Algorithm mirror image generation method based on container service

Abstract

The invention relates to mirror image generation, in particular to a container service-based algorithm mirror image generation method, which comprises the following steps of: carrying out business dimension classification on algorithms deployed by a meteorological big data cloud platform, carding operation dependency of each algorithm and constructing an algorithm dependency matrix; determining a mirror image hierarchical architecture, and constructing a hierarchical weight coefficient according to the function boundary and dependency relationship of each layer to measure the resource proportion and update priority of each layer in mirror image creation; extracting core calculation logic of the algorithm, separating a fixed calculation framework of the algorithm from configurable parameters, constructing a parameter configuration mapping model, and performing customized packaging on a core layer of the algorithm; packaging the triggering condition of the algorithm in the service adaptation layer, constructing a triggering priority model, and performing customized packaging on the service adaptation layer; the method can effectively overcome the defects that effective classification and dependent carding of service dimensions are lacked, mirror image construction and management are lacked scientific and reasonable architecture and strategy, and a mirror image verification and distribution mechanism is not perfect.

Description

Technical Field

[0001] This invention relates to image generation, and more specifically to an algorithm image generation method based on container services. Background Technology

[0002] In the development of meteorological big data cloud platforms, the efficient deployment and stable operation of algorithms are key factors in improving meteorological service capabilities. Traditional deployment methods for meteorological operational algorithms have many drawbacks and are unable to meet the ever-growing operational needs of meteorological big data cloud platforms.

[0003] On the one hand, traditional deployment methods lack effective classification and dependency analysis of business dimensions. There are many types of meteorological business algorithms and complex business scenarios. Different algorithms have different dependencies on various software, hardware resources and data during operation. If they are not clearly classified and analyzed, dependency gaps or conflicts are likely to occur during deployment, causing the algorithms to fail to operate normally and affecting the accuracy and timeliness of meteorological business processing.

[0004] On the other hand, image building and management lack a scientific and reasonable architecture and strategy. Traditional image building often fails to fully consider the functional boundaries and dependencies of each layer, resulting in a chaotic image structure, unreasonable resource allocation, and difficulties in updating and maintaining. At the same time, there is a lack of effective methods for the core computational logic and parameter configuration of algorithms, as well as business adaptation, which limits the flexibility and scalability of algorithms.

[0005] Furthermore, the image verification and distribution mechanism is inadequate. The lack of a multi-dimensional verification mechanism makes it difficult to ensure image quality, detect potential problems in a timely manner, and, during image distribution, it fails to reasonably calculate distribution priorities based on the business load of computing nodes, hindering the elastic deployment of algorithms and preventing the full utilization of the meteorological big data cloud platform's computing resources, thus restricting the overall performance improvement of the meteorological big data cloud platform. Summary of the Invention

[0006] (a) Technical problems to be solved In view of the above-mentioned shortcomings of the existing technology, the present invention provides an algorithm image generation method based on container services, which can effectively overcome the deficiencies of the existing technology, such as the lack of effective classification and dependency sorting of business dimensions, the lack of scientific and reasonable architecture and strategy for image building and management, and the imperfect image verification and distribution mechanism.

[0007] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: The method for generating algorithm images based on container services includes the following steps: S1. Classify the algorithms deployed on the meteorological big data cloud platform according to business dimensions, sort out the operation dependencies of each algorithm, and construct an algorithm dependency matrix; S2. Determine the image hierarchy architecture. Based on the functional boundaries and dependencies of each layer, construct layer weight coefficients to measure the resource share and update priority of each layer in image creation. S3. Extract the core computational logic of the algorithm, separate the fixed computational framework of the algorithm from the configurable parameters, construct a parameter configuration mapping model, and customize the encapsulation of the core layer of the algorithm. S4. Encapsulate the algorithm's triggering conditions in the business adaptation layer, construct a triggering priority model, and perform customized encapsulation of the business adaptation layer. S5. Introduce a multi-dimensional verification mechanism, construct a verification pass rate evaluation function to quantify the verification pass rate of the image, and push the verified images to the image library to establish a full lifecycle management ledger for image versions. S6. Construct a regional distribution node network for images based on the image library, calculate the distribution priority of images according to the business load of each computing node, and perform elastic deployment of the algorithm.

[0008] Preferably, in S1, the algorithms deployed on the meteorological big data cloud platform are categorized by business dimensions, and the operational dependencies of each algorithm are analyzed to construct an algorithm dependency matrix, including: The algorithms deployed on the meteorological big data cloud platform are classified according to business dimensions, including meteorological numerical forecasting algorithms, meteorological element assimilation algorithms, and meteorological disaster early warning algorithms, and business attribute tags are labeled for each algorithm; The runtime dependencies of each algorithm are identified, including basic environment dependencies, specialized tool dependencies, and data dependencies, and an algorithm dependency matrix M is constructed, with matrix elements m. ij satisfy: .

[0009] Preferably, in S2, the image hierarchy architecture is determined, and a layer weight coefficient is constructed based on the functional boundaries and dependencies of each layer to measure the resource share and update priority of each layer in image creation, including: The image hierarchy architecture is defined, with the following layers from bottom to top: basic environment layer, professional tools layer, algorithm core layer, and business adaptation layer. Each layer has independent update and reuse capabilities, and the layers interact through standardized interfaces. Based on the functional boundaries and dependencies of each layer, a layer weight coefficient is constructed to measure the resource share and update priority of each layer in image creation: ; in, The weight coefficients of the k-th layer in the mirror image are... k=1,2,3,4, the weight coefficients of the basic environment layer The adaptation algorithm is highly dependent on the basic environment, and the weight coefficients of the business adaptation layer... To ensure lightweight business configuration, the weight coefficient of the professional tool layer. Weight coefficients of the algorithm's core layer The algorithm dynamically allocates resources based on the degree of dependence on specialized tools.

[0010] Preferably, in S3, the core computational logic of the algorithm is extracted, the fixed computational framework of the algorithm is separated from the configurable parameters, and a parameter configuration mapping model is constructed, including: The core computational logic of the algorithm is extracted, separating the algorithm's fixed computational framework from its configurable parameters, including grid resolution. Time step and assimilated data weights Build a parameter configuration mapping model : ; in, For grid resolution The corresponding computational resource consumption function, For time step The corresponding iterative efficiency function, Weighting the assimilated data The corresponding forecast accuracy function, , , All are weighting coefficients, and , To compensate for parameter mismatch and ensure the accuracy of parameter configuration; Parameter configuration mapping model It can match the grid resolution according to the computing resources and operational accuracy requirements of the meteorological big data cloud platform. Time step Assimilation data weights and weighting coefficients , , The optimal combination of parameters, including those for the target.

[0011] Preferably, the algorithm core layer in S3 is customized and encapsulated, including: The input and output interfaces of the encapsulated algorithm are uniformly connected to the data storage system of the meteorological big data cloud platform. The input interface supports standardized parsing of multi-source heterogeneous data, including gridded data, station observation data and satellite remote sensing data. The output interface supports the synchronous generation of visualized metadata of forecast results and business alarm instructions, ensuring seamless connection between the container and the platform data link. An embedded status monitoring module is used to monitor the running status of real-time monitoring algorithms. The monitoring indicators include computing resource utilization, iteration convergence speed, and data processing latency. Threshold alarm rules are set, and when the monitoring indicators exceed the preset range, resource scheduling requests or fault tolerance mechanisms within the container are automatically triggered.

[0012] Preferably, in S4, the triggering conditions of the algorithm are encapsulated in the business adaptation layer, and a triggering priority model is constructed, including: The triggering conditions of the algorithm are encapsulated in the business adaptation layer, including timed triggering conditions, data arrival triggering conditions, and manual instruction triggering conditions, and a triggering priority model Q(t,p,s) is constructed: ; Where t is the time urgency coefficient, The closer to the forecast time, the greater the urgency coefficient, where p is the business priority coefficient. Meteorological disaster early warning services rank higher than meteorological numerical forecasting services, where 's' is the data completeness coefficient. The more complete the data, the higher the data completeness coefficient. , , All are weighting coefficients, and ; The meteorological big data cloud platform can prioritize the scheduling algorithm running in the container based on the size of the Q value.

[0013] Preferably, the business adaptation layer in S4 is customized and encapsulated, including: The business adaptation layer is connected to the business process orchestration interface of the meteorological big data cloud platform, and the permission authentication module of the meteorological big data cloud platform is integrated. The identity authentication and operation audit logic of the algorithm operation is embedded, allowing only authorized business accounts of the platform to trigger the start, pause and termination operations of the container. At the same time, the algorithm's operation log and data access log are recorded to achieve security control of the business. Adapt to the computing power scheduling strategy of the meteorological big data cloud platform. Based on the computing power requirement tags of various algorithms, configure the computing power application template of the container in the business adaptation layer. When the container starts, it automatically applies to the meteorological big data cloud platform for the corresponding type of computing power resources, so as to achieve precise matching between the computing power requirements of the algorithm and the computing power supply of the platform.

[0014] Preferably, S5 introduces a multi-dimensional verification mechanism, constructs a verification pass rate evaluation function to quantify the verification pass rate of the image, and pushes the verified images to the image library, establishing a full lifecycle management ledger for image versions, including: A multi-dimensional verification mechanism is designed, including environmental compatibility verification, algorithm functionality verification, and platform adaptability verification. A verification pass rate evaluation function V(e1,e2,e3) is constructed to quantify the verification pass rate of the image. ; Where e1 is the percentage of anomalies in environment compatibility verification, e2 is the error value in algorithm functionality verification, e3 is the percentage of failures in platform adaptability verification, and E 1max E 2max E 3max These are the maximum allowed values ​​for each validation dimension; Images with a V value greater than the preset pass threshold are considered to be valid images. Validated images are then version-fixed, and composite tags containing algorithm type, business attributes, and pass threshold are added to the images. The images are then pushed to the image library of the meteorological big data cloud platform to establish a full lifecycle management ledger for image versions.

[0015] Preferably, in S6, a regional distribution node network for images is built based on the image library, and the distribution priority of images is calculated according to the service load of each computing node, including: A regional distribution node network for images is built based on the image library, and the distribution priority D of the images is calculated according to the business load of each computing node. ; Where L represents the service load of the computing node, i.e., the current computing power utilization rate of the computing node. S represents the urgency of the business corresponding to the mirror image. , , All are weighting coefficients, and ; Based on the D value, the images are pre-distributed sequentially from high to low to ensure that business nodes can quickly pull the images to start the algorithm.

[0016] Preferably, S6 includes the following: Flexible deployment of the algorithm The elastic scheduling system, which connects to the meteorological big data cloud platform, will handle situations where the volume of business requests R exceeds a preset maximum request threshold R. max When the container is automatically expanded, the expansion operation is triggered, and the expansion quantity N satisfies a linear correlation model: ; Where k is the expansion coefficient, which is determined based on the parallelism of the algorithm. , This represents the floor function; When the number of business requests R is less than the preset minimum request threshold R min When the container is automatically recycled, the computing resources are dynamically optimized.

[0017] (III) Beneficial Effects Compared with existing technologies, the algorithm image generation method based on container services provided by this invention has the following advantages: 1) Precise hierarchical architecture, optimized resource management By determining the image hierarchy architecture and constructing layer weight coefficients, the resource proportion and update priority of each layer in image creation can be accurately measured based on the functional boundaries and dependencies of each layer. This allows for the reasonable allocation of computing, storage and other resources when building images, avoiding resource waste or insufficiency. This image hierarchy architecture effectively improves resource utilization efficiency and reduces resource management costs. 2) Comprehensive verification and control to ensure image quality. By introducing a multi-dimensional verification mechanism and constructing a verification pass rate evaluation function, the image can be comprehensively and quantitatively verified. By evaluating the image from multiple dimensions, problems and potential risks in the image can be identified in a timely manner, ensuring that only verified images can be pushed to the image library. At the same time, a full lifecycle management ledger for image versions is established to track and manage the entire process of image creation, updates, and usage, which facilitates the timely detection and resolution of quality issues. This comprehensive verification and control approach effectively guarantees the quality of the images and improves the stability and reliability of the meteorological big data cloud platform. 3) Intelligent and flexible deployment to improve deployment efficiency By building a regional distribution node network based on the image library and calculating the distribution priority of images according to the business load of each computing node, the algorithm can be deployed elastically. It can dynamically adjust image distribution and container capacity according to actual business needs and computing resources. This intelligent elastic deployment method improves the flexibility and adaptability of algorithm deployment and greatly enhances the overall deployment efficiency. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0019] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0021] The following describes the specific process of the container service-based algorithm image generation method provided by this invention, using concrete examples (e.g.) Figure 1 (as shown) and technical effects.

[0022] S1. Classify the algorithms deployed on the meteorological big data cloud platform according to business dimensions, and sort out the runtime dependencies of each algorithm to construct an algorithm dependency matrix, including: The algorithms deployed on the meteorological big data cloud platform are classified according to business dimensions, including meteorological numerical forecasting algorithms, meteorological element assimilation algorithms, and meteorological disaster early warning algorithms, and business attribute tags are labeled for each algorithm;

[0023] The runtime dependencies of each algorithm are identified, including basic environment dependencies, specialized tool dependencies, and data dependencies, and an algorithm dependency matrix M is constructed, with matrix elements m. ij satisfy: .

[0024] S2. Determine the image hierarchy architecture. Based on the functional boundaries and dependencies of each layer, construct layer weight coefficients to measure the resource share and update priority of each layer in image creation, including: The image hierarchy architecture is defined, with the following layers from bottom to top: basic environment layer, professional tools layer, algorithm core layer, and business adaptation layer. Each layer has independent update and reuse capabilities, and the layers interact through standardized interfaces. Based on the functional boundaries and dependencies of each layer, a layer weight coefficient is constructed to measure the resource share and update priority of each layer in image creation: ;

[0025] in, The weight coefficients of the k-th layer in the mirror image are... k=1,2,3,4, the weight coefficients of the basic environment layer The adaptation algorithm is highly dependent on the basic environment, and the weight coefficients of the business adaptation layer... To ensure lightweight business configuration, the weight coefficient of the professional tool layer. Weight coefficients of the algorithm's core layer The algorithm dynamically allocates resources based on the degree of dependence on specialized tools.

[0026] The above technical solution, by determining the image hierarchy architecture and constructing layer weight coefficients, can accurately measure the resource proportion and update priority of each layer in image creation based on the functional boundaries and dependencies of each layer. This allows for the reasonable allocation of computing, storage and other resources when building an image, avoiding resource waste or insufficiency. This image hierarchy architecture effectively improves resource utilization efficiency and reduces resource management costs.

[0027] S3. Extract the core computational logic of the algorithm, separate the fixed computational framework of the algorithm from the configurable parameters, construct a parameter configuration mapping model, and customize the core layer of the algorithm.

[0028] 1) Extract the core computational logic of the algorithm, separate the fixed computational framework from the configurable parameters, and construct a parameter configuration mapping model, including: The core computational logic of the algorithm is extracted, separating the algorithm's fixed computational framework from its configurable parameters, including grid resolution. Time step and assimilated data weights Build a parameter configuration mapping model : ; in, For grid resolution The corresponding computational resource consumption function, For time step The corresponding iterative efficiency function, Weighting the assimilated data The corresponding forecast accuracy function, , , All are weighting coefficients, and , To compensate for parameter mismatch and ensure the accuracy of parameter configuration; Parameter configuration mapping model It can match the grid resolution according to the computing resources and operational accuracy requirements of the meteorological big data cloud platform. Time step Assimilation data weights and weighting coefficients , , The optimal combination of parameters, including those for the target.

[0029] 2) Customize and encapsulate the core algorithm layer, including: The input and output interfaces of the encapsulated algorithm are uniformly connected to the data storage system of the meteorological big data cloud platform. The input interface supports standardized parsing of multi-source heterogeneous data, including gridded data, station observation data and satellite remote sensing data. The output interface supports the synchronous generation of visualized metadata of forecast results and business alarm instructions, ensuring seamless connection between the container and the platform data link. An embedded status monitoring module is used to monitor the running status of real-time monitoring algorithms. The monitoring indicators include computing resource utilization, iteration convergence speed, and data processing latency. Threshold alarm rules are set, and when the monitoring indicators exceed the preset range, resource scheduling requests or fault tolerance mechanisms within the container are automatically triggered.

[0030] S4. Encapsulate the algorithm's triggering conditions in the business adaptation layer, construct a triggering priority model, and perform customized encapsulation of the business adaptation layer.

[0031] 1) Encapsulate the algorithm's triggering conditions in the business adaptation layer and construct a triggering priority model, including: The triggering conditions of the algorithm are encapsulated in the business adaptation layer, including timed triggering conditions, data arrival triggering conditions, and manual instruction triggering conditions, and a triggering priority model Q(t,p,s) is constructed: ; Where t is the time urgency coefficient, The closer to the forecast time, the greater the urgency coefficient, where p is the business priority coefficient. Meteorological disaster early warning services rank higher than meteorological numerical forecasting services, where 's' is the data completeness coefficient. The more complete the data, the higher the data completeness coefficient. , , All are weighting coefficients, and ; The meteorological big data cloud platform can prioritize the scheduling algorithm running in the container based on the size of the Q value.

[0032] 2) Customize and encapsulate the business adaptation layer, including: The business adaptation layer is connected to the business process orchestration interface of the meteorological big data cloud platform, and the permission authentication module of the meteorological big data cloud platform is integrated. The identity authentication and operation audit logic of the algorithm operation is embedded, allowing only authorized business accounts of the platform to trigger the start, pause and termination operations of the container. At the same time, the algorithm's operation log and data access log are recorded to achieve security control of the business. Adapt to the computing power scheduling strategy of the meteorological big data cloud platform. Based on the computing power requirement tags of various algorithms, configure the computing power application template of the container in the business adaptation layer. When the container starts, it automatically applies to the meteorological big data cloud platform for the corresponding type of computing power resources, so as to achieve precise matching between the computing power requirements of the algorithm and the computing power supply of the platform.

[0033] S5. Introduce a multi-dimensional verification mechanism, construct a verification pass rate evaluation function to quantify the verification pass rate of images, and push verified images to the image repository to establish a full lifecycle management ledger for image versions, including: A multi-dimensional verification mechanism is designed, including environmental compatibility verification, algorithm functionality verification, and platform adaptability verification. A verification pass rate evaluation function V(e1,e2,e3) is constructed to quantify the verification pass rate of the image. ; Where e1 is the percentage of anomalies in environment compatibility verification, e2 is the error value in algorithm functionality verification, e3 is the percentage of failures in platform adaptability verification, and E 1max E 2max E 3max These are the maximum allowed values ​​for each validation dimension; Images with a V value greater than the preset pass threshold are considered to be valid images. Validated images are then version-fixed, and composite tags containing algorithm type, business attributes, and pass threshold are added to the images. The images are then pushed to the image library of the meteorological big data cloud platform to establish a full lifecycle management ledger for image versions.

[0034] The above technical solution introduces a multi-dimensional verification mechanism and constructs a verification pass rate evaluation function, which can comprehensively and quantitatively verify the images. By evaluating the images from multiple dimensions, problems and potential risks in the images can be identified in a timely manner, ensuring that only verified images can be pushed to the image library. At the same time, a full lifecycle management ledger for image versions is established to track and manage the entire process of image creation, updates, and usage, which facilitates the timely detection and resolution of quality issues. This comprehensive verification and control method effectively guarantees the quality of the images and improves the stability and reliability of the meteorological big data cloud platform.

[0035] S6. Construct a regional distribution node network for images based on the image library, calculate the distribution priority of images according to the business load of each computing node, and perform elastic deployment of the algorithm.

[0036] 1) Construct a regional distribution node network for images based on the image library, and calculate the distribution priority of images according to the business load of each computing node, including: A regional distribution node network for images is built based on the image library, and the distribution priority D of the images is calculated according to the business load of each computing node. ; Where L represents the service load of the computing node, i.e., the current computing power utilization rate of the computing node. S represents the urgency of the business corresponding to the mirror image. , , All are weighting coefficients, and ; Based on the D value, the images are pre-distributed sequentially from high to low to ensure that business nodes can quickly pull the images to start the algorithm.

[0037] 2) Flexible deployment of the algorithm, including: The elastic scheduling system, which connects to the meteorological big data cloud platform, will handle situations where the volume of business requests R exceeds a preset maximum request threshold R. max When the container is automatically expanded, the expansion operation is triggered, and the expansion quantity N satisfies a linear correlation model: ; Where k is the expansion coefficient, which is determined based on the parallelism of the algorithm. , This represents the floor function; When the number of business requests R is less than the preset minimum request threshold R min When the container is automatically recycled, the computing resources are dynamically optimized.

[0038] The above technical solution builds a regional distribution node network based on the image library and calculates the distribution priority of the image according to the business load of each computing node, realizing the elastic deployment of the algorithm. It can dynamically adjust the image distribution and container capacity according to actual business needs and computing resources. This intelligent elastic deployment method improves the flexibility and adaptability of algorithm deployment and greatly improves the overall deployment efficiency.

[0039] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating algorithm images based on container services, characterized in that: Includes the following steps: S1. Classify the algorithms deployed on the meteorological big data cloud platform according to business dimensions, sort out the operation dependencies of each algorithm, and construct an algorithm dependency matrix; S2. Determine the image hierarchy architecture. Based on the functional boundaries and dependencies of each layer, construct layer weight coefficients to measure the resource share and update priority of each layer in image creation. S3. Extract the core computational logic of the algorithm, separate the fixed computational framework of the algorithm from the configurable parameters, construct a parameter configuration mapping model, and customize the encapsulation of the core layer of the algorithm. S4. Encapsulate the algorithm's triggering conditions in the business adaptation layer, construct a triggering priority model, and perform customized encapsulation of the business adaptation layer. S5. Introduce a multi-dimensional verification mechanism, construct a verification pass rate evaluation function to quantify the verification pass rate of the image, and push the verified images to the image library to establish a full lifecycle management ledger for image versions. S6. Construct a regional distribution node network for images based on the image library, calculate the distribution priority of images according to the business load of each computing node, and perform elastic deployment of the algorithm.

2. The algorithm image generation method based on container services according to claim 1, characterized in that: S1 categorizes the algorithms deployed on the meteorological big data cloud platform according to business dimensions, identifies the operational dependencies of each algorithm, and constructs an algorithm dependency matrix, including: The algorithms deployed on the meteorological big data cloud platform are classified according to business dimensions, including meteorological numerical forecasting algorithms, meteorological element assimilation algorithms, and meteorological disaster early warning algorithms, and business attribute tags are labeled for each algorithm; The runtime dependencies of each algorithm are identified, including basic environment dependencies, specialized tool dependencies, and data dependencies, and an algorithm dependency matrix M is constructed, with matrix elements m. ij satisfy: 。 3. The algorithm image generation method based on container services according to claim 1, characterized in that: In S2, the image hierarchy architecture is determined. Based on the functional boundaries and dependencies of each layer, a layer weight coefficient is constructed to measure the resource share and update priority of each layer in image creation, including: The image hierarchy architecture is defined, with the following layers from bottom to top: basic environment layer, professional tools layer, algorithm core layer, and business adaptation layer. Each layer has independent update and reuse capabilities, and the layers interact through standardized interfaces. Based on the functional boundaries and dependencies of each layer, a layer weight coefficient is constructed to measure the resource share and update priority of each layer in image creation: ； in, The weight coefficients of the k-th layer in the mirror image are... k=1,2,3,4, the weight coefficients of the basic environment layer The adaptation algorithm is highly dependent on the basic environment, and the weight coefficients of the business adaptation layer... To ensure lightweight business configuration, the weight coefficient of the professional tool layer. Weight coefficients of the algorithm's core layer The algorithm dynamically allocates resources based on the degree of dependence on specialized tools.

4. The algorithm image generation method based on container services according to claim 1, characterized in that: S3 extracts the core computational logic of the algorithm, separates the algorithm's fixed computational framework from its configurable parameters, and constructs a parameter configuration mapping model, including: The core computational logic of the algorithm is extracted, separating the algorithm's fixed computational framework from its configurable parameters, including grid resolution. Time step and assimilated data weights Build a parameter configuration mapping model : ； in, For grid resolution The corresponding computational resource consumption function, For time step The corresponding iterative efficiency function, Weighting the assimilated data The corresponding forecast accuracy function, , , All are weighting coefficients, and , To compensate for parameter mismatch and ensure the accuracy of parameter configuration; Parameter configuration mapping model It can match the grid resolution according to the computing resources and operational accuracy requirements of the meteorological big data cloud platform. Time step Assimilation data weights and weighting coefficients , , The optimal combination of parameters, including those for the target.

5. The algorithm image generation method based on container services according to claim 4, characterized in that: S3 provides a customized encapsulation of the algorithm's core layer, including: The input and output interfaces of the encapsulated algorithm are uniformly connected to the data storage system of the meteorological big data cloud platform. The input interface supports standardized parsing of multi-source heterogeneous data, including gridded data, station observation data and satellite remote sensing data. The output interface supports the synchronous generation of visualized metadata of forecast results and business alarm instructions, ensuring seamless connection between the container and the platform data link. An embedded status monitoring module is used to monitor the running status of real-time monitoring algorithms. The monitoring indicators include computing resource utilization, iteration convergence speed, and data processing latency. Threshold alarm rules are set, and when the monitoring indicators exceed the preset range, resource scheduling requests or fault tolerance mechanisms within the container are automatically triggered.

6. The algorithm image generation method based on container services according to claim 1, characterized in that: In S4, the triggering conditions of the algorithm are encapsulated in the business adaptation layer, and a triggering priority model is constructed, including: The triggering conditions of the algorithm are encapsulated in the business adaptation layer, including timed triggering conditions, data arrival triggering conditions, and manual instruction triggering conditions, and a triggering priority model Q(t,p,s) is constructed: ； Where t is the time urgency coefficient, The closer to the forecast time, the greater the urgency coefficient, where p is the business priority coefficient. Meteorological disaster early warning services rank higher than meteorological numerical forecasting services, where 's' is the data completeness coefficient. The more complete the data, the higher the data completeness coefficient. , , All are weighting coefficients, and ; The meteorological big data cloud platform can prioritize the scheduling algorithm running in the container based on the size of the Q value.

7. The algorithm image generation method based on container services according to claim 6, characterized in that: S4 provides customized encapsulation for the business adaptation layer, including: The business adaptation layer is connected to the business process orchestration interface of the meteorological big data cloud platform, and the permission authentication module of the meteorological big data cloud platform is integrated. The identity authentication and operation audit logic of the algorithm operation is embedded, allowing only authorized business accounts of the platform to trigger the start, pause and termination operations of the container. At the same time, the algorithm's operation log and data access log are recorded to achieve security control of the business. Adapt to the computing power scheduling strategy of the meteorological big data cloud platform. Based on the computing power requirement tags of various algorithms, configure the computing power application template of the container in the business adaptation layer. When the container starts, it automatically applies to the meteorological big data cloud platform for the corresponding type of computing power resources, so as to achieve precise matching between the computing power requirements of the algorithm and the computing power supply of the platform.

8. The method for generating algorithm images based on container services according to claim 1, characterized in that: S5 introduces a multi-dimensional verification mechanism, constructs a verification pass rate evaluation function to quantify the verification pass rate of images, and pushes verified images to the image repository, establishing a full lifecycle management ledger for image versions, including: A multi-dimensional verification mechanism is designed, including environmental compatibility verification, algorithm functionality verification, and platform adaptability verification. A verification pass rate evaluation function V(e1,e2,e3) is constructed to quantify the verification pass rate of the image. ； Where e1 is the percentage of anomalies in environment compatibility verification, e2 is the error value in algorithm functionality verification, e3 is the percentage of failures in platform adaptability verification, and E 1max E 2max E 3max These are the maximum allowed values ​​for each validation dimension; Images with a V value greater than the preset pass threshold are considered to be valid images. Validated images are then version-fixed, and composite tags containing algorithm type, business attributes, and pass threshold are added to the images. The images are then pushed to the image library of the meteorological big data cloud platform to establish a full lifecycle management ledger for image versions.

9. The method for generating algorithm images based on container services according to claim 1, characterized in that: In S6, a regional distribution node network for images is built based on the image library. The distribution priority of images is calculated according to the service load of each computing node, including: A regional distribution node network for images is built based on the image library, and the distribution priority D of the images is calculated according to the business load of each computing node. ； Where L represents the service load of the computing node, i.e., the current computing power utilization rate of the computing node. S represents the urgency of the business corresponding to the mirror image. , , All are weighting coefficients, and ; Based on the D value, the images are pre-distributed sequentially from high to low to ensure that business nodes can quickly pull the images to start the algorithm.

10. The algorithm image generation method based on container services according to claim 9, characterized in that: S6 enables flexible deployment of algorithms, including: The elastic scheduling system, which connects to the meteorological big data cloud platform, will handle situations where the volume of business requests R exceeds a preset maximum request threshold R. max When the container is automatically expanded, the expansion operation is triggered, and the expansion quantity N satisfies a linear correlation model: ； Where k is the expansion coefficient, which is determined based on the parallelism of the algorithm. , This represents the floor function; When the number of business requests R is less than the preset minimum request threshold R min When the container is automatically recycled, the computing resources are dynamically optimized.

Images