Cloud service method of computational analysis class model
Through the cloud service method of computing and analysis models, the model resource service platform is used to realize the cloud encapsulation, deployment, management and call of the model, which solves the problems of poor reusability, low resource degree and high system coupling of the existing models. , improves the utilization rate and development economics of model resources.
Patent Information
- Application Number
- CN202411527589.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-30
- Publication Date
- 2025-02-21
AI Technical Summary
In actual application, existing computing analysis models have problems such as poor reusability, low resource degree, high system coupling, difficulty in updating and maintaining, and difficulty in collaboration between different models, resulting in waste of resources, low utilization rate and high development costs. .
A cloud service method for computing and analysis models is proposed, including model cloud encapsulation, model push deployment, model cloud management, model resource discovery, task-oriented model call orchestration, model service cloud call, model resource readiness audit and service status monitoring. The model registration, deployment, management and call is realized through the model resource service platform, supporting model reuse and collaboration.
It improves the reusability and resource utilization of the computational analysis model, reduces development costs, simplifies the update and maintenance of models, supports the needs of complex tasks, and realizes the wide application of model resources and system decoupling.
Smart Images
Figure CN119496811A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of cloud service for computing and analyzing models, and in particular to a cloud service method for computing and analyzing models. Background Art
[0002] In today's era of rapid digital development, the demand for computational analysis models in various fields is growing. With the continuous advancement of technology, a large number of computational analysis models have been developed. However, these models face many challenges in practical applications. On the one hand, many models have the problem of poor reusability. Different development teams may repeatedly develop similar models, resulting in a waste of resources. At the same time, due to the lack of a unified management and deployment method, models are difficult to reuse on different platforms, limiting their scope of application. On the other hand, the degree of resourceization of models is not high. After development, many models are often confined to specific projects or systems and cannot be widely used as a shared resource. This not only reduces the utilization rate of the model, but also increases the cost of developing new models. In addition, traditional model application methods are usually independent, and the degree of coupling between systems is high. This makes it difficult to update and maintain the model. Once a model changes, it may affect multiple systems related to it. At the same time, collaboration between different models is also difficult, and it is difficult to meet the needs of complex tasks.
[0003] In order to solve these problems and improve the utilization rate of model resources and the economic efficiency of development, a new cloud service method for computing and analyzing models is urgently needed. Summary of the invention
[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions: a cloud service method for computing and analyzing models, including model cloud encapsulation, model push deployment, model cloud management, model resource discovery, task-oriented model call orchestration, model service cloud call and model resource readiness review and service status monitoring.
[0005] Preferably, the model cloud packaging includes cloud registration and publishing of the model. The cloud registration and publishing of the model is achieved by registering with a model resource service platform, packaging the model into a platform-independent, arbitrarily deployable shared resource, and obtaining a unique identifier.
[0006] Preferably, the model push deployment is used to deploy model resources on a resource service platform, and various information resources such as corresponding computing, storage, and bandwidth are allocated.
[0007] Preferably, the model cloud management is the maintenance and management of registered models, including searching and discovering specific model resources through the model resource service platform to apply for shared applications, manage background information resource allocation, etc.
[0008] Preferably, the model resource discovery searches and discovers model resources through a model resource service platform to apply for shared applications.
[0009] Preferably, the task-oriented model call orchestration is used to construct a task-oriented model application process, express the order, branching, loop and union or relationship of tasks, adapt the model call data interface, and publish the task execution process in the form of a service.
[0010] Preferably, the model service cloud call is a process in which a user uses a model by operating a simple remote call interface or development kit, automatically pushing model input data to the model, returning calculation results, and the calculation results will be pushed to the user as a data product.
[0011] Preferably, the model resource readiness review and service status monitoring is for the model user to check whether the corresponding model resources are ready for service, so as to avoid the situation where the model resources are not in place or the service is unavailable.
[0012] Preferably, it also includes establishing a distributed deployment environment for model management and services based on the model resource service platform, specifically including: (1) Model registration management: The model resource service platform is used to register and manage the entire life cycle of model resource construction, management, use and evaluation. In the model construction stage, the model resource service platform is responsible for collecting and managing model construction requirements and model development status. In the model management stage, the model resource service platform is used to encapsulate the software program model into the library through the registration management service, install and deploy it to servers, container clouds, virtualized clouds and other environments, build model services and rule services, register them to the model resource service platform, and monitor the model operation. In the model application stage, the model directory-based model browsing and retrieval, model parameter configuration, model effect testing, model service online scheduling and model installation package offline download functions are provided, and three integration modes such as full model integration, semi-integration and plug-in are supported. In the model evaluation stage, the model registration service supports recording the usage feedback of users at all levels and types, and can adjust the model retrieval results according to the usage feedback. (2) Model discovery and scheduling: The model resource service platform provides three model discovery and scheduling services, namely, model directory service, model scheduling service and model theme service. The model directory service supports the model resource service platform to browse model information, locate model computing resources and obtain model computing results through the model directory service; the model scheduling service distributes and deploys models, so that general or special domain models can be deployed on servers, containers and cloud environments, and registered with the model directory service, and regularly send heartbeat information to the model directory to ensure status synchronization; model data resources can be registered as theme resources through the model theme service, and model computing can obtain data resources through the data bus for computing without sensing, and push back the computing results to users. The model directory service, model scheduling service and model theme service can be published to the outside world through the model portal according to rights and needs; (3) Model combination service: The model combination application service capability is provided through the model organization engine and model integration interaction. The model organization engine can efficiently and flexibly organize the call of basic models and provide model support for plan formulation and solution generation. The model integration interaction is oriented to model users and application systems. It uses the model integration framework and service development framework to integrate desktop, access scheduling and data management functions, integrate model data, schedule model access, and realize unified access to model services and model data. It supports unified access and services for external model users and application systems. The model combination service has visual model combination, combination model recommendation and functions and model process engine, etc., and supports rapid combination of model visualization based on specific tasks, programmatic operation of combination models and comprehensive monitoring of operation status. (4) Model computing application: Organize the computing application of the model according to the computing meta-model, provide model computing output for model discovery scheduling and combination services, support the model service application of the front-end and back-end, call the model through cloud services and input relevant model parameters, directly obtain the front-end display of the calculation results and the packaging and storage of data products. Model computing application depends on model operation parameters, which usually include two types: calculation parameters and ad hoc parameters. The calculation parameters are substituted into the data-driven algorithm calculation with specific meanings, and its parameters are required to use data resources as much as possible to solve the authority, timeliness and timeliness of the parameters. In the subsequent construction of supporting data resources, the existing data system and the construction results of other related projects are introduced. Ad hoc parameters refer to the data that needs to be specified, such as time, place, event, and target, which are usually accurately input by the model user when the model is called. The calculation results of the model can be pushed to the front-end user or application system as a specific data product, or stored in the data resource pool and data subject library to provide input data for other model applications with calling relationships.
[0013] The present invention provides a cloud service method for computing and analyzing models, and proposes a cloud service application framework and process for a large number of reusable, resource-based, and system-decoupled computing and analyzing models, providing a universal method for the resource-based application of specific fields or general models, thereby improving the utilization rate of model resources and the economic efficiency of development. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 A basic flow chart of the model resource service application of the present invention; Figure 2 This is a schematic diagram of the calculation application of the model of the present invention. DETAILED DESCRIPTION
[0015] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0016] See also Figure 1-Figure 2 The present invention provides a technical solution: a cloud service method for computing and analyzing models. The basic logical flow of the model service application of the present invention includes the following seven steps: Step 1: Model cloud packaging, including cloud registration and publishing of the model. Cloud registration and publishing of the model is to register with the model resource service platform, encapsulate the model into a shared resource that is independent of the platform and can be deployed arbitrarily, and obtain a unique identifier; Step 2: Model push deployment, deploying model resources on the resource service platform, and allocating corresponding computing, storage, bandwidth and other information resources (distributing model resources in the form of images to complete the initialization of model services); Step 3: Model cloud management is the maintenance and management of registered models, including searching and discovering specific model resources through the model resource service platform to apply for shared applications, manage backend information resource allocation, etc. (as well as provide automatic recovery, backup disaster recovery, high availability, horizontal expansion and other capabilities for applications); Step 4: Model resource discovery: search and discover model resources through the model resource service platform to apply for shared applications; Step 5: Task-oriented model call orchestration, build task-oriented model application process, express the order, branching, loop and union / OR relationship of tasks, adapt the model call data interface, and publish the task execution process in the form of service; Step 6: Model service cloud call, the user uses the model through a simple remote call interface or development kit, automatically pushes model input data to the model, returns the calculation results, and the calculation results will be pushed to the user as a data product; Step 7: Model resource readiness review and service status monitoring. Model users check whether the corresponding model resources are ready for service to avoid model resources being in place or service being unavailable. Based on the basic process of the above model service application, relying on the model resource service platform, a distributed deployment environment for model management and services is established, as follows: (1) Model registration management Relying on the model resource service platform, we register and manage the entire life cycle of model resource construction, management, use and evaluation, and effectively ensure the authority, accuracy and credibility of the model through quality management and control of model results at each stage; In the model construction stage, the model resource service platform is responsible for collecting and managing model construction requirements and model development status, and timely grasping the model development status; in the model management stage, relying on the model code generator and other software tools provided by the model resource management platform, the software program model is encapsulated into the library through the registration management service, installed and deployed to servers, container clouds, virtualized clouds and other environments, and model services and rule services are built. Register to the model resource service platform and monitor the model operation; in the model application stage, cloud service technology is used to provide model browsing and retrieval based on the model directory, model parameter configuration, model effect testing, model service online scheduling and model installation package offline download functions, and support three integration modes such as full model integration, semi-integration and plug-in; in the model evaluation stage, the model registration service supports recording the usage feedback of users at all levels and types, and can adjust the model retrieval results according to the usage feedback; (2) Model Discovery Scheduling The model resource service platform provides three model discovery and scheduling services, namely model directory service, model scheduling service and model theme service. The model directory service supports the model resource service platform to browse model information, locate model computing resources and obtain model computing results through the model directory service; the model scheduling service is deployed through model distribution, and general or special domain models can be deployed on servers, containers and cloud environments, and registered with the model directory service, and heartbeat information is regularly exchanged with the model directory to ensure status synchronization; model data resources can be registered as theme resources through the model theme service, and model calculations can obtain data resources through the data bus for calculation without sensing, and push back the calculation results to users. The model directory service, model scheduling service and model theme service can be published to the outside world through the model portal according to rights and needs; (3) Model combination service The model combination service provides the model combination application service capability through the model organization engine and model integration interaction. The model organization engine can efficiently and flexibly organize the call of basic models to provide model support for plan formulation and solution generation. The model integration interaction is oriented to model users and application systems. It uses the model integration framework and service development framework to integrate desktop, access scheduling and data management functions, integrate model data, schedule model access, and unify model services and model data access, supporting unified access and services for external model users and application systems; The model combination service has functions such as visual model combination, combination model recommendation and model process engine, etc., which supports rapid combination of model visualization based on specific tasks, programmed operation of combination models and comprehensive monitoring of operation status; (4) Model calculation application Model computing applications provide model computing outputs for model discovery scheduling and combination services, support front-end and back-end model service applications, call models through cloud services and input relevant model parameters, directly obtain front-end display of computing results and data product packaging and storage; Model calculation applications rely on model operation parameters, which usually include calculation parameters and ad hoc parameters. The calculation parameters are substituted into the data-driven algorithm calculation with specific meanings, and the parameters are required to use data resources as much as possible to solve the authority, timeliness and timeliness of the parameters, and to connect the existing data system and the construction results of other related projects in the subsequent construction of supporting data resources. Ad hoc parameters refer to the data that needs to be specified, such as time, place, event, and target, which are usually accurately input by the model user when the model is called; The calculation results of the model can be pushed to front-end users or application systems as specific data products, or stored in the data resource pool and data subject library to provide input data for other model applications that have a calling relationship.
[0017] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field and related fields without creative work should fall within the scope of protection of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention are implemented according to the conventional means in the field unless otherwise specified and limited.
Claims
1. A cloud service method for computing and analyzing models, characterized in that: It includes model cloud packaging, model push deployment, model cloud management, model resource discovery, task-oriented model call orchestration, model service cloud call, model resource readiness review and service status monitoring.
2. A cloud service method for computing and analyzing models according to claim 1, characterized in that: The model cloud packaging includes cloud registration and publishing of the model. The cloud registration and publishing of the model is achieved by registering with the model resource service platform, packaging the model into a shared resource that is independent of the platform and can be deployed arbitrarily, and obtaining a unique identifier.
3. The cloud service method for computing and analyzing models according to claim 1, characterized in that: The model push deployment is used to deploy model resources on the resource service platform, and allocate corresponding computing, storage, bandwidth and other information resources.
4. The cloud service method for computing and analyzing models according to claim 1, characterized in that: The model cloud management is the maintenance and management of registered models, including searching and discovering specific model resources through the model resource service platform to apply for shared applications, manage background information resource allocation, etc.
5. The cloud service method for computing and analyzing models according to claim 1, characterized in that: The model resource discovery searches and discovers model resources through the model resource service platform to apply for shared applications.
6. The cloud service method for computing and analyzing models according to claim 1, characterized in that: The task-oriented model call orchestration is used to build a task-oriented model application process, express the order, branching, loop and union or relationship of tasks, adapt the model call data interface, and publish the task execution process in the form of a service.
7. The cloud service method for computing and analyzing models according to claim 1, characterized in that: The model service cloud call is a process in which users use the model by operating a simple remote call interface or development kit, automatically pushing model input data to the model, returning calculation results, and the calculation results will be pushed to users as a data product.
8. The cloud service method for computing and analyzing models according to claim 1, characterized in that: The model resource readiness review and service status monitoring are for model users to check whether the corresponding model resources are ready for service, to avoid model resources being in place or services being unavailable.
9. The cloud service method for computing and analyzing models according to claim 1, characterized in that: It also includes establishing a distributed deployment environment for model management and services based on the model resource service platform, including: (1) Model registration management: The model resource service platform is used to register and manage the entire life cycle of model resource construction, management, use and evaluation. In the model construction stage, the model resource service platform is responsible for collecting and managing model construction requirements and model development status. In the model management stage, the model resource service platform is used to encapsulate the software program model into the library through the registration management service, install and deploy it to servers, container clouds, virtualized clouds and other environments, build model services and rule services, register them to the model resource service platform, and monitor the model operation. In the model application stage, the model directory-based model browsing and retrieval, model parameter configuration, model effect testing, model service online scheduling and model installation package offline download functions are provided, and three integration modes such as full model integration, semi-integration and plug-in are supported. In the model evaluation stage, the model registration service supports recording the usage feedback of users at all levels and types, and can adjust the model retrieval results according to the usage feedback. (2) Model discovery and scheduling: The model resource service platform provides three model discovery and scheduling services, namely, model directory service, model scheduling service and model theme service. The model directory service supports the model resource service platform to browse model information, locate model computing resources and obtain model computing results through the model directory service; the model scheduling service distributes and deploys models, so that general or special domain models can be deployed on servers, containers and cloud environments, and registered with the model directory service, and regularly send heartbeat information to the model directory to ensure status synchronization; model data resources can be registered as theme resources through the model theme service, and model computing can obtain data resources through the data bus for computing without sensing, and push back the computing results to users. The model directory service, model scheduling service and model theme service can be published to the outside world through the model portal according to rights and needs; (3) Model combination service: The model combination application service capability is provided through the model organization engine and model integration interaction. The model organization engine can efficiently and flexibly organize the call of basic models and provide model support for plan formulation and solution generation. The model integration interaction is oriented to model users and application systems. It uses the model integration framework and service development framework to integrate desktop, access scheduling and data management functions, integrate model data, schedule model access, and realize unified access to model services and model data. It supports unified access and services for external model users and application systems. The model combination service has visual model combination, combination model recommendation and functions and model process engine, etc., and supports rapid combination of model visualization based on specific tasks, programmatic operation of combination models and comprehensive monitoring of operation status. (4) Model computing application: Organize the computing application of the model according to the computing meta-model, provide model computing output for model discovery scheduling and combination services, support the model service application of the front-end and back-end, call the model through cloud services and input relevant model parameters, directly obtain the front-end display of the calculation results and the packaging and storage of data products. Model computing application depends on model operation parameters, which usually include two types: calculation parameters and ad hoc parameters. The calculation parameters are substituted into the data-driven algorithm calculation with specific meanings, and its parameters are required to use data resources as much as possible to solve the authority, timeliness and timeliness of the parameters. In the subsequent construction of supporting data resources, the existing data system and the construction results of other related projects are introduced. Ad hoc parameters refer to the data that needs to be specified, such as time, place, event, and target, which are usually accurately input by the model user when the model is called. The calculation results of the model can be pushed to the front-end user or application system as a specific data product, or stored in the data resource pool and data subject library to provide input data for other model applications with calling relationships.