Job scheduling method and device based on cloud native

By adopting cloud-native job scheduling methods in large-scale operation scenarios, generating pipeline models and matching container pods, and using a decentralized distributed scheduling controller, the low resource utilization and poor flexibility in the existing technology are solved. The problem has achieved efficient resource scheduling and diverse job support.

CN119088517APending Publication Date: 2024-12-06BEIJING SOHU NEW MEDIA INFORMATION TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411143339.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-20
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The existing operation scheduling scheme based on cluster management systems lacks containerized application support in large-scale operation scenarios, which has poor flexibility, resulting in low resource utilization and tends to static resource allocation, making it difficult to meet the needs of diverse operation scheduling.

Method used

The cloud-native job scheduling method is adopted, and the container resources are dynamically scheduled and managed by generating pipeline models and matching container pods, using a decentralized distributed job scheduling controller, combined with preset resource management strategies.

Benefits of technology

It improves resource scheduling efficiency in large-scale operation scenarios, supports the needs of diverse operation scenarios, improves resource utilization, avoids resource limitations during operation execution, and enhances system stability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119088517A_ABST
    Figure CN119088517A_ABST
Patent Text Reader

Abstract

The invention discloses a job scheduling method and device based on cloud native, and the method comprises the steps: generating a pipeline model corresponding to a plurality of to-be-processed jobs based on a customized resource of a preset cloud native environment; determining a container pod matched with each assembly line model based on a dependency relationship among a plurality of tasks in each ordered task set; and starting the operation of each container by using a decentralized distributed design job scheduling controller, and controlling the operation condition of each container according to a preset resource management and control strategy. The method comprises the following steps: determining container pods corresponding to a plurality of to-be-processed jobs by utilizing self-defined resources of a preset cloud native environment and combining a dependency relationship among a plurality of tasks, starting the operation of containers in the plurality of container pods through a job scheduling controller, and controlling the operation condition of each container according to a preset resource management and control strategy. The resource scheduling efficiency in a large-scale operation scene can be effectively guaranteed.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of job scheduling, and in particular to a cloud-native based job scheduling method and device. Background Art

[0002] In solving the problem of large-scale job scheduling, the current mainstream solution is a solution based on a cluster management system. This solution mainly uses the cluster management system to implement resource allocation and job scheduling, and usually focuses more on the resource management and scheduling of the entire cluster. The scheduling system obtains resource information in the cluster through the cluster manager, including available nodes, CPU, memory and other resources. The user submits a job request to the scheduling system, specifying information such as the job type and resource requirements. The scheduling system decides which job to allocate resources to based on the resource requirements of the job and the cluster status. The scheduling system starts a corresponding ApplicationMaster (task scheduling monitoring) for the job. The ApplicationMaster is responsible for coordinating resource applications and task scheduling with the scheduling system. The ApplicationMaster obtains cluster resources through the scheduling system, starts tasks on the corresponding nodes, and executes the job. After the job is executed, the scheduling system releases the occupied resources and returns them to the resource pool, waiting for the next job.

[0003] However, since most solutions based on cluster management systems are designed to meet the needs of a specific scenario (such as big data computing scenarios), they lack the ability to support containerized applications and are difficult to meet more general job scheduling scenarios. The flexibility of this solution is poor. Secondly, for the unified management and planning of resources, this solution tends to use a static resource allocation method. Fixed resources lead to low overall resource utilization, and a large number of resource restrictions may occur during job execution. Summary of the invention

[0004] The present application provides a cloud-native-based job scheduling method and device, aiming to ensure resource scheduling efficiency in large-scale job scenarios.

[0005] In order to achieve the above objectives, this application provides the following technical solutions:

[0006] A cloud-native-based job scheduling method includes:

[0007] Based on the custom resources of the preset cloud native environment, a pipeline model corresponding to a plurality of pending jobs is generated; the pending jobs include an ordered task set, and there are dependencies between the plurality of tasks in the ordered task set; the pipeline model includes a directed acyclic graph composed of task models corresponding to each task in the ordered task set;

[0008] Based on the dependency relationship between multiple tasks in each of the ordered task sets, determine a container pod that matches each of the pipeline models; the container pod includes multiple containers, and each of the containers deploys multiple task models with a dependency relationship;

[0009] A job scheduling controller with a decentralized distributed design is used to start the operation of each of the containers, and the operation status of each of the containers is controlled according to a preset resource management and control strategy.

[0010] Optionally, after determining the container pod matching each of the pipeline models, the method further includes:

[0011] Using a preset container pod template, an operation specification of each container pod is generated; the operation specification is used to limit resources available for the container pod.

[0012] Optionally, after determining the container pod matching each of the pipeline models, the method further includes:

[0013] The admission controller of the preset cloud native environment is used to inject the configuration information required by the task model deployed in each of the containers into each of the containers.

[0014] Optionally, the preset resource control policy includes a resource restriction policy and a resource allocation policy, and the operation status of each container is controlled according to the preset resource control policy, including:

[0015] Determine the resource type and resource quantity corresponding to each of the containers according to the resource restriction policy, and schedule running resources matching the resource type and resource quantity for each of the containers;

[0016] According to the resource allocation strategy, the operating environment corresponding to each of the containers is determined, and each of the containers is allocated to the corresponding operating environment.

[0017] A cloud-native-based job scheduling device, comprising:

[0018] A model generation unit, configured to generate a pipeline model corresponding to a plurality of pending jobs based on a custom resource of a preset cloud native environment; the pending jobs include an ordered task set, wherein there is a dependency relationship between the plurality of tasks in the ordered task set; the pipeline model includes a directed acyclic graph composed of task models corresponding to each task in the ordered task set;

[0019] A container allocation unit, configured to determine a container pod matching each of the pipeline models based on the dependency relationship between multiple tasks in each of the ordered task sets; the container pod includes multiple containers, each of which deploys multiple task models with a dependency relationship;

[0020] The container operation unit is used to start the operation of each container using a job scheduling controller with a decentralized distributed design, and to control the operation status of each container according to a preset resource management and control strategy.

[0021] Optionally, the container dispensing unit is further used for:

[0022] Using a preset container pod template, an operation specification of each container pod is generated; the operation specification is used to limit resources available for the container pod.

[0023] Optionally, the container dispensing unit is further used for:

[0024] The admission controller of the preset cloud native environment is used to inject the configuration information required by the task model deployed in each of the containers into each of the containers.

[0025] Optionally, the preset resource management and control policy includes a resource restriction policy and a resource allocation policy, and the container operation unit is specifically used to:

[0026] Determine the resource type and resource quantity corresponding to each of the containers according to the resource restriction policy, and schedule running resources matching the resource type and resource quantity for each of the containers;

[0027] According to the resource allocation strategy, the operating environment corresponding to each of the containers is determined, and each of the containers is allocated to the corresponding operating environment.

[0028] A storage medium includes a stored program, wherein the program executes the cloud-native-based job scheduling method when executed by a processor.

[0029] An electronic device, comprising: a processor, a memory and a bus; the processor and the memory are connected via the bus;

[0030] The memory is used to store programs, and the processor is used to run programs, wherein the program executes the cloud-native-based job scheduling method when the processor runs the program.

[0031] The technical solution provided in the present application generates a plurality of pipeline models corresponding to pending jobs based on the custom resources of a preset cloud native environment. Based on the dependencies between multiple tasks in each ordered task set, a container pod matching each pipeline model is determined. A job scheduling controller with a decentralized distributed design is used to start the operation of each container, and the operating status of each container is controlled according to a preset resource management strategy. The present application utilizes the custom resources of a preset cloud native environment, combined with the dependencies between multiple tasks, to determine the container pods corresponding to multiple pending jobs, starts the operation of containers in multiple container pods through a job scheduling controller, and controls the operating status of each of the containers according to a preset resource management strategy, which can effectively ensure the resource scheduling efficiency in large-scale operation scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0033] Figure 1 A flowchart of a cloud-native-based job scheduling method provided in an embodiment of the present application;

[0034] Figure 2 A flowchart of another cloud-native-based job scheduling method provided in an embodiment of the present application;

[0035] Figure 3 A flowchart of another cloud-native-based job scheduling method provided in an embodiment of the present application;

[0036] Figure 4 A schematic diagram of the architecture of a cloud-native based job scheduling device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0038] In this application, relational terms such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of more restrictions, the elements defined by the sentence "comprise one..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0039] like Figure 1 As shown, a flow chart of a cloud-native-based job scheduling method provided in an embodiment of the present application includes the following steps.

[0040] S101: Generate a pipeline model corresponding to a plurality of jobs to be processed based on custom resources of a preset cloud native environment.

[0041] The jobs to be processed include an ordered task set, and there are dependencies between multiple tasks in the ordered task set. The pipeline model includes a directed acyclic graph composed of task models corresponding to each task in the ordered task set.

[0042] It should be noted that cloud native is a software development and deployment methodology that aims to fully utilize the elasticity, flexibility, and scalability of the cloud computing environment. Cloud native application and service design pursues higher scalability, reliability, and agility to adapt to the dynamic characteristics of the cloud environment. The cloud native environment is a new type of cloud computing environment built on technologies such as containers and microservices. It can use the elasticity, scalability, and efficiency of the cloud platform to support the rapid iteration and deployment of applications.

[0043] In some examples, the preset cloud native environment can be Kubernetes, which is an open source container orchestration and management platform for automating the deployment, expansion, and operation of containerized applications.

[0044] In a possible implementation, a pipeline model corresponding to multiple jobs to be processed can be generated based on the CRD object of Kubernetes. Generally speaking, CRD is a custom resource definition for Kubernetes. CRD can be simply understood as a part of the API (Application Programming Interface) of Kubernetes, allowing users to create and operate custom resources. CRD includes the definition of custom resources and controllers. The definition of custom resources is used to describe the structure, properties, and behavior of custom resources. The controller is a custom controller running in the Kubernetes cluster, which is responsible for handling the life cycle and behavior of custom resources.

[0045] In a possible implementation, the pipeline model corresponding to the job to be processed can be understood as Pipeline, and the task model can be understood as Task. Task is the smallest unit of cloud native, representing a task model, which includes multiple steps, and each step can be defined as a step in Task. Pipeline can be understood as a directed acyclic graph composed of multiple Tasks.

[0046] A so-called directed acyclic graph is a graph structure in which each edge has a direction and there are no paths in the graph that form a cycle.

[0047] In some examples, the dependency relationship between tasks includes, but is not limited to, data dependency and task dependency.

[0048] It should be noted that when generating Pipelines corresponding to multiple pending jobs based on Kubernetes' CRD objects, it is also necessary to define Pipeline Run as the corresponding pipeline instance to generate a pipeline record. In addition, when generating Tasks corresponding to multiple tasks based on Kubernetes' CRD objects, it is also necessary to define TaskRun as the corresponding task instance to generate task status records.

[0049] S102: Determine a container pod matching each pipeline model based on the dependency relationship between multiple tasks in each ordered task set.

[0050] The container pod includes multiple containers, and each container deploys multiple task models with dependencies.

[0051] It should be noted that the container pod can be the smallest resource management component Pod in Kubernetes, and Pod is also the resource object that minimizes the running of containerized applications. A Pod represents a process running in passion. In a possible implementation, a Pod can run a container (it is necessary to ensure that the containers in a Pod must run on the same node), or a Pod can run multiple containers at the same time (it is necessary to ensure that a Pod simultaneously encapsulates multiple containers that need to be tightly coupled and collaborate with each other, and multiple containers share resources).

[0052] It is understandable that, by utilizing the functional characteristics of the container pod, the container pod matching each pipeline model is determined based on the dependencies between multiple tasks in each ordered task set, wherein each container pod deploys multiple task models with dependencies. Generally speaking, allocating resources to the container pod to run is actually allocating resources to the multiple task models in the container pod to run.

[0053] It should be noted that after determining the container pod that matches each pipeline model, a series of resource configurations are required to ensure that the container pod can run normally. Figure 2 Steps shown and explanations of the steps.

[0054] S103: Use a decentralized distributed job scheduling controller to start the operation of each container, and control the operation status of each container according to a preset resource management strategy.

[0055] Among them, Kubernetes' native scheduler Pipeline Controller is based on a decentralized distributed design. Therefore, Kubernetes' native scheduler can be used as a job scheduling controller to ensure the scheduling efficiency of each pending job.

[0056] In some examples, the main responsibilities of the Pipeline Controller are to monitor resource changes of the Pipeline and Task, parse the directed acyclic graph, parse the preset resource management policy, and coordinate and control the execution status of pending jobs corresponding to multiple directed acyclic graphs.

[0057] Optionally, the preset resource control strategy includes resource restriction strategy and resource allocation strategy. For the implementation process of controlling the operation status of each container according to the resource restriction strategy and resource allocation strategy, please refer to Figure 3 Steps shown and explanations of the steps.

[0058] It should be noted that, combined with Figure 2 as well as Figure 3The method shown, the embodiment of the present application can have the following advantages: First, based on the cloud native environment, it supports containerized applications and can meet the scheduling needs of various job scenarios; Second, the dependencies between the tasks in the pending job are calculated in the job definition stage, and there is no need to calculate when the pending job is executed, thereby avoiding the impact of a large number of complex jobs on scheduling efficiency; Third, the job scheduling controller is based on a decentralized distributed design, can schedule shards, support horizontal expansion, and cope with large-scale job scheduling; Fourth, multi-level resource restriction rules and multi-resource pool design provide multi-tenant job resource isolation strategies to avoid the impact of abnormal jobs and improve job execution and cluster stability.

[0059] It is understandable that by adopting the native scheduling, asynchronous execution, failure retry and other mechanisms provided by Kubernetes, designing Pipeline, Task and containerized execution environment, key issues such as job dependency management, dynamic resource management, scheduling performance and system stability are well solved. Specifically, Pipeline and Task are used, where Pipeline consists of multiple Tasks, and Task is the smallest executable unit. This resource involves allowing the definition and management of dependencies between tasks, ensuring that multiple tasks are executed in the correct order, so as to effectively solve the problems of job dependency management. In addition, tasks are executed in containers. This containerized execution environment enables tasks to run on different infrastructures, achieving better elasticity and resource management, and can support dynamic adaptation to different resource requirements to effectively solve the problems of dynamic resource management. Finally, the native scheduler of Kubernetes is used to schedule tasks, and a decentralized Pipeline Controller is designed to coordinate scheduling to ensure scheduling efficiency. The isolation of jobs between different tenants and scenarios is implemented through multi-level resource restriction rules and multi-level resource pools for resource allocation to ensure the overall stability of the system, so as to effectively solve the problems of scheduling performance and system stability.

[0060] The process shown in the above S101-S103 utilizes the custom resources of the preset cloud native environment and combines the dependencies between multiple tasks to determine the container pods corresponding to multiple jobs to be processed. Through the job scheduling controller, the operation of containers in multiple container pods is started, and the operation status of each container is controlled according to the preset resource management and control strategy, which can effectively ensure the resource scheduling efficiency in large-scale operation scenarios.

[0061] like Figure 2 As shown, a flow chart of another cloud-native-based job scheduling method provided in an embodiment of the present application includes the following steps.

[0062] S201: Generate operation specifications for each container pod using a preset container pod template.

[0063] Run specifications are used to limit the resources available to container pods.

[0064] It should be noted that the preset container pod template may be a Pod Template of Kubernetes, and the PodTemplate is used to define the operation specification of the container pod (i.e., Pod), and the operation specification includes but is not limited to the structure, configuration, and initialization script of the Pod.

[0065] S202: Use the admission controller of the preset cloud native environment to inject the configuration information required by the task model deployed in each container into each container.

[0066] Among them, the admission controller for the preset cloud native environment can be PodPreset of Kubernetes. PodPreset can be used to provide configuration information such as runtime environment variables and volume mounting for the task model deployed in the container pod.

[0067] It can be understood that using Pod Preset to inject the configuration information required by the task model deployed in each container into each container can ensure that the task model deployed in each container can run normally.

[0068] The process shown in S201-S202 above can effectively implement resource configuration of the container pod through the container pod template and the admission controller.

[0069] like Figure 3 As shown, a flow chart of another cloud-native-based job scheduling method provided in an embodiment of the present application includes the following steps.

[0070] S301: Determine the resource type and resource quantity corresponding to each container according to the resource restriction policy, and schedule running resources that match the resource type and resource quantity for each container.

[0071] Among them, resource types include but are not limited to CPU, memory, cache and other resources. By parsing the resource restriction policy, the resource type and resource quantity corresponding to each container can be determined, and the running resources matching the resource type and resource quantity can be scheduled for each container to ensure that the task model in each container can run normally.

[0072] S302: Determine the operating environment corresponding to each container according to the resource allocation strategy, and allocate each container to the corresponding operating environment.

[0073] The operating environment can be understood as the hardware physical environment of the container (for example, any node in Kubernetes is a computing device deployed in any region, etc.). Allocating each container to the corresponding operating environment can provide multi-tenant job resource isolation, prevent abnormal jobs from affecting the operation of other normal jobs, and improve job execution and cluster stability.

[0074] The process shown in S301-S302 above can effectively control the operating status of each container (such as operating resources and operating environment, etc.) by using resource limitation strategy and resource allocation strategy.

[0075] like Figure 4 As shown, it is a schematic diagram of the architecture of a cloud-native based job scheduling device provided in an embodiment of the present application, including the units shown below.

[0076] The model generation unit 100 is used to generate a pipeline model corresponding to multiple jobs to be processed based on custom resources of a preset cloud native environment; the jobs to be processed include an ordered task set, and there are dependencies between multiple tasks in the ordered task set; the pipeline model includes a directed acyclic graph composed of task models corresponding to each task in the ordered task set.

[0077] The container allocation unit 200 is used to determine a container pod matching each pipeline model based on the dependency relationship between multiple tasks in each ordered task set; the container pod includes multiple containers, and each container deploys multiple task models with dependencies.

[0078] The container operation unit 300 is used to start the operation of each container using a job scheduling controller with a decentralized distributed design, and to control the operation status of each container according to a preset resource management and control strategy.

[0079] Optionally, the container allocation unit 200 is further used to: generate an operation specification of each container pod using a preset container pod template; the operation specification is used to limit resources available for the container pod.

[0080] Optionally, the container allocation unit 200 is further used to: use the admission controller of the preset cloud native environment to inject the configuration information required by the task model deployed in each container into each container.

[0081] Optionally, the preset resource management strategy includes a resource limitation strategy and a resource allocation strategy, and the container operation unit 300 is specifically used to: determine the resource type and resource quantity corresponding to each container according to the resource limitation strategy, and schedule operating resources that match the resource type and resource quantity for each container; determine the operating environment corresponding to each container according to the resource allocation strategy, and allocate each container to the corresponding operating environment.

[0082] Each unit shown above uses the custom resources of the preset cloud native environment and combines the dependencies between multiple tasks to determine the container pods corresponding to multiple jobs to be processed. Through the job scheduling controller, the operation of containers in multiple container pods is started, and the operation status of each container is controlled according to the preset resource management and control strategy, which can effectively ensure the resource scheduling efficiency in large-scale operation scenarios.

[0083] The present application also provides a computer-readable storage medium, which includes a stored program, wherein the program executes the cloud-native-based job scheduling method provided by the present application.

[0084] The present application also provides an electronic device, including: a processor, a memory, and a bus. The processor and the memory are connected via a bus, the memory is used to store programs, and the processor is used to run programs, wherein when the program is running, the above-mentioned cloud-native-based job scheduling method provided by the present application is executed, including the following steps:

[0085] Based on the custom resources of the preset cloud native environment, a pipeline model corresponding to a plurality of pending jobs is generated; the pending jobs include an ordered task set, and there are dependencies between the plurality of tasks in the ordered task set; the pipeline model includes a directed acyclic graph composed of task models corresponding to each task in the ordered task set;

[0086] Based on the dependency relationship between multiple tasks in each of the ordered task sets, determine a container pod that matches each of the pipeline models; the container pod includes multiple containers, and each of the containers deploys multiple task models with a dependency relationship;

[0087] A job scheduling controller with a decentralized distributed design is used to start the operation of each of the containers, and the operation status of each of the containers is controlled according to a preset resource management and control strategy.

[0088] In a specific implementation process, after determining the container pod matching each of the pipeline models, the method further includes:

[0089] Using a preset container pod template, an operation specification of each container pod is generated; the operation specification is used to limit resources available for the container pod.

[0090] In a specific implementation process, after determining the container pod matching each of the pipeline models, the method further includes:

[0091] The admission controller of the preset cloud native environment is used to inject the configuration information required by the task model deployed in each of the containers into each of the containers.

[0092] In the specific implementation process, the preset resource control strategy includes a resource restriction strategy and a resource allocation strategy. According to the preset resource control strategy, the operation status of each container is controlled, including:

[0093] Determine the resource type and resource quantity corresponding to each of the containers according to the resource restriction policy, and schedule running resources matching the resource type and resource quantity for each of the containers;

[0094] According to the resource allocation strategy, the operating environment corresponding to each of the containers is determined, and each of the containers is allocated to the corresponding operating environment.

[0095] In addition, the functions described above in the embodiments of the present application may be at least partially performed by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0096] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.

[0097] Although several specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present application. Certain features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination.

[0098] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other to form a technical solution.

Claims

1. A cloud-native job scheduling method, characterized in that: include: Generate a pipeline model corresponding to a plurality of jobs to be processed based on custom resources of a preset cloud native environment; the jobs to be processed include an ordered task set, and there are dependencies between the plurality of tasks in the ordered task set; The pipeline model includes a directed acyclic graph composed of task models corresponding to each task in the ordered task set; Based on the dependency relationship between multiple tasks in each of the ordered task sets, determine a container pod that matches each of the pipeline models; the container pod includes multiple containers, and each of the containers deploys multiple task models with a dependency relationship; A job scheduling controller with a decentralized distributed design is used to start the operation of each of the containers, and the operation status of each of the containers is controlled according to a preset resource management and control strategy.

2. The method according to claim 1, characterized in that After determining the container pods matching each of the pipeline models, the method further includes: Using a preset container pod template, an operation specification of each container pod is generated; the operation specification is used to limit resources available for the container pod.

3. The method according to claim 1, characterized in that After determining the container pods matching each of the pipeline models, the method further includes: The admission controller of the preset cloud native environment is used to inject the configuration information required by the task model deployed in each of the containers into each of the containers.

4. The method according to claim 1, characterized in that: The preset resource control strategy includes a resource restriction strategy and a resource allocation strategy. According to the preset resource control strategy, the operation status of each container is controlled, including: Determine the resource type and resource quantity corresponding to each of the containers according to the resource restriction policy, and schedule running resources matching the resource type and resource quantity for each of the containers; According to the resource allocation strategy, the operating environment corresponding to each of the containers is determined, and each of the containers is allocated to the corresponding operating environment.

5. A cloud-native job scheduling device, characterized in that: include: A model generation unit, configured to generate a pipeline model corresponding to a plurality of pending jobs based on a custom resource of a preset cloud native environment; the pending jobs include an ordered task set, and there are dependencies between the plurality of tasks in the ordered task set; The pipeline model includes a directed acyclic graph composed of task models corresponding to each task in the ordered task set; A container allocation unit, configured to determine a container pod matching each of the pipeline models based on the dependency relationship between multiple tasks in each of the ordered task sets; the container pod includes multiple containers, each of which deploys multiple task models with a dependency relationship; The container operation unit is used to start the operation of each container using a job scheduling controller with a decentralized distributed design, and to control the operation status of each container according to a preset resource management and control strategy.

6. The device according to claim 5, characterized in that The container dispensing unit is also used for: Using a preset container pod template, an operation specification of each container pod is generated; the operation specification is used to limit resources available for the container pod.

7. The device according to claim 5, characterized in that The container dispensing unit is also used for: The admission controller of the preset cloud native environment is used to inject the configuration information required by the task model deployed in each of the containers into each of the containers.

8. The device according to claim 5, characterized in that The preset resource management and control strategy includes a resource restriction strategy and a resource allocation strategy, and the container operation unit is specifically used to: Determine the resource type and resource quantity corresponding to each of the containers according to the resource restriction policy, and schedule running resources matching the resource type and resource quantity for each of the containers; According to the resource allocation strategy, the operating environment corresponding to each of the containers is determined, and each of the containers is allocated to the corresponding operating environment.

9. A storage medium, characterized in that: The storage medium includes a stored program, wherein the program, when executed by a processor, executes the cloud-native-based job scheduling method described in any one of claims 1 to 4.

10. An electronic device, characterized in that: include: processor, memory, and bus; The processor is connected to the memory via the bus; The memory is used to store programs, and the processor is used to run programs, wherein the program, when run by the processor, executes the cloud-native-based job scheduling method described in any one of claims 1 to 4.