client-side device and the server 200 may be referred to as a cloud-side device. Multiple clients 100 may establish a communication connection through the server 200. In the network cluster data acquisition scenario, the server 200 is used to provide network cluster data acquisition services between multiple clients 100. Multiple clients 100 may serve as a sender or a receiver, respectively, and achieve communication through the server 200.

[0225] The user can interact with the server 200 through the client 100 to receive data sent by other clients 100, or send data to other clients 100, etc. In the network cluster data acquisition scenario, the user can publish a data stream to the server 200 through the client 100, and the server 200 generates a reply text according to the data stream and pushes the reply text to other clients that have established communication.

[0226] The client 100 and the server 200 are connected via a network. The network provides a medium for a communication link between the client 100 and the server 200. The network may include various connection types, such as wired or wireless communication links or optical fiber cables, etc. The data transmitted by the client 100 may need to be encoded, transcoded, compressed, etc. before being released to the server 200.

[0227] The client 100 can be a browser, an APP (Application), or a web application such as an H5 (HyperText Markup Language 5, Hypertext Markup Language Version 5) application, or a light application (also known as a mini-program, a lightweight application) or a cloud application, etc. The client 100 can be based on the software development kit (SDK, Software Development Kit) of the corresponding service provided by the server 200, such as based on the real-time communication (RTC, Real Time Communication) SDK development and acquisition. The client 100 can be deployed in an electronic device and needs to rely on the device to run or some APPs in the device to run. For example, the electronic device can have a display screen and support information browsing, such as a personal mobile terminal such as a mobile phone, a tablet computer, a personal computer, etc. Various other types of applications can also be configured in the electronic device, such as human-computer dialogue applications, model training applications, text processing applications, web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0228] The server 200 may include servers that provide various services, such as servers that provide communication services to multiple clients, servers for background training that support models used on clients, and servers that process data sent by clients. It should be noted that the server 200 can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. The server can also be a server of a distributed system, or a server combined with a blockchain. The server can also be a cloud server for basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content distribution networks (CDN, Content Delivery Network), and big data and artificial intelligence platforms, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology.

[0229] It is worth noting that the network cluster data acquisition method provided in the embodiments of this specification is generally executed by the server, but in other embodiments of this specification, the client may also have similar functions to the server, thereby executing the network cluster data acquisition method provided in the embodiments of this specification. In other embodiments, the network cluster data acquisition method provided in the embodiments of this specification may also be jointly executed by the client and the server.

[0230] The following combination Figure 3 Taking the application of the network cluster data acquisition method provided in this specification in acquiring the top N device information of network cluster indicators as an example, the network cluster data acquisition method is further described. Figure 3 A flowchart of a processing process of a method for obtaining the top N device information of a network cluster indicator provided by an embodiment of the present specification is shown, which specifically includes the following steps.

[0231] Step 302: Get the question text: "What are the two machine IPs with the largest number of active sessions and input bytes per second in network cluster A in the past 10 minutes?".

[0232] Step 304: The cluster determination prompt word template and the above question text set are combined to generate cluster determination prompt words which are input into the target data processing model to obtain the data acquisition interface call information output by the data processing model.

[0233] In the application of this embodiment, the prompt word template is: Please tell me how to obtain the data of the cluster described in "<question text>" within the corresponding time period by calling the interface "Metric API and device information API"; the prompt word obtained is: Please tell me how to obtain the data of the cluster described in "What are the two machine IPs with the largest number of active sessions and input bytes per second in network cluster A in the past 10 minutes?" within the corresponding time period by calling the interface "Metric API and device information API".

[0234] According to the above, the data processing model outputs the data acquisition interface call information as follows:

[0235] "cluster": [

[0236] {

[0237] "id": "cluster1",

[0238] "begin_ts ": "1729561389000 ",

[0239] "end_ts": "1729561989000"

[0240] }]” Cluster1 is the ID of network cluster A in the system.

[0241] Step 306: Use the above interface calling information to call the cluster information acquisition interface to obtain the network cluster data of network cluster A within 10 minutes.

[0242] In the application of this embodiment, the obtained network cluster data example is:

[0243] "[{

[0244] "OutPps": 5.2941175,

[0245] "OutBps": 18919.0,

[0246] "IP": "IP1",

[0247] "InBps": 78912.18,

[0248] "NewSessionPerSecond": 0.0,

[0249] "InPps": 7.7058825,

[0250] "ActiveSessionCount": 0.8947368

[0251] },{

[0252] "OutPps": 0.6666667,

[0253] "OutBps": 385.7544,

[0254] "IP": "IP2",

[0255] "InBps": 53636.246,

[0256] "NewSessionPerSecond": 0.0,

[0257] "InPps": 4.4736843,

[0258] "ActiveSessionCount": 0.46478873

[0259] },…,{

[0260] "OutPps": 0.0,

[0261] "OutBps": 210.33333,

[0262] "IP": "IPn",

[0263] "InBps": 2334.3333,

[0264] "NewSessionPerSecond": 0.0,

[0265] "InPps": 0.33333334,

[0266] "ActiveSessionCount": 0.25

[0267] }]”

[0268] Step 308: Generate question answer prompt words by combining the question answer prompt word template, network cluster data and the above question text.

[0269] In the application of this embodiment, the answer prompt word template is:

[0270] "Please conduct a detailed analysis based on the network cluster data and the question '<question text>' as follows:

[0271] 1. Extract relevant indicator information from network cluster data.

[0272] 2. Perform necessary data processing and calculations on the extracted indicators based on the problem requirements.

[0273] 3. Sort or perform other statistical operations on the processed data to find devices or values ​​that meet the conditions.

[0274] 4. Describe in detail the execution process and intermediate results of each step.

[0275] 5. Output the analysis results and explain the significance and relevant context of the results.

[0276] Network cluster data: <network cluster data>” Then fill the above question text and network cluster data into the corresponding part to obtain the question answer prompt words, which will not be repeated here.

[0277] Step 310: Input the above question answer prompt words into the target data processing model so that the target data processing model analyzes the question text and generates an answer text including the analysis step.

[0278] In the application of this embodiment, the answer text obtained is:

[0279] "Based on the data you provided, we need to find the two IP addresses with the largest ActiveSessionCount and InBps in each machine. This can be analyzed by the following steps:

[0280] ## Find the two machine IPs with the largest ActiveSessionCount:

[0281] First, we sort all machines by ActiveSessionCount from largest to smallest and take the first two:

[0282] IP3-ActiveSessionCount: 3.4833333;

[0283] IP4 - ActiveSessionCount: 0.8108108.

[0284] Find the two machine IPs with the largest InBps:

[0285] ## Next, we sort all machines by InBps from largest to smallest and take the first two:

[0286] IP1 - InBps: 78912.18;

[0287] IP2 - InBps: 53636.246.

[0288] ##result:

[0289] The two machine IPs with the largest ActiveSessionCount are IP3 and IP4.

[0290] The two machine IPs with the largest InBps are: IP1 and IP2.

[0291] I hope this analysis was helpful! Please let me know if further information or additional analysis is required.”

[0292] Among them, ActiveSessionCount is the ID of the number of active sessions in the network cluster data, and InBps is the ID of the number of input bytes per second in the network cluster data.

[0293] Step 312: Extract the answer data in the answer text, and generate the IP addresses of the first two devices of the two indicators based on the answer data.

[0294] By applying the solution of the embodiments of this specification, by combining the question text and network cluster data, the question answer prompt words are automatically generated, and the target data processing model can perform detailed data processing, sorting and calculation according to different indicator requirements, so as to accurately obtain the device information that meets the conditions. This method not only improves the efficiency of multi-dimensional and multi-indicator data processing, but also ensures the transparency of the analysis steps. Users can clearly understand the analysis process and intermediate results of each step, and obtain accurate device IP and its related indicator data. Through the automated analysis and answer generation process, the complexity of manual operations is significantly reduced, adapting to the dynamic network environment.

[0295] Corresponding to the above method embodiment, this specification also provides a network cluster data acquisition device embodiment, Figure 4 FIG. 1 is a schematic diagram showing the structure of a network cluster data acquisition device provided by an embodiment of the present specification. Figure 4 As shown, the device comprises:

[0296] The acquisition module 402 is configured to acquire a question text, wherein the question text is for at least one network cluster;

[0297] The generation module 404 is configured to process the question text based on the target data processing model to generate an answer text, wherein the answer text includes an answer text and a step analysis text, the answer text is generated by the target data processing model based on the step analysis text, and the step analysis text is generated by the target data processing model according to the question text and the network cluster data corresponding to each network cluster.

[0298] Optionally, the acquisition module 402 is further configured to:

[0299] Receiving question request information input by a user, wherein the question request information includes a request question text;

[0300] Inputting the request question text into a question judgment model to obtain a question judgment result generated by the question judgment model;

[0301] When the model judgment result is related to the network cluster, the request question text is determined to be a question text.

[0302] Optionally, the generating module 404 is further configured to:

[0303] Acquire network cluster data corresponding to at least one network cluster according to the question text;

[0304] The question text and each network cluster data are processed based on the target data processing model to generate an answer text.

[0305] Optionally, the generating module 404 is further configured to:

[0306] Acquire a cluster determination prompt word template, and based on the cluster determination prompt word template and the question text, acquire a cluster determination prompt word corresponding to the question text, wherein the cluster determination prompt word template is a formatting template for enabling a data processing model to generate interface call information;

[0307] Inputting the cluster determination prompt word into the target data processing model, and acquiring data acquisition interface call information output by the target data processing model, wherein the data acquisition interface call information corresponds to at least one network cluster;

[0308] The network cluster data corresponding to each network cluster is acquired based on the data acquisition interface call information.

[0309] Optionally, the generating module 404 is further configured to:

[0310] Acquire a question answer prompt word template, wherein the question answer prompt word template is used to enable the target data processing model to output an answer text based on the question text;

[0311] Generate question answer prompt words based on the question answer prompt word template, the question text and each network cluster data;

[0312] The question answer prompt words are input into the target data processing model to obtain the answer text generated by the target data processing model.

[0313] Optionally, the generating module 404 is further configured to:

[0314] According to the question text, obtaining answer step information corresponding to the question text;

[0315] The question text, the answer step information and each network cluster data are input into the target data processing model to obtain the answer text generated by the target data processing model.

[0316] Optionally, the generating module 404 is further configured to:

[0317] Obtain a step parsing prompt word template corresponding to the question text;

[0318] Generate step parsing prompt words according to the step parsing prompt word template and the question text;

[0319] The step analysis prompt words are input into the target data processing model to obtain the answer step information generated by the target data processing model.

[0320] Optionally, the network cluster data acquisition device further includes an answer determination module configured to:

[0321] According to the answer text, target answer information corresponding to the question text is determined.

[0322] Optionally, the network cluster data acquisition device further includes a model training module configured to:

[0323] Acquire network cluster data feature information and an initial data processing model, wherein the network cluster data feature information includes at least one network cluster indicator feature information and network cluster question and answer information;

[0324] The initial data processing model is trained based on the network cluster data feature information to obtain a target data processing model.

[0325] Optionally, the model training module is further configured as follows:

[0326] Obtaining an initial model training layer and an initial model training layer, wherein the initial model training layer is connected to a self-attention layer in the initial data processing model;

[0327] Adjust the parameters of the initial model training layer based on the network cluster data feature information to obtain a target model training layer;

[0328] The target model training layer and the initial data processing model are combined to obtain a target data processing model.

[0329] By applying the scheme of the embodiment of this specification, the design of the network cluster data acquisition device greatly improves the automation of data processing, especially in complex multi-dimensional data analysis, showing good efficiency. The acquisition module can accurately obtain the question text, and in combination with the generation module, automatically generate the answer text according to the pre-trained target data processing model, avoiding the tedious operation of manually writing queries in the traditional method. In addition, by generating clusters to determine the prompt word template and the question answer prompt word, the device can efficiently call the data interface, obtain the real-time data of the network cluster, and generate detailed analysis steps and results based on these data to ensure the accuracy and timeliness of the answer text. The confirmation module further reduces the user's intervention requirements based on the answer information generated by the analysis text and optimizes the user experience. The entire device not only improves the speed of data acquisition and processing, but also improves the generalization ability of the data processing model through the adaptive adjustment of the model, providing support for the rapid analysis and automated answering of network cluster data.

[0330] The above is a schematic scheme of a network cluster data acquisition device of this embodiment. It should be noted that the technical scheme of the network cluster data acquisition device and the technical scheme of the above-mentioned network cluster data acquisition method belong to the same concept, and the details not described in detail in the technical scheme of the network cluster data acquisition device can be referred to the description of the technical scheme of the above-mentioned network cluster data acquisition method.

[0331] Figure 5 The structure block diagram of a computing device 500 provided according to an embodiment of the present specification is shown. The components of the computing device 500 include but are not limited to a memory 510 and a processor 520. The processor 520 is connected to the memory 510 via a bus 530, and the database 550 is used to store data.

[0332] The computing device 500 also includes an access device 540 that enables the computing device 500 to communicate via one or more networks 560. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 540 may include one or more of any type of network interface (e.g., a network interface card (NIC)) that is wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a world-wide interoperability for microwave access (Wi-MAX) interface, an Ethernet interface, a universal serial bus (USB) interface, a cellular network interface, a Bluetooth interface, and a near field communication (NFC).

[0333] In one embodiment of the present specification, the above components of the computing device 500 and Figure 5 Other components not shown in the figure may also be connected to each other, for example, via a bus. It should be understood that Figure 5 The computing device structure block diagram shown is only for the purpose of illustration, and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0334] The computing device 500 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable computing device (e.g., a smart watch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or a personal computer (PC). The computing device 500 may also be a mobile or stationary server.

[0335] The processor 520 is used to execute the following computer executable instructions, which, when executed by the processor, implement the steps of the above-mentioned network cluster data acquisition method.

[0336] The above is a schematic scheme of a computing device of this embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the above-mentioned network cluster data acquisition method belong to the same concept, and the details not described in detail in the technical scheme of the computing device can be referred to the description of the technical scheme of the above-mentioned network cluster data acquisition method.

[0337] An embodiment of the present specification further provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the above-mentioned network cluster data acquisition method.

[0338] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of the storage medium and the technical scheme of the above-mentioned network cluster data acquisition method belong to the same concept, and the details not described in detail in the technical scheme of the storage medium can be referred to the description of the technical scheme of the above-mentioned network cluster data acquisition method.

[0339] An embodiment of the present specification further provides a computer program product, including a computer program / instruction, which implements the steps of the above-mentioned network cluster data acquisition method when executed by a processor.

[0340] The above is a schematic scheme of a computer program of this embodiment. It should be noted that the technical scheme of the computer program and the technical scheme of the above-mentioned network cluster data acquisition method belong to the same concept, and the details not described in detail in the technical scheme of the computer program can be referred to the description of the technical scheme of the above-mentioned network cluster data acquisition method.

[0341] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0342] The computer instructions include computer program codes, which may be in source code form, object code form, executable files or some intermediate forms, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0343] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0344] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0345] The preferred embodiments of this specification disclosed above are only used to help explain this specification. The optional embodiments do not describe all the details in detail, nor do they limit the invention to only the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that technicians in the relevant technical field can well understand and use this specification. This specification is only limited by the claims and their full scope and equivalents.

Claims

1. A network cluster data acquisition method, comprising: Obtaining a question text, wherein the question text is for at least one network cluster; According to the question text, network cluster data corresponding to at least one network cluster is obtained; based on the target data processing model, the question text and each network cluster data are processed to generate an answer text, wherein the answer text includes an answer text and a step analysis text, the answer text is generated by the target data processing model based on the step analysis text, the step analysis text is generated by the target data processing model based on the question text and the network cluster data corresponding to each network cluster, the answer text is generated by the target data processing model based on the question text, each network cluster data and a question answer prompt word template or by the target data processing model based on the question text, each network cluster data and answer step information, the answer step information is generated by the target data processing model based on the question text and the step analysis prompt word template corresponding to the question text, the question answer prompt word template is used to enable the target data processing model to output the answer text based on the question text, and the step analysis prompt word template is used to enable the target data processing model to generate answer step information based on the question text.

2. The method according to claim 1, obtaining network cluster data corresponding to at least one network cluster according to the question text, comprises: Acquire a cluster determination prompt word template, and based on the cluster determination prompt word template and the question text, acquire a cluster determination prompt word corresponding to the question text, wherein the cluster determination prompt word template is a formatting template for enabling a data processing model to generate interface call information; Inputting the cluster determination prompt word into the target data processing model, and acquiring data acquisition interface call information output by the target data processing model, wherein the data acquisition interface call information corresponds to at least one network cluster; The network cluster data corresponding to each network cluster is acquired based on the data acquisition interface call information.

3. The method according to claim 1, processing the question text and each network cluster data based on the target data processing model to generate an answer text, comprising: Get the question answer prompt word template; Generate question answer prompt words based on the question answer prompt word template, the question text and each network cluster data; The question answer prompt words are input into the target data processing model to obtain the answer text generated by the target data processing model.

4. The method according to claim 1, processing the question text and each network cluster data based on the target data processing model to generate an answer text, comprising: According to the question text, obtaining answer step information corresponding to the question text; The question text, the answer step information and each network cluster data are input into the target data processing model to obtain the answer text generated by the target data processing model.

5. The method according to claim 4, obtaining answer step information corresponding to the question text according to the question text, comprising: Obtain a step parsing prompt word template corresponding to the question text; Generate step parsing prompt words according to the step parsing prompt word template and the question text; The step analysis prompt words are input into the target data processing model to obtain the answer step information generated by the target data processing model.

6. The method according to claim 1, after processing the question text based on the target data processing model to generate the answer text, the method further comprises: According to the answer text, target answer information corresponding to the question text is determined.

7. The method according to claim 1, wherein the target data processing model is obtained by training the following steps, including: Acquire network cluster data feature information and an initial data processing model, wherein the network cluster data feature information includes at least one network cluster indicator feature information and network cluster question and answer information; The initial data processing model is trained based on the network cluster data feature information to obtain a target data processing model.

8. The method according to claim 7, training the initial data processing model based on the network cluster data feature information to obtain the target data processing model, comprising: Obtaining an initial model training layer, wherein the initial model training layer is connected to a self-attention layer in the initial data processing model; Adjust the parameters of the initial model training layer based on the network cluster data feature information to obtain a target model training layer; The target model training layer and the initial data processing model are combined to obtain a target data processing model.

9. The method according to claim 1, obtaining the question text, comprising: Receiving question request information input by a user, wherein the question request information includes a question request text; Inputting the request question text into a question judgment model to obtain a question judgment result generated by the question judgment model; When the problem determination result is related to the network cluster, the request question text is determined to be the question text.

10. A network cluster data acquisition device, comprising: An acquisition module is configured to acquire a question text, wherein the question text is for at least one network cluster; A generation module is configured to obtain network cluster data corresponding to at least one network cluster according to the question text; process the question text and each network cluster data based on a target data processing model to generate an answer text, wherein the answer text includes an answer text and a step analysis text, the answer text is generated by the target data processing model based on the step analysis text, the step analysis text is generated by the target data processing model based on the question text and the network cluster data corresponding to each network cluster, the answer text is generated by the target data processing model based on the question text, each network cluster data and a question answer prompt word template or by the target data processing model based on the question text, each network cluster data and answer step information, the answer step information is generated by the target data processing model based on the question text and a step analysis prompt word template corresponding to the question text, the question answer prompt word template is used to enable the target data processing model to output the answer text based on the question text, and the step analysis prompt word template is used to enable the target data processing model to generate answer step information based on the question text.

11. A computing device comprising: Memory and processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer program / instructions are executed by the processor, the steps of the method according to any one of claims 1 to 9 are implemented.

12. A computer-readable storage medium storing a computer program / instruction, wherein the computer program / instruction, when executed by a processor, implements the steps of the method according to any one of claims 1 to 9.

13. A computer program product, comprising a computer program / instruction, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 9.

Images