Container-based scheduling method and device for city-level computing power task

Abstract

The invention discloses a container-based scheduling method and device for city-level computing power tasks. According to the method, a computing power demand side obtains a computing power task runtime environment from a computing power task management module of a computing power service side and sends computing power task information and computing power task data to the computing power task management module; the computing power service party automatically constructs a computing power task mirror image according to the computing power task information and the computing power task data, then generates computing power task requirements, and issues the computing power task requirements to each computing power task matching module; the computing power supplier draws computing power task data from the computing power task management module in a preemptive manner through the computing power task matching module, executes a computing power task in the computing power infrastructure, and finally uploads a computing power task result to the computing power task management module; and the computing power demand side obtains a computing power task result from the computing power task management module. According to the method, a preemptive scheduling strategy oriented to the city-level computing power task is realized, and the overall resource utilization rate of the city-level computing power center is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer application technology, and in particular to a container-based scheduling method and device for city-level computing tasks. Background Art

[0002] With the accelerated advancement of digitalization, the demand for computing power across various sectors has exploded. Efficient scheduling and management of city-level computing tasks has become a critical issue that needs to be addressed. In the current technological landscape, computing tasks involve parties involved in computing power demanders, computing power service providers, and computing power suppliers. However, the collaborative process between these parties is plagued by complexities and inefficiencies. Computing power demanders often rely on complex and unstandardized interfaces and modules to obtain the computing power task runtime environment and transmit task information and data to computing power service providers. This not only increases communication costs but also easily leads to data transmission errors or delays due to inconsistent interfaces. After computing power service providers construct computing power task images based on computing power task information and data and generate requirements, they then publish the requirements to various computing power task matching modules. However, due to the lack of an efficient coordination mechanism, it is difficult to ensure that the task requirements reach the appropriate computing power suppliers accurately and promptly. When computing power providers preemptively pull computing power task data from computing power service providers and execute tasks within the infrastructure, they often lack effective resource allocation and scheduling strategies, resulting in low resource utilization, idle computing power resources, and the inability to complete some tasks on time due to insufficient resources. Furthermore, in the feedback and acquisition of computing power task results, the interaction between the parties is not smooth enough, which can easily lead to delayed results and affect the timeliness of the entire business process. Existing computing power scheduling technologies cannot effectively meet the complex needs of large-scale and diverse computing power tasks at the municipal level, making it difficult to implement efficient preemptive scheduling strategies to improve the overall resource utilization of municipal computing power centers.

[0003] Therefore, how to design an innovative container-based scheduling method and device for city-level computing power tasks, optimize the collaboration process between computing power task participants, and improve the allocation efficiency and overall utilization of computing power resources has become an important research direction and urgent need in the current field of computing power technology. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a container-based scheduling method and device for city-level computing power tasks, and constructs a complete scheduling method and device system. The computing power demander uses this module to obtain the computing power task runtime environment from the computing power task management module of the computing power service provider, and smoothly and accurately sends the computing power task information and data to the computing power task management module, thereby ensuring the accurate transmission of the demand information and the adaptive acquisition of the environment. The computing power task management module of the computing power service provider plays a key role. Based on the task information and data received from the computing power demander, it automatically and efficiently constructs a computing power task image, and then generates accurate computing power task requirements. These requirements are published to various computing power task matching modules to ensure the effective dissemination of task requirements. The computing power supplier uses the computing power task matching module to quickly pull computing power task data from the computing power service provider's computing power task management module in a preemptive manner, and then effectively executes the computing power task in the computing power infrastructure. After the task is completed, the results are promptly uploaded to the computing power task management module, realizing an efficient process of task execution and result feedback.

[0005] The present invention is achieved through the following technologies:

[0006] The first aspect of the present invention is a container-based scheduling method for city-level computing tasks, which is applied to a computing network for city-level computing tasks. The computing network is composed of various participants, including computing power demanders, computing power service providers, and computing power suppliers. The method specifically includes the following steps:

[0007] (1) The computing power demander pulls the computing power task runtime environment through the computing power task release module and configures the task dependency library;

[0008] (2) The computing power demander creates computing power task information and computing power task data through the computing power task publishing module and submits them to the computing power task management module;

[0009] (3) The computing power service provider uses the computing power task management module to build a container image tar package based on the computing power task runtime environment and submits it to the persistent storage module to create computing power task requirements and push them to each computing power task matching module;

[0010] (4) The computing power supplier, based on the computing power task requirements, pulls computing power task information from the computing power task management module through the computing power task matching module and then preempts computing power tasks concurrently;

[0011] (5) The computing power supplier loads computing power task data and executes computing power tasks through the computing power task matching module;

[0012] (6) The computing power supplier uploads the computing power task execution log and computing power task results, and the computing power demander obtains the computing power task execution log and computing power task results.

[0013] Furthermore, the step (1) specifically includes the following sub-steps:

[0014] (1.1) The computing power demander pulls the computing power task runtime environment through the computing power task publishing module. This environment is provided in the form of a container image. The components of the container image include the operating system, Python, Cuda, Python dependency libraries, and AI framework.

[0015] This component can be combined into different runtime environment container images;

[0016] (1.2) The computing power demander configures the task dependency library based on the runtime environment of the pulled computing power task, including the computing power task environment variables and the computing power task code dependency library.

[0017] Furthermore, the computing task name, computing task description, computing specifications and computing task script in the computing task information of step (2) are filled in by the computing task demander; the model address, data set address and result address in the computing task information are filled in by the computing service provider, which specifically includes the following sub-steps:

[0018] (2.1) The computing power demander creates computing power task information through the computing power task release module and submits it to the computing power task management module. This information includes the computing power task name, computing power task description, computing power specifications, computing power task script, model address, dataset address, and result address;

[0019] (2.2) The computing power demander uploads the computing power task runtime environment to the computing power task management module through the computing power task release module;

[0020] (2.3) The computing power demander uploads the computing power task data to the computing power task management module through the computing power task release module. The computing power task data includes the computing power task code and the computing power task data set.

[0021] Furthermore, the step (3) includes the following sub-steps:

[0022] (3.1) The computing power service provider uses the computing power task management module to build a computing power task image tarball based on the computing power task data uploaded by the computing power demander. The image tarball includes file system layer data, image metadata, and other file data.

[0023] (3.2) The computing power service provider generates computing power task requirements based on the computing power task information created by the computing power demander through the computing power task management module. The requirements include the computing power task ID and computing power specifications.

[0024] (3.3) The computing power service provider pushes computing power task requirements to each computing power task matching module through the computing power task management module.

[0025] Furthermore, the step (4) includes the following sub-steps:

[0026] (4.1) The computing power supplier receives computing power task requirements through the computing power task matching module, pulls the computing power task information list from the computing power task management module, and concurrently preempts computing power tasks, that is, sends a computing power task preemption request to the computing power task management module; the computing power task preemption request includes the computing power supplier ID and the computing power task ID;

[0027] (4.2) The computing power service provider receives the computing power task preemption request from the computing power task matching module through the computing power task management module, and ensures that each computing power task can only be preempted by one computing power task matching module; the computing power task management module generates a computing power task instance for the request based on the preemption status. The instance information includes the computing power task start time, computing power task end time, computing power task execution log and computing power task result, and is associated with the computing power task instance;

[0028] (4.3) The computing power supplier pulls computing power task data from the computing power task management module through the computing power task matching module.

[0029] Furthermore, the step (5) includes the following sub-steps:

[0030] (5.1) The computing power provider loads the computing power task data through the computing power task matching module to generate a local container image and load it into the computing power infrastructure;

[0031] (5.2) The computing power provider starts the container image through the computing power task matching module until the computing power task is completed.

[0032] Furthermore, the step (6) includes the following sub-steps:

[0033] (6.1) The computing power supplier uses the computing power task matching module to package the computing power task results according to the result directory in the computing power task information and upload them to the computing power task management module;

[0034] (6.2) The computing power service provider submits the uploaded computing power task execution log and execution results to the persistent storage module through the computing power task management module, and then associates the computing power task execution log and execution results with the computing power task instance;

[0035] (6.3) The computing power demander uses the computing power task release module to pull the computing power task execution log and execution results from the computing power task management module based on the computing power task instance.

[0036] A fourth aspect of the present invention is a container-based scheduling device for city-level computing tasks, located in a server of a computing network participant, and specifically comprising the following modules:

[0037] Computing power task publishing module: This module is deployed in the server of the computing power demander. The computing power demander first pulls the computing power task runtime environment from the computing power task management module through the computing power task publishing module, configures the task dependency library, and then creates computing power task information and computing power task data in the computing power task management module. Finally, the computing power task release module obtains the computing power task results from the computing power task management module.

[0038] Computing task management module: This module is deployed on the computing service provider's server. The computing service provider first obtains computing task information and computing task data from the persistent storage module through the computing task management module, builds a computing task image tarball, publishes computing task requirements, and then pushes the computing task requirements to each computing task matching module.

[0039] Computing Task Matching Module: This module is deployed on the server of the computing power supplier. The computing power supplier first pulls computing task information from the computing task management module through the computing task matching module and concurrently preempts computing tasks. It then pulls computing task data from the computing power infrastructure, executes computing tasks, and finally uploads the computing task results to the computing task management module.

[0040] Persistent storage unit: This module uses physical devices to persistently store various data of the task management module, including the computing task runtime environment, computing task information, computing task data, computing task container image tarball, and computing task result tarball.

[0041] The third aspect of the present invention: an electronic device, comprising: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the container-based scheduling method for city-level computing power tasks.

[0042] The fourth aspect of the present invention: A computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of any one of the container-based scheduling methods for city-level computing tasks.

[0043] The beneficial effects of the present invention are as follows:

[0044] By integrating regional computing resources, this invention effectively lowers the threshold for acquiring computing resources, allowing small and medium-sized enterprises to fully benefit from intelligent computing. Through the computing task publishing module, scientific research institutions, enterprises, and other entities can easily obtain powerful computing resources, significantly saving time and cost investment, and accelerating the process of technological innovation. Through the computing task matching module, resource allocation is optimized, equipment utilization is improved, and the risk of idle single-point resources and cross-vendor collaboration costs are reduced. Through the computing task management module, value-added services such as computing power leasing, algorithmic trading, and data labeling are derived, building a multi-level profit system and providing innovative momentum for the sustainable development of the science and technology field. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0046] Figure 1 This is the overall architecture diagram of the computing power network participants of the present invention;

[0047] Figure 2 This is a flow chart of the container-based scheduling method for city-level computing tasks of the present invention;

[0048] Figure 3 This is a flow chart of the method for pulling the computing power task runtime environment and configuring the task dependency library of the present invention;

[0049] Figure 4 A flow chart of the method for creating computing task information and computing task data and submitting them to the computing task management module of the present invention;

[0050] Figure 5 A flowchart of the method for creating computing task requirements and pushing them to various computing task matching modules in the present invention;

[0051] Figure 6 This is a flow chart of the method for concurrently preempting computing tasks after pulling computing task information in the present invention;

[0052] Figure 7 A flow chart of the method for loading computing task data and executing computing tasks according to the present invention;

[0053] Figure 8 A flowchart of the method for obtaining computing task execution logs and computing task results of the present invention;

[0054] Figure 9 This is the overall architecture diagram of the container-based scheduling device for city-level computing tasks in the present invention. DETAILED DESCRIPTION

[0055] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0056] like Figures 1-8 As shown, the present invention also discloses a container-based scheduling method for city-level computing tasks, and the implementation of each step is as follows:

[0057] Step 1: The computing power demander pulls the computing power task runtime environment through the computing power task release module and configures the task dependency libraries. In some instances, the computing power task release runtime environment is implemented as a Docker container image, with an Ubuntu operating system and Python as the runtime language. The Cuda library is configured to use NVIDIA GPU cards, and programming frameworks such as Caffe and Torch, as well as AI frameworks such as PyTorch and Tensorflow, are installed. Table 1 shows the environment variables used during the computing power task runtime and the components related to configuring the task code dependency libraries:

[0058] Table 1

[0059]

[0060]

[0061] Step 2: The computing power demander creates computing power task information and computing power task data through the computing power task release module and submits them to the computing power task management module. The computing power demander uploads the computing power task runtime environment to the computing power task management module through the computing power task release module. The computing power demander uploads the computing power task data to the computing power task management block through the computing power task release module. The computing power task data includes the computing power task code and the computing power task data set. Specifically, in the example, the computing power demander fills in the computing power task name, computing power task description, computing power specifications and computing power task script through the Web interface, and uploads the task model and task data set on the interface, where the computing power specifications are optional, such as NVIDIA GPU: 1; video memory: 40GB; CPU: 8; memory: 50GB. It is then sent to the computing power task management module through the HTTP interface, which uses the OSS storage device to store the data submitted by the computing power demander, and supplements the model address and data set address with the corresponding OSS storage address respectively. The result address is the directory in the computing power task runtime environment, such as / home / task / result. Table 2 shows the computing power task information and related data of the computing power task:

[0062] Table 2

[0063]

[0064]

[0065] Step 3: The computing power service provider uses the computing power task management module to build a container image tarball based on the computing power task runtime environment, submits it to the persistent storage module, creates a computing power task requirement, and pushes it to each computing power task matching module. In some instances, the computing power task management module builds a container image tarball based on Docker technology, implements the packaging process through the dockersave command, and then stores the container image tarball in the OSS object storage system to generate a computing power task requirement. Finally, the computing power task requirement is pushed to each computing power task matching module using the HTTP interface. Table 3 shows the relevant components of the container image tarball, and Table 4 shows the relevant data of the computing power task requirement:

[0066] Table 3

[0067]

[0068] Table 4

[0069]

[0070]

[0071] Step 4: The computing power supplier pulls computing power task information from the computing power task management module through the computing power task matching module according to the computing power task demand, and then preempts the computing power task concurrently; the computing power service provider receives the computing power task preemption application from the computing power task matching module through the computing power task management module, and ensures that each computing power task can only be preempted by one computing power task matching module; the computing power supplier pulls computing power task data from the computing power task management module through the computing power task matching module. Specifically, in some instances, the computing power task matching module uses the HTTP interface to pull the computing power task information list from the computing power task management module, and then uses the HTTP interface to preempt the computing power task information from the computing power task management module. Table 5 shows the relevant data of preempted computing power tasks, and Table 6 shows the relevant data of computing power task instances:

[0072] Table 5

[0073]

[0074] Table 6

[0075]

[0076] Step 5: The computing power provider loads the computing power task data and executes the computing power task through the computing power task matching module. Specifically, the computing power provider uses the computing power task matching module to load the computing power task data, generate a local container image, and load it into the computing power infrastructure. The computing power provider then uses the computing power task matching module to start the container image until the computing power task is completed. Specifically, in some instances, the computing power matching module is based on Docker technology, using the docker load command to load the computing power task container image tarball, and then using the docker run command to run the computing power task container image.

[0077] Step 6: The computing power supplier uploads the computing power task execution log and results, and the computing power demander obtains the computing power task execution log and results. Specifically, the computing power supplier uses the computing power task matching module to block the computing power task results based on the result directory in the computing power task information and uploads them to the computing power task management module. The computing power service provider submits the uploaded computing power task execution log and results to the persistent storage module through the computing power task management module and then associates the computing power task execution log and results with the computing power task instance. The computing power demander uses the computing power task publishing module to pull the computing power task execution log and results from the computing power task management module based on the computing power task instance. In some instances, the computing power matching module is based on Docker technology, using the docker commit command to save the running computing power task container image, then using the docker save command to package the computing power task into a container image tarball, and finally uploading it to the computing power task management module using the HTTP interface. The computing power task publishing module uses the HTTP interface to download the computing power task management module from the computing power task management module and obtain the task execution results in the result directory.

[0078] like Figure 9 As shown, a container-based scheduling device for city-level computing tasks is disclosed. Examples of implementation methods of each module are as follows:

[0079] Module 1: Computing Task Release Module. In some instances, the computing task release module serves as the core infrastructure of the computing task scheduling system, providing functions such as pulling, pushing, and configuring container images. For example, Harbor is an open-source enterprise-level registry server for storing and distributing container images. The Harbor client is a tool for users to interact with the Harbor service. It provides a series of commands to perform operations such as image pushing, pulling, and repository management, and can serve as a specific implementation of the task release module.

[0080] Module 2: Computing Task Management Module. In some cases, the computing task management module serves as the core infrastructure of the computing task scheduling system, providing functions such as task management, demand management, data management, and matchmaking and preemption. Different technology stacks are used to build business logic and expose business interfaces for different business services. For example, in a Java-based technology architecture, the computing task management module is typically built using the SpringMVC framework, connected to the persistent storage unit through the Mybatis framework, connected to the Harbor storage repository through the DockerJava framework, and implemented through the HttpClient component for request forwarding.

[0081] Module 3: Computing Task Matching Module. In some instances, the computing task matching module serves as the core infrastructure of the computing task scheduling system, providing functions such as container creation and operation, container stopping and starting, container viewing and monitoring, and data volume creation and deletion. For example, the Docker client pulls images from a remote image repository (such as Docker Hub or an enterprise's internal Harbor repository) to the local machine, or pushes locally built images to a remote image repository, using the docker run command to create and run a container.

[0082] Module 4: Persistent Storage Module. In some cases, the persistent storage module serves as core infrastructure in the computing task scheduling system, providing data persistence, data reading and writing, file systems or key-value storage systems, and data security and integrity. Currently, commonly used persistent storage modules in computing task scheduling systems include ETCD, MySQL, TiKV, and MongoDB.

[0083] The present invention also discloses an electronic device comprising: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the container-based scheduling method for city-level computing tasks; and a computer-readable storage medium storing computer instructions that, when executed by the processor, implement any of the steps of the container-based scheduling method for city-level computing tasks.

[0084] Those skilled in the art will readily appreciate other embodiments of the present application after considering the description and practicing the contents disclosed herein. This application is intended to encompass any variations, uses, or adaptations of the present application, which follow the general principles of the present application and include common knowledge or customary techniques in the art that are not disclosed herein. The description and examples are intended to be exemplary only. It should be understood that the present application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A container-based scheduling method for city-level computing tasks, characterized by: Applied to a computing power network for city-level computing tasks, the computing power network consists of various participants, including computing power demanders, computing power service providers, and computing power suppliers. Specifically, the following steps are included: (1) The computing power demander pulls the computing power task runtime environment through the computing power task release module and configures the task dependency library; (2) The computing power demander creates computing power task information and computing power task data through the computing power task publishing module and submits them to the computing power task management module; (3) The computing power service provider uses the computing power task management module to build a container image tar package based on the computing power task runtime environment and submits it to the persistent storage module to create computing power task requirements and push them to each computing power task matching module; (4) The computing power supplier, based on the computing power task requirements, pulls computing power task information from the computing power task management module through the computing power task matching module and then preempts computing power tasks concurrently; (5) The computing power supplier loads computing power task data and executes computing power tasks through the computing power task matching module; (6) The computing power supplier uploads the computing power task execution log and computing power task results, and the computing power demander obtains the computing power task execution log and computing power task results.

2. The scheduling method according to claim 1, characterized in that: The step (1) specifically includes the following sub-steps: (1.1) The computing power demander pulls the computing power task runtime environment through the computing power task publishing module. This environment is provided in the form of a container image. The components of the container image include the operating system, Python, Cuda, Python dependency libraries, and AI frameworks. These components can be combined into different runtime environment container images; (1.2) The computing power demander configures the task dependency library based on the runtime environment of the pulled computing power task, including the computing power task environment variables and the computing power task code dependency library.

3. The scheduling method according to claim 1, characterized in that: The computing task name, computing task description, computing task specifications and computing task script in the computing task information of step (2) are filled in by the computing task demander; the model address, data set address and result address in the computing task information are filled in by the computing service provider, which specifically includes the following sub-steps: (2.1) The computing power demander creates computing power task information through the computing power task release module and submits it to the computing power task management module. This information includes the computing power task name, computing power task description, computing power specifications, computing power task script, model address, dataset address, and result address; (2.2) The computing power demander uploads the computing power task runtime environment to the computing power task management module through the computing power task release module; (2.3) The computing power demander uploads the computing power task data to the computing power task management module through the computing power task release module. The computing power task data includes the computing power task code and the computing power task data set.

4. The scheduling method according to claim 1, characterized in that: The step (3) includes the following sub-steps: (3.1) The computing power service provider uses the computing power task management module to build a computing power task image tarball based on the computing power task data uploaded by the computing power demander. The image tarball includes file system layer data, image metadata, and other file data. (3.2) The computing power service provider generates computing power task requirements based on the computing power task information created by the computing power demander through the computing power task management module. The requirements include the computing power task ID and computing power specifications. (3.3) The computing power service provider pushes computing power task requirements to each computing power task matching module through the computing power task management module.

5. The scheduling method according to claim 1, characterized in that: The step (4) includes the following sub-steps: (4.1) The computing power supplier receives computing power task requirements through the computing power task matching module, pulls the computing power task information list from the computing power task management module, and concurrently preempts computing power tasks, that is, sends a computing power task preemption request to the computing power task management module; The computing power task preemption request includes the computing power supplier ID and the computing power task ID; (4.2) The computing power service provider receives the computing power task preemption request from the computing power task matching module through the computing power task management module, and ensures that each computing power task can only be preempted by one computing power task matching module; the computing power task management module generates a computing power task instance for the request based on the preemption status. The instance information includes the computing power task start time, computing power task end time, computing power task execution log and computing power task result, and is associated with the computing power task instance; (4.3) The computing power supplier pulls computing power task data from the computing power task management module through the computing power task matching module.

6. The scheduling method according to claim 1, characterized in that: The step (5) includes the following sub-steps: (5.1) The computing power provider loads the computing power task data through the computing power task matching module to generate a local container image and load it into the computing power infrastructure; (5.2) The computing power provider starts the container image through the computing power task matching module until the computing power task is completed.

7. The scheduling method according to claim 1, characterized in that: The step (6) includes the following sub-steps: (6.1) The computing power supplier uses the computing power task matching module to package the computing power task results according to the result directory in the computing power task information and upload them to the computing power task management module; (6.2) The computing power service provider submits the uploaded computing power task execution log and execution results to the persistent storage module through the computing power task management module, and then associates the computing power task execution log and execution results with the computing power task instance; (6.3) The computing power demander uses the computing power task release module to pull the computing power task execution log and execution results from the computing power task management module based on the computing power task instance.

8. A container-based scheduling device for city-level computing tasks, characterized in that: The device is located in the server of the computing power network participant and specifically includes the following modules: Computing power task publishing module: This module is deployed in the server of the computing power demander. The computing power demander first pulls the computing power task runtime environment from the computing power task management module through the computing power task publishing module, configures the task dependency library, and then creates computing power task information and computing power task data in the computing power task management module. Finally, the computing power task release module obtains the computing power task results from the computing power task management module. Computing task management module: This module is deployed on the computing service provider's server. The computing service provider first obtains computing task information and computing task data from the persistent storage module through the computing task management module, builds a computing task image tarball, publishes computing task requirements, and then pushes the computing task requirements to each computing task matching module. Computing Task Matching Module: This module is deployed on the server of the computing power supplier. The computing power supplier first pulls computing task information from the computing task management module through the computing task matching module and concurrently preempts computing tasks. It then pulls computing task data from the computing power infrastructure, executes computing tasks, and finally uploads the computing task results to the computing task management module. Persistent storage unit: This module uses physical devices to persistently store various data of the task management module, including the computing task runtime environment, computing task information, computing task data, computing task container image tarball, and computing task result tarball.

9. An electronic device, characterized in that: include: one or more processors; a memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the container-based scheduling method for city-level computing power tasks as described in any one of claims 1-7.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed by the processor, the steps of the container-based scheduling method for city-level computing power tasks as described in any one of claims 1 to 7 are implemented.

Images