Model training method and device based on data center, medium and equipment
By providing business definition interface and distributed acquisition points in the data center, customizing services and generating monitoring models, the problem of high data acquisition and labeling costs in large model training is solved, training efficiency and adaptability are improved, and labor costs are reduced.
Patent Information
- Application Number
- CN202411195779.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-12-06
AI Technical Summary
During the training process of large-scale model, especially in the data acquisition and labeling stages, there are problems such as high cost, insufficient data set coverage, inconsistency in labeling and time-consuming, which leads to delaying the progress of model development.
Provide a business definition interface in the data center. Users can customize services, collect business-related data through distributed acquisition points, perform pre-processing and conduct model training, generate monitoring models, and adjust the training process through model evaluation results to reduce manual labeling requirements.
It improves model training efficiency, reduces computational burden and labor costs, ensures that the model can adapt to business changes in a timely manner and provides better services.
Smart Images
Figure CN119089244A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of model training technology, and in particular to a data center-based model training method, device, medium, and equipment. Background Art
[0002] With the rapid development of data center technology, cloud network systems have become more and more complex, and the frequency of failures has also increased, which has a great impact on our production and life. Therefore, after a failure occurs, staff need to make operation and maintenance decisions quickly and accurately to find problems in the network. Today's cloud network systems are intricate, and it is very difficult for operation and maintenance personnel to troubleshoot problems in the massive amount of data generated by such a complex system. We need to improve our operation and maintenance methods and ideas to help operation and maintenance engineers quickly diagnose problems.
[0003] Although large models have shown significant advantages in many fields, their training process, especially in the data collection, training and annotation stages, also has some significant disadvantages and challenges. Large models often require massive amounts of data for training, which requires efficient data collection and storage systems. The size of the data can be costly, especially when it comes to special or professional data. In order for the model to have good generalization capabilities, it is necessary to cover a wide range of diverse datasets. Collecting a comprehensive and balanced dataset is a complex and time-consuming process. In addition, data annotation also faces problems such as high cost and lack of consistency in annotation. High-quality annotation work usually needs to be done manually, which is an extremely expensive task for large-scale datasets. Ensuring consistency and accuracy between different annotators is also a challenge, especially in complex or subjective tasks. The annotation of large-scale datasets can take months or even years, which can seriously delay the development of models. Summary of the invention
[0004] In response to at least one of the above technical problems, an embodiment of the present invention provides a data center-based model training method, device, medium, and equipment.
[0005] According to a first aspect, a data center-based model training method provided by an embodiment of the present invention is performed by the data center, and the method includes:
[0006] Providing a service definition interface so that users can add new services or adjust existing services according to business needs on the service definition interface;
[0007] Determine the collection points and collection strategies of each distributed device of the data center, and notify each collection point to collect business-related data according to the collection strategy;
[0008] Determine whether the newly added or adjusted business is in a normal state;
[0009] If yes, preprocess the service-related data, use the preprocessed service-related data as training positive samples, and perform model training based on the training positive samples to obtain a monitoring model for monitoring the network status of the newly added service or the adjusted service;
[0010] According to the business flow of the newly added business or the adjusted business, the monitoring model is evaluated to obtain a model evaluation result; if the model evaluation result is qualified, the training is completed; if the model evaluation result is unqualified, data collection and model training are continued.
[0011] In one embodiment, the data center is deployed with an operation and maintenance platform; correspondingly, the method further includes: monitoring the model training process through the operation and maintenance platform, and performing operation and maintenance processing on abnormal situations occurring during the model training process.
[0012] In one embodiment, the data center is deployed with a training platform; correspondingly, providing a business definition interface includes: providing a business definition interface through the training platform; determining the collection points and collection strategies of each distributed device in the data center, and notifying each collection point to collect business-related data according to the collection strategy, includes: determining the collection points and collection strategies of each distributed device in the data center through the training platform, and notifying each collection point to collect business-related data according to the collection strategy; judging whether the newly added business or the adjusted business is in a normal state includes: judging whether the newly added business or the adjusted business is in a normal state through the training platform; preprocessing the business-related data includes: preprocessing the business-related data through the training platform; evaluating the monitoring model according to the business flow of the newly added business or the adjusted business to obtain a model evaluation result includes: evaluating the monitoring model according to the business flow of the newly added business or the adjusted business through the training platform to obtain a model evaluation result.
[0013] In one embodiment, the data center is also deployed with a large model training cluster; correspondingly, the model training based on the training positive samples includes: sending the training positive samples to the large model training cluster through the training platform, so that the large model training cluster performs model training based on the training positive samples.
[0014] In one embodiment, the distributed device includes a firewall, a switch, and a virtual machine.
[0015] In one embodiment, the collection strategy includes collection rounds, collection scope, and expected data format.
[0016] In one embodiment, providing a service definition interface so that users can add new services or adjust existing services according to business needs on the service definition interface includes: providing a service definition interface so that users can add new services or adjust existing services according to business needs on the service definition interface, and defining at least one of the service IP, port, external network IP for receiving service requests and traffic characteristics of the new or adjusted service.
[0017] According to a second aspect, a data center-based model training device provided by an embodiment of the present invention is deployed on the data center, and the device includes:
[0018] An interface providing module is used to provide a service definition interface so that users can add new services or adjust existing services according to business needs on the service definition interface;
[0019] A collection determination module, used to determine the collection points and collection strategies of each distributed device in the data center, and to notify each collection point to collect business-related data according to the collection strategy;
[0020] A status judgment module is used to judge whether a new service or adjusted service is in a normal state;
[0021] A model training module is used to pre-process the service-related data, use the pre-processed service-related data as training positive samples, and perform model training based on the training positive samples to obtain a monitoring model for monitoring the network status of the newly added service or the adjusted service;
[0022] The model evaluation module is used to evaluate the monitoring model according to the business flow of the new business or the adjusted business to obtain the model evaluation result; if the model evaluation result is qualified, the training is completed; if the model evaluation result is unqualified, data collection and model training will continue.
[0023] In one embodiment, the data center is deployed with an operation and maintenance platform;
[0024] Correspondingly, the device further includes:
[0025] The model monitoring module is used to monitor the model training process through the operation and maintenance platform, and perform operation and maintenance processing on abnormal situations that occur during the model training process.
[0026] In one embodiment, the data center is deployed with a training platform;
[0027] Correspondingly, the interface providing module is specifically used to: provide a business definition interface through the training platform; the collection determination module is specifically used to: determine the collection points and collection strategies of each distributed device in the data center through the training platform, and notify each collection point to collect business-related data according to the collection strategy; the status judgment module is specifically used to: determine whether the new business or the adjusted business is in a normal state through the training platform; the model training module is specifically used to: pre-process the business-related data through the training platform; the model evaluation module is specifically used to: evaluate the monitoring model according to the business flow of the new business or the adjusted business through the training platform to obtain the model evaluation result.
[0028] In one embodiment, the data center is also deployed with a large model training cluster; correspondingly, the model training module is specifically used to: send the training positive samples to the large model training cluster through the training platform, so that the large model training cluster performs model training according to the training positive samples.
[0029] In one embodiment, the distributed device includes a firewall, a switch, and a virtual machine.
[0030] In one embodiment, the collection strategy includes collection rounds, collection scope, and expected data format.
[0031] In one embodiment, the interface providing module is specifically used to: provide a service definition interface so that users can add new services or adjust existing services according to business needs on the service definition interface, and define at least one of the service IP, port, external network IP for receiving service requests and traffic characteristics of the new or adjusted services.
[0032] According to a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, enables the computer to execute the method provided in the first aspect.
[0033] According to a fourth aspect, an embodiment of the present invention provides a computing device, including a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, the method provided in the first aspect is implemented.
[0034] The data center-based model training method, device, medium, and equipment provided by the embodiment of the present invention, the data center provides a service definition interface, so that users can add new services or adjust existing services according to business needs on the service definition interface. Determine the collection points and collection strategies of each distributed device of the data center, and notify each collection point to collect service-related data according to the collection strategy. When the new service or adjusted service is in a normal state, pre-process the service-related data, use the pre-processed service-related data as a training positive sample, and perform model training based on the training positive sample to obtain a monitoring model for monitoring the network status of the new service or adjusted service; evaluate the monitoring model according to the service flow of the new service or adjusted service to obtain a model evaluation result; if the model evaluation result is qualified, the training is completed; if the model evaluation result is unqualified, continue data collection and model training. It can be seen that users can customize services on the service definition interface of the data center, and perform model training for the new service or adjusted service, so that the trained large model matches the new service or adapts to the changes of the adjusted service, so that the trained large model can provide better services. Moreover, through distributed collection points, diverse data can be collected in a timely manner. Since the collected data is timely and high-quality, it can effectively improve the training efficiency and reduce the computing burden of the training cluster. The data is collected when the business is running normally, and can be used as positive samples. There is no need for a lot of manual labeling work, which reduces the cost of manpower and material resources. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 It is a flowchart of a data center-based model training method in one embodiment of the present invention;
[0036] Figure 2 A schematic diagram of various parts involved in an embodiment of the present invention;
[0037] Figure 3 It is a structural block diagram of a data center-based model training device in one embodiment of the present invention. DETAILED DESCRIPTION
[0038] In a first aspect, an embodiment of the present invention provides a data center-based model training method, which is executed by the data center, see Figure 1 , the method includes the following S110 to S150:
[0039] S110, providing a service definition interface, so that the user can add a new service or adjust the original service according to the service demand on the service definition interface;
[0040] In other words, the data center provides a unified business definition interface, on which users can add a new business according to business needs, or modify and adjust the original business. It can be seen that at this time, the user's business changes, thereby triggering the training task.
[0041] In one embodiment, providing a service definition interface in S110 so that a user can add a new service or adjust an existing service according to service requirements on the service definition interface may include:
[0042] A service definition interface is provided so that users can add new services or adjust existing services according to business needs on the service definition interface, and define at least one of the service IP, port, external network IP for receiving service requests and traffic characteristics of the new or adjusted services.
[0043] It can be seen that users can add new services or adjust existing services on the service definition interface. At the same time, they will also define the relevant information of the new or adjusted services. This information includes at least one of the service IP, port, external network IP for receiving service requests, and traffic characteristics. By defining this information, the relevant configuration of the new or adjusted services can be achieved.
[0044] Among them, see Figure 2 The data center is deployed with a training platform, an operation and maintenance platform, and a large model training cluster. The subsequent collection points are deployed on various distributed devices to form distributed collection points.
[0045] The providing of the service definition interface in S110 may include: providing the service definition interface through the training platform, that is, the training platform provides the service definition interface.
[0046] S120, determining the collection points and collection strategies of each distributed device of the data center, and notifying each collection point to collect business-related data according to the collection strategy;
[0047] The distributed equipment may include a firewall, a switch and a virtual machine.
[0048] The collection strategy includes collection rounds, collection scope and expected data format.
[0049] In one embodiment, determining the collection points and collection strategies of each distributed device in the data center in S120, and notifying each collection point to collect business-related data according to the collection strategy, may include: determining the collection points and collection strategies of each distributed device in the data center through the training platform, and notifying each collection point to collect business-related data according to the collection strategy;
[0050] That is to say, the training platform determines which collection points are used to perform the collection task and the collection strategy used by the collection points for data collection, and then collects data through these collection points according to the collection strategy, and returns the collected business-related data to the training platform.
[0051] S130, determining whether the newly added service or the adjusted service is in a normal state;
[0052] In one embodiment, the step of determining whether the newly added service or the adjusted service is in a normal state in S130 may include: determining whether the newly added service or the adjusted service is in a normal state by the training platform. That is, the training platform determines whether the newly added service or the adjusted service is in a normal state.
[0053] Specifically, the training platform can determine whether the newly added business or the adjusted business is in a normal operating state based on the data returned by the collection point.
[0054] S140: If yes, preprocess the service-related data, use the preprocessed service-related data as training positive samples, and perform model training based on the training positive samples to obtain a monitoring model for monitoring the network status of the newly added service or the adjusted service;
[0055] It can be seen that subsequent steps are only executed when the business is running normally. This ensures that the collected data is data generated by the business under normal circumstances. The data is preprocessed and the processed business-related data is used as training positive samples. The training positive samples are used in model training, and the trained model can be used to monitor the network status of the business.
[0056] The preprocessing of the business-related data may include: preprocessing the business-related data through the training platform. That is, the training platform executes S140.
[0057] Among them, the performing model training according to the training positive samples may include: sending the training positive samples to a large model training cluster through the training platform, so that the large model training cluster performs model training according to the training positive samples.
[0058] It can be seen that after the training platform preprocesses the business-related data, it uses the preprocessed business-related data as training positive samples, and then sends the training positive samples to the large model training cluster, which performs model training based on the training positive samples.
[0059] In one embodiment, the method may further include: monitoring the model training process through the operation and maintenance platform, and performing operation and maintenance processing on abnormal situations that occur during the model training process.
[0060] It can be seen that during the model training process of the large model training cluster, the operation and maintenance platform monitors the training process to determine whether there are any abnormalities in the model training process. If there are any abnormalities, the abnormal situation will be handled by operation and maintenance to ensure that the large model training cluster can train the model normally.
[0061] S150. Evaluate the monitoring model according to the business flow of the newly added business or the adjusted business to obtain a model evaluation result; if the model evaluation result is qualified, the training is completed; if the model evaluation result is unqualified, continue data collection and model training.
[0062] In one embodiment, the step of evaluating the monitoring model according to the business flow of the new business or the adjusted business to obtain the model evaluation result in S150 may include: evaluating the monitoring model according to the business flow of the new business or the adjusted business through the training platform to obtain the model evaluation result.
[0063] It can be seen that the training platform evaluates the monitoring model based on the business flow generated by the business in the application scenario to obtain the model evaluation result, which is more authoritative. If the model evaluation result is unqualified, continue to collect data and continue model training until the evaluation result of the trained model is qualified.
[0064] Among them, the network traffic of the data center is uncertain and will fluctuate with time and business needs, and the data types are also diverse, but at the same time, the services of the data center are relatively fixed, and the data center will provide a relatively unified management interface. Based on these characteristics, the embodiment of the present invention proposes a large model training solution, where users can customize services on the service definition interface of the data center, and perform model training for new services or adjusted services, so that the trained large model matches the new services or adapts to the changes in the adjusted services, so that the trained large model can provide better services.
[0065] The basic idea of the present invention is that users customize their services on the service definition interface provided by the data center, thereby distributing the collection of service-related data, such as service-related traffic data. When the service is normal, the service-related data is organized into training positive samples, so that model training is performed based on the training positive samples, and the performance of the trained model is evaluated. Since the service definition interface is user-oriented, users can change services at any time according to their needs. When the service changes, timely collection of service-related data and timely model training can improve the adaptability of the model. The trained model can monitor the traffic of the network center in real time, issue fault alarms, or assist in network operation and maintenance.
[0066] The present invention mainly involves the following parts:
[0067] (1) Training platform: On the user side, the management interface of the data center is relatively unified, and users can quickly define business changes of cloud servers and networks through the unified interface.
[0068] (2) Collection points: Based on the characteristics of unified network management in data centers, multiple collection points can be defined in the network, and the data collected at different stages can be merged into a logical data set.
[0069] (3) Large model training cluster: After obtaining the collected data, the training platform uniformly calls the training cluster interface to enable the large model training cluster to perform model training.
[0070] (4) Operation and maintenance platform: can be used to manage and maintain the training status of large model training clusters.
[0071] The training platform can interact with the data collection service and the large model training cluster through APIs. The data collection service can be responsible for collecting data in the real environment, and the large model training cluster uses the acquired data for training and provides an API interface to the operation and maintenance platform.
[0072] The large model training cluster is a group of interconnected and collaborative high-performance computing nodes deployed in the data center, specifically used to support large-scale machine learning model training tasks. The large model training cluster provides APIs to the training platform, operation and maintenance platform, and data collection services. Using the data collected by the data collection service, combined with user-defined scenario strategies, etc., high-quality data can be used for efficient training.
[0073] In the large model training process of the prior art, a massive amount of data is needed for training, which requires an efficient data collection and storage system. In addition, high-quality data requires a lot of manpower and material resources for labeling, etc. At the same time, because the labeling work may take months or even years, it will seriously delay the development progress of the model. For this reason, the embodiment of the present invention proposes a model training method based on a data center, which can effectively solve the above problems:
[0074] 1. The embodiments of the present invention can effectively reduce the burden of large model training clusters, because various data can be collected in a timely manner through distributed collection points, and the training platform can quickly adjust the data collection strategy and scope to maintain the pertinence and timeliness of data collection. It can be seen that the collected data is high-quality. Since the collected data is timely and high-quality, it can effectively improve the training efficiency and reduce the computational burden of the training cluster;
[0075] Second, the data is collected when the business is operating normally, which can be used as positive samples. It does not require a lot of manual labeling work, thus reducing the cost of manpower and material resources;
[0076] 3. When a user-deployed service in the data center is changed or added, it can be defined through the business definition interface for end users, thereby triggering the model training task to ensure that the model can adapt to service changes in a timely manner.
[0077] It can be seen that the embodiments of the present invention can better solve the problems of insufficient training resources, cost, timeliness and adaptability.
[0078] In a second aspect, an embodiment of the present invention provides a data center-based model training device, wherein the device is deployed on the data center, see Figure 3 , the device 100 comprises:
[0079] The interface providing module 110 is used to provide a service definition interface, so that the user can add a new service or adjust the original service according to the service demand on the service definition interface;
[0080] The collection determination module 120 is used to determine the collection points and collection strategies of each distributed device in the data center, and notify each collection point to collect business-related data according to the collection strategy;
[0081] The status determination module 130 is used to determine whether the newly added service or the adjusted service is in a normal state;
[0082] A model training module 140 is used to pre-process the service-related data, use the pre-processed service-related data as training positive samples, and perform model training based on the training positive samples to obtain a monitoring model for monitoring the network status of the newly added service or the adjusted service;
[0083] The model evaluation module 150 is used to evaluate the monitoring model according to the business flow of the new business or the adjusted business to obtain the model evaluation result; if the model evaluation result is qualified, the training is completed; if the model evaluation result is unqualified, data collection and model training will continue.
[0084] In one embodiment, the data center is deployed with an operation and maintenance platform;
[0085] Correspondingly, the device further includes:
[0086] The model monitoring module is used to monitor the model training process through the operation and maintenance platform, and perform operation and maintenance processing on abnormal situations that occur during the model training process.
[0087] In one embodiment, the data center is deployed with a training platform;
[0088] Correspondingly, the interface providing module is specifically used to: provide a business definition interface through the training platform; the collection determination module is specifically used to: determine the collection points and collection strategies of each distributed device in the data center through the training platform, and notify each collection point to collect business-related data according to the collection strategy; the status judgment module is specifically used to: determine whether the new business or the adjusted business is in a normal state through the training platform; the model training module is specifically used to: pre-process the business-related data through the training platform; the model evaluation module is specifically used to: evaluate the monitoring model according to the business flow of the new business or the adjusted business through the training platform to obtain the model evaluation result.
[0089] In one embodiment, the data center is also deployed with a large model training cluster; correspondingly, the model training module is specifically used to: send the training positive samples to the large model training cluster through the training platform, so that the large model training cluster performs model training according to the training positive samples.
[0090] In one embodiment, the distributed device includes a firewall, a switch, and a virtual machine.
[0091] In one embodiment, the collection strategy includes collection rounds, collection scope, and expected data format.
[0092] In one embodiment, the interface providing module is specifically used to: provide a service definition interface so that users can add new services or adjust existing services according to business needs on the service definition interface, and define at least one of the service IP, port, external network IP for receiving service requests and traffic characteristics of the new or adjusted services.
[0093] It is understandable that the explanation, specific implementation, beneficial effects, examples, etc. of the relevant contents in the device provided in the embodiment of the present invention can be found in the corresponding parts of the method provided in the first aspect, and will not be repeated here.
[0094] In a third aspect, an embodiment of the present invention provides a computer-readable medium, on which computer instructions are stored. When the computer instructions are executed by a processor, the processor executes the method provided in the first aspect.
[0095] Specifically, a system or device equipped with a storage medium can be provided, on which software program code that implements the functions of any of the above-mentioned embodiments is stored, and a computer (or CPU or MPU) of the system or device can be enabled to read and execute the program code stored in the storage medium.
[0096] In this case, the program code itself read from the storage medium can realize the function of any one of the above-mentioned embodiments, and thus the program code and the storage medium storing the program code constitute a part of the present invention.
[0097] The storage medium embodiments for providing the program code include a floppy disk, a hard disk, a magneto-optical disk, an optical disk (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), a magnetic tape, a non-volatile memory card, and a ROM. Alternatively, the program code can be downloaded from a server computer by a communication network.
[0098] In addition, it should be clear that the functions of any of the above embodiments can be implemented not only by executing the program code read by the computer, but also by enabling an operating system operating on the computer to complete part or all of the actual operations based on instructions from the program code.
[0099] In addition, it can be understood that the program code read from the storage medium is written to a memory provided in an expansion board inserted into the computer or to a memory provided in an expansion module connected to the computer, and then based on the instructions of the program code, a CPU installed on the expansion board or expansion module is enabled to perform part or all of the actual operations, thereby realizing the functions of any of the above-mentioned embodiments.
[0100] It is understandable that the explanation, specific implementation methods, beneficial effects, examples, etc. of the relevant contents in the computer-readable medium provided in the embodiment of the present invention can be found in the corresponding parts of the method provided in the first aspect, and will not be repeated here.
[0101] In a fourth aspect, an embodiment of the present specification provides a computing device, including a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, it implements the method in any embodiment of the specification.
[0102] It is understandable that the explanation, specific implementation, beneficial effects, examples, etc. of the relevant contents in the computing device provided in the embodiment of the present invention can be found in the corresponding parts of the method provided in the first aspect, and will not be repeated here.
[0103] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0104] Those skilled in the art should be aware that in one or more of the above examples, the functions described in the present invention can be implemented by hardware, software, widgets, or any combination thereof. When implemented by software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium.
[0105] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solution of the present invention should be included in the scope of protection of the present invention.
Claims
1. A data center-based model training method, characterized in that: Executed by the data center, the method includes: Providing a service definition interface so that users can add new services or adjust existing services according to business needs on the service definition interface; Determine the collection points and collection strategies of each distributed device of the data center, and notify each collection point to collect business-related data according to the collection strategy; Determine whether the newly added or adjusted business is in a normal state; If yes, preprocess the service-related data, use the preprocessed service-related data as training positive samples, and perform model training based on the training positive samples to obtain a monitoring model for monitoring the network status of the newly added service or the adjusted service; According to the business flow of the newly added business or the adjusted business, the monitoring model is evaluated to obtain a model evaluation result; if the model evaluation result is qualified, the training is completed; if the model evaluation result is unqualified, data collection and model training are continued.
2. The method according to claim 1, characterized in that The data center is deployed with an operation and maintenance platform; correspondingly, the method further includes: The model training process is monitored through the operation and maintenance platform, and operation and maintenance processing is performed on abnormal situations that occur during the model training process.
3. The method according to claim 1, characterized in that The data center is deployed with a training platform; Correspondingly, providing a business definition interface includes: providing a business definition interface through the training platform; determining the collection points and collection strategies of each distributed device in the data center, and notifying each collection point to collect business-related data according to the collection strategy, includes: determining the collection points and collection strategies of each distributed device in the data center through the training platform, and notifying each collection point to collect business-related data according to the collection strategy; judging whether the new business or the adjusted business is in a normal state includes: judging whether the new business or the adjusted business is in a normal state through the training platform; preprocessing the business-related data includes: preprocessing the business-related data through the training platform; evaluating the monitoring model according to the business flow of the new business or the adjusted business to obtain a model evaluation result includes: evaluating the monitoring model according to the business flow of the new business or the adjusted business through the training platform to obtain a model evaluation result.
4. The method according to claim 3, characterized in that: The data center is also deployed with a large model training cluster; Correspondingly, the performing model training according to the training positive samples includes: sending the training positive samples to a large model training cluster through the training platform, so that the large model training cluster performs model training according to the training positive samples.
5. The method according to claim 1, characterized in that The distributed equipment includes a firewall, a switch and a virtual machine.
6. The method according to claim 1, characterized in that The collection strategy includes collection rounds, collection scope and expected data format.
7. The method according to claim 1, characterized in that The provision of a service definition interface so that the user can add a new service or adjust the original service according to business needs on the service definition interface includes: providing a service definition interface so that the user can add a new service or adjust the original service according to business needs on the service definition interface, and defining at least one of the service IP, port, external network IP for receiving service requests and traffic characteristics of the new or adjusted service.
8. A data center-based model training device, characterized in that: The device is deployed in the data center, and includes: An interface providing module is used to provide a service definition interface so that users can add new services or adjust existing services according to business needs on the service definition interface; A collection determination module, used to determine the collection points and collection strategies of each distributed device in the data center, and to notify each collection point to collect business-related data according to the collection strategy; A status judgment module is used to judge whether a new service or adjusted service is in a normal state; A model training module is used to pre-process the service-related data, use the pre-processed service-related data as training positive samples, and perform model training based on the training positive samples to obtain a monitoring model for monitoring the network status of the newly added service or the adjusted service; The model evaluation module is used to evaluate the monitoring model according to the business flow of the new business or the adjusted business to obtain the model evaluation result; if the model evaluation result is qualified, the training is completed; if the model evaluation result is unqualified, data collection and model training will continue.
9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed in a computer, the computer is caused to execute the method according to any one of claims 1 to 7.
10. A computing device, characterized in that The method comprises a memory and a processor, wherein the memory stores executable codes, and when the processor executes the executable codes, the method according to any one of claims 1 to 7 is implemented.