CN115643249A - Construction method of AI teaching practical training programming platform based on Web page - Google Patents

Abstract

The invention discloses a construction method of an AI teaching practical training programming platform based on a Web page. The method of the invention comprises the following steps: (1) Constructing a bottom cloud native environment, including a Kubernets cluster, a Ceph storage layer, a Harbor mirror image management and a Nacos service discovery; (2) The back end adopts the micro-service technology, is deployed on a Kubernets cluster in a Pod form, is registered on Nacos, and performs service discovery and access through a unified gateway entrance; (3) The front end also adopts the micro-service technology and realizes unified access through the Nginx agent; (4) And establishing a database for storing some platform-related information such as users and resources of the platform. The invention enables users to conveniently, quickly and efficiently learn and apply artificial intelligence technology. The system solves a series of problems that the personal development and construction environment of the user is not uniform, multi-user teaching is not convenient, the user management efficiency is low and the like.

Description

A construction method of a web page-based AI teaching training programming platform

technical field

The invention relates to the fields of artificial intelligence, recommendation algorithms and automatic machine learning, and in particular to a method for constructing a web page-based AI teaching training programming platform.

Background technique

With the advent of the era of artificial intelligence, more and more universities and research institutes are carrying out AI course teaching, but the versions of the AI learning environment are complicated and depend on a lot. There are many versions of Tensorflow alone, not to mention other AI software such as Pytorch It is not friendly to newcomers to install, and it will cause a situation of multiple people and multiple environments, which is not conducive to the unified teaching management of teachers. A unified, efficient and stable AI learning environment must be adopted.

In order to solve these problems, many AI orchestration platforms have emerged on the market, such as KubeFlow, MLFlow, etc., but these platforms also have some problems, some of which are heavy platforms and difficult to install and deploy for individuals; some lack efficient and useful unified management function, the teacher cannot manage it conveniently and effectively; at the same time, the user's personal environment resource configuration does not necessarily meet the minimum requirements of the platform and cannot experience the whole process of smooth learning.

Contents of the invention

Purpose of the invention: In view of the problems and deficiencies in the prior art above, the purpose of the invention is to provide a method for building a web page-based AI teaching and training programming platform, which integrates the current mainstream AI technology and cloud native framework, including Kubernetes , Ceph, Harbor, etc., from learning, storage, query, to computing, covering all parts of the AI teaching system, but also abandoning the disadvantages of these native frameworks: inflexibility in interacting with users In view of the disadvantages of difficulty in getting started, the methods of using these technologies are reconstructed on the webpage, providing an easy-to-learn and easy-to-use method, so that users can learn and apply artificial intelligence technology conveniently, quickly and efficiently. It solves a series of problems such as the inconsistency of the user's personal development and construction environment, inconvenient multi-person teaching, and low user management efficiency.

Technical solution: In order to achieve the purpose of the above invention, the technical solution adopted in the present invention provides a method for building a web page-based AI teaching training programming platform, including the following steps:

(1) Construct the underlying cloud-native environment, including Kubernetes cluster, Ceph storage layer, Harbor image management and Nacos service discovery;

(2) The backend adopts microservice technology, deployed on the Kubernetes cluster in the form of Pod, and registered on Nacos, and performs service discovery and access through the unified gateway entrance;

(3) The front end also adopts micro-service technology, and realizes unified entrance access through Nginx proxy;

(4) Establish a database to store some information related to the platform, such as users and resources of the platform;

Among them, remote communication can be performed between the backend service in step (2) and the frontend service in step (3), and the two are deployed in an independent environment; in step (2), the backend service receives every request from the frontend service Permission verification will be performed to prevent malicious attacks.

Further, in the step (1), first deploy the cloud-native basic environment Kubernetes cluster; deploy the Ceph storage layer based on the Kubernetes cluster. After the deployment of Ceph, you need to set the default storage of the bottom layer of the Kubernetes cluster to Ceph, and Ceph is used for the entire Kubernetes. The cluster and various services on it provide underlying distributed storage services; then deploy the image management warehouse Harbor based on the Kubernetes cluster, and Harbor provides management functions for users to use the default training environment image and custom training environment image; finally, based on the Kubernetes cluster Deploy Nacos services for service discovery of backend microservices.

Further, in the step (2), first develop corresponding functional microservice modules according to the interface functions of the front-end services, and each microservice module is responsible for processing corresponding interface functions of different types; then for the IP address, port number, host list These configuration parameters that need to be dynamically changed are uniformly extracted into the configuration file or set to the environment variables of the Kubernetes cluster. The configuration files and the environment variables of the Kubernetes cluster include the internal VIP of some services, the database configuration, and the configuration parameters of some supporting services. Each function Microservice modules have their own configuration parameters. Finally, configure all the parameters corresponding to microservices in the startup file or set them in the Kubernetes cluster environment variables, so as to correctly start the corresponding microservice modules; when processing front-end interface function requests , each request has a special method for response processing, which is controlled by the method implemented by SpringCloud, and then the specific requirements are sent to the corresponding underlying cluster architecture or microservice for implementation, and then the running results are fed back to the interface for display; at the same time All data generated by the user during use, including operation logs, will be stored in the database.

Further, in the step (3), first compile and package the front-end microservices of the platform to a specified directory, and write its corresponding configuration into the configuration file, including the communication IP and port configuration information with the backend microservices, and The unified entrance access is realized through the Nginx proxy. When the front-end micro-service starts, it will read the configuration file, so as to realize the communication with the back-end micro-service; the main function of each interface is an independent micro-application module, and the whole platform consists of multiple The composition of different micro-application modules, the administrator function includes the following micro-application modules: home page, big data service, my classroom, educational administration management, data center, experimental project, user service, resource management, operation log, a total of nine modules, each Each module has its own corresponding functions; teacher functions include the following micro-application modules: home page, educational administration management, data center, experimental project, user service, resource management, a total of six modules, each module has its own corresponding functions; student functions include The following micro-application modules: Homepage, My Classroom, Data Center, Experimental Project, and Dataset Configuration are five modules in total, and each module has its own corresponding functions.

Further, in the step (4), each microservice corresponds to a database, and different libraries are isolated from each other, and the database service is installed and deployed, and then its user permissions are opened. If a specified user uses the correct password, then You can log in to use the database service at any address, and the backend will store user data, experimental data, user group data, document data and other platform-related data in the database.

Further, remote communication is performed between the backend service in step (2) and the frontend service in step (3), and the two are deployed in an independent environment. Specifically, after the frontend and backend are separated, the frontend service and the backend service can be separated Run on different nodes, complete decoupling between the two, or build a highly available system through load balancing.

Furthermore, in step (2), each request received by the front-end service from the back-end service will be authenticated to prevent malicious attacks. The specific method is to assign a unique token to each user of the platform. This token It is an identity identifier. Every step of operation on the platform must verify the token to ensure that the operation is not a malicious attack.

Beneficial effects: the present invention can deploy a web-based AI teaching and training programming platform based on some common servers, which solves some problems that users often encounter when building their own AI platform, such as: difficulty in model selection, high threshold for installation and configuration , Unified user management is inconvenient, local development environment is difficult to build, etc. It is easy for novices to learn to use AI technology or experienced workers to conduct development or scientific research operations, which has a very positive effect on technical education in the AI industry.

Description of drawings

Fig. 1 is a schematic diagram of the overall process under the administrator role of the present invention;

Fig. 2 is a schematic diagram of the overall process under the teacher role of the present invention;

Fig. 3 is a schematic diagram of the overall flow of the student role in the present invention;

Fig. 4 is a schematic diagram of front-end service to administrator role function module;

Fig. 5 is a schematic diagram of front-end service to teacher's role function module;

Fig. 6 is a schematic diagram of the front-end service to the student role function module;

FIG. 7 is a schematic diagram of backend services.

Detailed ways

Below in conjunction with accompanying drawing and specific embodiment, further illustrate the present invention, should be understood that these embodiments are only for illustrating the present invention and are not intended to limit the scope of the present invention, after having read the present invention, those skilled in the art will understand various aspects of the present invention Modifications in equivalent forms all fall within the scope defined by the appended claims of this application.

The present invention proposes a method for constructing an AI teaching and training programming platform. Deploying microservices based on cloud-native clusters built on general-purpose servers enables users to interact with them through web pages, and users can perform interactive operations on the web pages. AI course learning and AI project practice, etc., all user code execution operations are delivered to the cloud native cluster through microservices, and the results are fed back to the web page for display. All code development, data set operations, mirroring operations, etc., Users can directly perform management operations on the unified page, without the need to switch environments, open multiple pages, etc., thereby realizing the construction of a one-stop AI learning and training programming environment.

As shown in Figures 1 to 3, the complete process of the present invention includes four parts: customized construction and configuration of the underlying cloud-native cluster, deployment of back-end microservices based on cloud-native, database construction, and front-end service construction. The specific implementation process is described as follows:

The customized construction and configuration of the underlying cloud-native cluster corresponds to the technical solution step (1). The specific implementation method is: first deploy the cloud-native basic environment Kubernetes cluster (hereinafter referred to as K8S); deploy the Ceph storage layer based on K8S, and after the deployment of Ceph is completed, It is necessary to set the bottom layer default storage of K8S to Ceph. Ceph is used to provide bottom layer distributed storage services for the entire K8S and various services on it; then deploy mirror management warehouse Harbor based on K8S. Harbor uses the default training environment mirror image and self-service for users. Define the training environment image to provide management functions; finally, deploy Nacos services based on K8S for service discovery of backend microservices. So far, the entire cloud native environment construction has been completed.

The construction of the back-end micro-service corresponds to step (2) of the technical solution. The specific implementation method is: first, develop the corresponding functional micro-service module according to the interface function of the front-end service. Each micro-service module is responsible for processing the corresponding different types of interface functions, such as building The AI training environment is a micro-service module, and the AI course learning management is another micro-service module; then for the configuration parameters that need to be changed dynamically, such as IP address, port number, and host list, they are uniformly extracted to the configuration file or set to the K8S environment Among the variables, configuration files and environment variables of K8S include the internal VIP of some services, database configuration, and configuration parameters of some supporting services. Each functional microservice module has its own configuration parameters. In the startup file or set to the K8S environment variable, so as to correctly start the corresponding microservice module; when processing the front-end interface function request, each request has a special method for response processing, which is controlled by the method implemented by SpringCloud. Then send the specific requirements to the corresponding underlying cluster architecture or microservices for implementation, and then feed back the running results to the interface for display; at the same time, all the data generated by the user during use, including operation logs, will be stored in the database. All back-end micro-services are deployed and run based on the cloud-native environment, which is convenient for deployment and use in new server clusters. The environment includes all the software required for back-end micro-services and the AI operating environment. All teaching and learning on the interface The practical training is based on these microservices.

The construction of the front-end service corresponds to step (3) of the technical solution. The specific implementation method is: first compile and package the front-end micro-service of the platform to the specified directory, and at the same time write its corresponding configuration into the configuration file, including the communication IP with the back-end micro-service , port and other configuration information, and achieve unified entrance access through the Nginx proxy. When the front-end micro-service starts, it will read the configuration file to realize communication with the back-end micro-service. The main function of each interface is an independent micro-application module. The whole platform is composed of multiple different micro-application modules. The administrator function mainly includes the following micro-application modules: home page, big data service, my classroom, educational administration management, Center, experimental project, user service, resource management, and operation log are nine modules in total, and each module has its own corresponding function. Teacher functions mainly include the following micro-application modules: home page, educational administration management, data center, experimental project, user service, and resource management, a total of six modules, each of which has its own corresponding functions. The student function mainly includes the following micro-application modules: Homepage, My Classroom, Data Center, Experimental Project, and Dataset Configuration. There are five modules in total, and each module has its own corresponding function. The request initiated by the front-end microservice will be processed in the following three ways:

Some independent verifications will return results directly after front-end processing;

Module function requests involving permission verification will be processed by the front end, and then sent to the back end for authentication and return the result;

Operation requests related to cloud-native clusters, such as building a training environment, will be sent to the back-end microservices for preliminary processing, and the back-end microservices will be uniformly dispatched to the cluster for execution, and the corresponding information will be fed back to the front-end interface.

The construction of the database corresponds to step (4) of the technical solution. The specific implementation method is: install and deploy the database service, and then open its user permissions. If a specified user uses the correct password, he can log in to use the database service at any address. The backend will store user data, experimental data, user group data, document data and other platform-related data in the database.

Claims (7)

1. A method for building an AI teaching and training programming platform based on a Web page, characterized in that it comprises the following steps: (1) Construct the underlying cloud-native environment, including Kubernetes cluster, Ceph storage layer, Harbor image management and Nacos service discovery; (2) The backend adopts microservice technology, deployed on the Kubernetes cluster in the form of Pod, and registered on Nacos, and performs service discovery and access through the unified gateway entrance; (3) The front end also adopts micro-service technology, and realizes unified entrance access through Nginx proxy; (4) Establish a database to store some information related to the platform, such as users and resources of the platform; Among them, remote communication can be performed between the backend service in step (2) and the frontend service in step (3), and the two are deployed in an independent environment; in step (2), the backend service receives every request from the frontend service Permission verification will be performed to prevent malicious attacks. 2. A method for constructing a web page-based AI teaching and training programming platform according to claim 1, characterized in that, in the step (1), first deploy the cloud-native basic environment Kubernetes cluster; deploy based on the Kubernetes cluster Ceph storage layer, after the deployment of Ceph, you need to set the default storage of the bottom layer of the Kubernetes cluster to Ceph, which is used to provide the underlying distributed storage service for the entire Kubernetes cluster and various services on it; and then deploy the image management warehouse based on the Kubernetes cluster Harbor, Harbor provides management functions for users to use the default training environment image and custom training environment image; finally, deploy Nacos services based on Kubernetes clusters for service discovery of back-end microservices. 3. The method for constructing a web page-based AI teaching training programming platform according to claim 1, characterized in that, in the step (2), firstly, the corresponding functional microservices are developed according to the interface functions of the front-end services module, each microservice module is responsible for processing corresponding interface functions of different types; then for the configuration parameters that need to be changed dynamically, such as IP address, port number, and host list, they are uniformly extracted into the configuration file or set to the environment variables of the Kubernetes cluster, and the configuration The file and the environment variables of the Kubernetes cluster include the internal VIP of some services, the database configuration, and the configuration parameters of some supporting services. Each functional microservice module has its own configuration parameters. Finally, the parameters corresponding to all microservices are configured in the startup file In or set to the Kubernetes cluster environment variable, so as to correctly start the corresponding microservice module; when processing the front-end interface function request, each request has a special method for response processing, which is controlled by the method implemented by SpringCloud, and then the The specific requirements are sent to the corresponding underlying cluster architecture or microservices for implementation, and then the running results are fed back to the interface for display; at the same time, all data generated by the user during use, including operation logs, are stored in the database. 4. The method for building a web page-based AI teaching and training programming platform according to claim 1, characterized in that in the step (3), first compile and package the front-end microservices of the platform to a specified directory, At the same time, write its corresponding configuration into the configuration file, including communication IP and port configuration information with the back-end microservice, and realize unified entrance access through the Nginx proxy. When the front-end microservice starts, it will read the configuration file, so as to realize the Communication between micro-services; the main function of each interface is an independent micro-application module, and the entire platform is composed of multiple different micro-application modules. The administrator function includes the following micro-application modules: home page, big data service, my Classroom, educational administration management, data center, experimental project, user service, resource management, and operation log are nine modules in total, and each module has its own corresponding functions; teacher functions include the following micro-application modules: home page, educational administration management, data center, There are six modules in total including experimental project, user service and resource management, and each module has its own corresponding function; the student function includes the following micro-application modules: Homepage, My Classroom, Data Center, Experimental Project and Dataset Configuration, a total of five modules , each module has its own corresponding function. 5. The method for constructing a web page-based AI teaching training programming platform according to claim 1, characterized in that, in the step (4), each microservice corresponds to a database, and between different databases Isolate each other, install and deploy database services, and then open its user permissions. If a specified user uses the correct password, you can log in to use the database service at any address. The backend will store user data, experimental data, and user group data. , document data and other platform-related data are stored in the database. 6. A method for building a web page-based AI teaching training programming platform according to claim 1, characterized in that the back-end service in step (2) and the front-end service in step (3) perform remote communication , the two are deployed in an independent environment. Specifically, after the front-end and back-end are separated, the front-end service and the back-end service can run independently on different nodes, and the decoupling between the two is completed, or a high-speed available systems. 7. The method for constructing a web page-based AI teaching training programming platform according to claim 1, characterized in that, in step (2), each request received by the back-end service from the front-end service will perform authority verification, To prevent malicious attacks, the specific method is that each user of the platform will be assigned a unique token, which is an identity identifier, and every step of operation on the platform must verify the token to ensure that the operation is not malicious attack.

Images