Scheduling system, method and device and readable storage medium

Abstract

The invention discloses a scheduling system, method and device and a readable storage medium, relates to the technical field of resource scheduling, and is used for matching supply and demand among computing power resources, improving the resource utilization rate and realizing global scheduling optimization of the resources. The scheduling system comprises a resource access module used for obtaining available resource data from at least one docking system; the scheduling management module is used for determining a scheduling strategy for executing the scheduling task according to resources required for executing the scheduling task and available resource data, and executing the scheduling task according to the scheduling strategy, and the scheduling management module comprises a scheduling decision module, a scheduling memory module and a scheduling tool module; the scheduling decision module is used for carrying out resource matching and allocation on the scheduling task based on the scheduling task information, determining a scheduling strategy for executing the scheduling task, and guiding acceptance and execution of the scheduling task according to the scheduling strategy; the scheduling memory module is used for providing historical scheduling information; the scheduling tool module is used for providing a scheduling tool.

Description

Technical Field

[0001] The present application relates to the technical field of resource scheduling, and in particular to a scheduling system, method, device and readable storage medium. Background Art

[0002] In power production and operation, computing resources mainly refer to the computing power used to process and analyze power system data. At present, the computing network can be used to expand the network control plane capabilities and achieve optimal allocation of computing resources.

[0003] However, with the development of computing power networks, available computing power resources are increasing. In this case, how to match the supply and demand of computing power resources and improve resource utilization is an urgent problem to be solved. Summary of the invention

[0004] The present application provides a scheduling system, method, device and readable storage medium for matching supply and demand between computing resources, improving resource utilization, and achieving global scheduling optimization of resources.

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

[0006] In a first aspect, a scheduling system is provided, which includes: a resource access module and a scheduling management module; wherein the resource access module is used to obtain available resource data from at least one docking system; the scheduling management module is used to determine a scheduling strategy for executing a scheduling task based on the resources required for executing the scheduling task and the available resource data, and execute the scheduling task according to the scheduling strategy, and the scheduling management module includes: a scheduling decision module, a scheduling memory module and a scheduling tool module; the scheduling decision module is used to match and allocate resources for the scheduling task based on the scheduling task information, determine the scheduling strategy for executing the scheduling task, and guide the acceptance and execution of the scheduling task according to the scheduling strategy; the scheduling memory module is used to provide historical scheduling information; and the scheduling tool module is used to provide a scheduling tool.

[0007] Based on the above scheduling system, the resource access module and the scheduling management module can be used to utilize the available resource data in at least one docking system to perform resource scheduling for scheduling tasks, thereby breaking the current computing power island phenomenon and solving the problem of resource scheduling and coordination, thereby achieving supply and demand matching between computing power resources, improving resource utilization, and realizing global scheduling optimization of resources.

[0008] In a possible implementation, the above-mentioned scheduling system further includes: a system management module; wherein the system management module is used to maintain the operation of the scheduling system.

[0009] In a possible implementation, the system management module includes a role management module; wherein the role management module is used to manage the resource access rights of tenants and users in the scheduling system, and the resource access rights are resource access rights across docking systems.

[0010] In one possible implementation, the resource access module includes at least one of the following: an access management module, a computing power management module, a network management module, a sample management module, a model management module and a task management module; wherein the access management module is used to support authentication and authorization for docking with at least one docking system, and supports multiple resource data access methods; the computing power management module is used to support the management and monitoring of computing resources of at least one docking system; the network management module is used to obtain network resource data of at least one docking system; the sample management module is used to support the acquisition of sample data information of at least one docking system; the model management module is used to support the acquisition of model information of at least one docking system; and the task management module is used to support the acquisition of computing task information of at least one docking system.

[0011] In a possible implementation, the above-mentioned scheduling management module further includes: a scheduling agent module; wherein the scheduling agent module is used to maintain the operation of the scheduling management module.

[0012] In one possible implementation, the above-mentioned scheduling agent module is used to maintain the operation of the scheduling management module, including: parsing the scheduling task information of the scheduling task through the scheduling decision module, matching and allocating resources of the scheduling task for the parsed scheduling task information based on the historical scheduling information provided by the scheduling memory module, determining the scheduling strategy for executing the scheduling task, and generating scheduling instructions based on the scheduling strategy, using the scheduling tools and available resource data provided by the scheduling tool to execute the scheduling task corresponding to the scheduling instruction.

[0013] In one possible implementation, the above-mentioned scheduling decision module includes a parsing module, a resource matching and allocation module, and a task acceptance and execution module; wherein the parsing module is used to perform at least one of the following processing on the scheduling task information: intent recognition, entity recognition, parameter parsing and constraint condition parsing; the resource matching and allocation module is used to perform availability check, scheduling strategy matching and scheduling algorithm matching on the scheduling task information, and determine the scheduling strategy for executing the scheduling task; the task acceptance and execution module is used for at least one of the following: scheduling service activation, scheduling work order acceptance and scheduling task acceptance.

[0014] In a second aspect, a scheduling method is provided, which is applied to a scheduling system, and the method includes: obtaining available resource data from at least one docking system; parsing scheduling task information of a scheduling task; based on historical scheduling information, matching and allocating resources of the scheduling task for the parsed scheduling task information, determining a scheduling strategy for executing the scheduling task, and generating a scheduling instruction based on the scheduling strategy; using a scheduling tool and available resource data to execute the scheduling task corresponding to the scheduling instruction.

[0015] Based on the above scheduling method, the resource access module and the scheduling management module can be used to utilize the available resource data in at least one docking system to perform resource scheduling for scheduling tasks, thereby breaking the current computing power island phenomenon and solving the problem of resource scheduling and coordination, thereby achieving supply and demand matching between computing power resources, improving resource utilization, and realizing global scheduling optimization of resources.

[0016] In a possible implementation, the above-mentioned resource matching and allocation of the scheduling task for the parsed scheduling task information based on the historical scheduling information, and determining the scheduling strategy for executing the scheduling task, include:

[0017] Based on the historical scheduling information, the parsed scheduling task information is used to match and allocate resources for the scheduling task to obtain at least one scheduling strategy;

[0018] Based on the user's selection of at least one scheduling strategy, a scheduling strategy for executing the scheduling task is determined.

[0019] In a possible implementation manner, the method further includes: managing resource access rights of tenants and users in the scheduling system, where the resource access rights are resource access rights across docking systems.

[0020] In a possible implementation, the obtaining of available resource data from at least one docking system includes:

[0021] Obtain at least one of the following from at least one docking system:

[0022] The computing resources of at least one docking system, the network resource data of at least one docking system, the sample data information of at least one docking system, the model information of at least one docking system, and the computing task information of at least one docking system.

[0023] In a possible implementation, the above-mentioned parsing of the scheduling task information includes:

[0024] The scheduling task information is processed by at least one of the following: intent recognition, entity recognition, parameter parsing, and constraint condition parsing.

[0025] In a possible implementation, the above-mentioned resource matching and allocation of the scheduling task for the parsed scheduling task information includes:

[0026] Perform availability check, scheduling strategy matching and scheduling algorithm matching on scheduling task information.

[0027] In a possible implementation, the scheduling task corresponding to the above-mentioned execution of the scheduling instruction includes at least one of the following: scheduling service activation, scheduling work order acceptance and scheduling task acceptance.

[0028] In a possible implementation, the scheduling task corresponding to the execution scheduling instruction includes:

[0029] According to the scheduling instructions, the scheduling tasks are sent to the docking system indicated by the scheduling strategy.

[0030] In the third aspect, a scheduling device is provided, which can be a scheduler or a chip or system on chip in the scheduler. The scheduling device can implement the functions performed by the scheduler in the above aspects or possible designs, and the functions can be implemented by hardware, such as: in one possible design, the scheduling device may include: a processor and a communication interface, and the processor can be used to support the scheduling device to implement the functions involved in the above second aspect or any possible design of the second aspect, for example: the scheduling device obtains available resource data from at least one docking system through the communication interface; and parses the scheduling task information of the scheduling task through the processor; based on the historical scheduling information, the parsed scheduling task information is used to match and allocate the resources of the scheduling task, determine the scheduling strategy for executing the scheduling task, and generate the scheduling instruction based on the scheduling strategy; use the scheduling tool and the available resource data to execute the scheduling task corresponding to the scheduling instruction.

[0031] In another possible design, the scheduling device may further include a memory, the memory being used to store computer-executable instructions and data necessary for the scheduling device. When the scheduling device is running, the processor executes the computer-executable instructions stored in the memory, so that the scheduling device executes the scheduling method described in the second aspect or any possible design of the second aspect.

[0032] In a fourth aspect, a computer-readable storage medium is provided, which may be a readable non-volatile storage medium, and which stores computer instructions or programs. When the computer-readable storage medium is run on a computer, the computer can execute the scheduling method described in the second aspect or any possible design of the above aspects.

[0033] In a fifth aspect, a computer program product comprising instructions is provided, which, when executed on a computer, enables the computer to execute the scheduling method described in the second aspect or any possible design of the above aspects.

[0034] In a sixth aspect, a chip system is provided, the chip system including a processor and a communication interface, the chip system can be used to implement the functions performed by the scheduling device in the above-mentioned second aspect or any possible design of the second aspect, for example, the processor is used to obtain available resource data from at least one docking system through the communication interface. In one possible design, the chip system also includes a memory, and the memory is used to store program instructions and / or data. The chip system can be composed of chips, and can also include chips and other discrete devices, without limitation.

[0035] Among them, the technical effects brought about by any design method from the second aspect to the sixth aspect can refer to the technical effects brought about by the above-mentioned first aspect, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 One of the architecture diagrams of a scheduling system provided in an embodiment of the present application;

[0037] Figure 2 A second architecture diagram of a scheduling system provided in an embodiment of the present application;

[0038] Figure 3 An architectural diagram of a scheduling management module of a scheduling system provided in an embodiment of the present application;

[0039] Figure 4 A third architectural diagram of a scheduling system provided in an embodiment of the present application;

[0040] Figure 5 A scheduling system resource docking logic diagram provided in an embodiment of the present application;

[0041] Figure 6 One of the flowcharts of a scheduling method provided in an embodiment of the present application;

[0042] Figure 7 A second flowchart of a scheduling method provided in an embodiment of the present application;

[0043] Figure 8 A third flowchart of a scheduling method provided in an embodiment of the present application;

[0044] Fig. 9 A fourth flowchart of a scheduling method provided in an embodiment of the present application;

[0045] Fig.10The fifth flowchart of a scheduling method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0047] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the attached claims.

[0048] It should also be understood that the term “comprising” indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements and / or components.

[0049] The term "and / or" in this article is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0050] It should be noted that, in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0051] In the description of the present application, unless otherwise specified, “plurality” means two or more.

[0052] The power sector is one of the key areas related to the national economy and people's livelihood. The power grid is large in scale, rich in data resources, and complex in operation and maintenance, which has an urgent need for artificial intelligence technology. In recent years, artificial intelligence technology has been integrated into all aspects of power production and operation, and has played an increasingly important role in power grid load forecasting, power dispatching, power transmission and transformation facility monitoring, equipment fault diagnosis, operation risk warning, construction site safety management and control, and intelligent customer service. The development of artificial intelligence is inseparable from the driving force of computing resources.

[0053] In order to support the application of power artificial intelligence technology and promote the breakthrough development of power artificial intelligence technology, the power sector has carried out a series of artificial intelligence basic capacity building in recent years. Build an artificial intelligence platform to serve the support of artificial intelligence computing resource scheduling, model building, model deployment and operation, and provide unified model training and model reasoning services for various units and professions by building an operating environment and training environment. Build sample libraries and model libraries related to artificial intelligence applications to provide sample and model resources and services for various units and professions.

[0054] However, due to the lack of early unified planning for the rapidly developing power artificial intelligence applications, the existing computing resources are scattered and unevenly laid out, resulting in problems such as low resource utilization, lack of coordination capabilities, and mismatch between supply and demand.

[0055] In view of this, an embodiment of the present application provides a scheduling system for matching supply and demand between computing resources, improving resource utilization, and realizing global scheduling optimization of resources. The scheduling system includes: a resource access module and a scheduling management module; wherein the resource access module is used to obtain available resource data from at least one docking system; the scheduling management module is used to determine the scheduling strategy for executing the scheduling task according to the resources required for the scheduling task execution and the available resource data, and execute the scheduling task according to the scheduling strategy, and the scheduling management module includes: a scheduling decision module, a scheduling memory module and a scheduling tool module; the scheduling decision module is used to match and allocate resources for the scheduling task based on the scheduling task information, determine the scheduling strategy for executing the scheduling task, and guide the acceptance and execution of the scheduling task according to the scheduling strategy; the scheduling memory module is used to provide historical scheduling information; the scheduling tool module is used to provide a scheduling tool.

[0056] Based on the above scheme, the scheduling system provided in the embodiment of the present application can use the available resource data in at least one docking system to perform resource scheduling for scheduling tasks through the resource access module and the scheduling management module, break the current computing power island phenomenon, and solve the problem of resource scheduling and coordination, thereby achieving supply and demand matching between computing power resources, improving resource utilization, and realizing global scheduling optimization of resources.

[0057] The scheduling system provided in the embodiment of the present application can be applied to the scenario of computing power scheduling.

[0058] like Figure 1 As shown, it is a schematic diagram of the architecture of the scheduling system provided in an embodiment of the present application. The scheduling system 10 may include a resource access module 11 and a scheduling management module 12.

[0059] In a possible implementation, the scheduling system 10 may also be referred to as a scheduler, which have the same meaning and can be used interchangeably in the text.

[0060] In the embodiment of the present application, the resource access module 11 may be used to obtain available resource data from at least one docking system.

[0061] In one possible implementation, the resource access module 11 can obtain resource data from docking systems such as a local artificial intelligence training platform, a local artificial intelligence reasoning platform, a model library, a sample library, and a network management platform to facilitate resource management and scheduling.

[0062] In one possible implementation, combining Figure 1 ,like Figure 2 As shown, the resource access module 11 may include at least one of the following: an access management module 111 , a computing power management module 112 , a network management module 113 , a sample management module 114 , a model management module 115 and a task management module 116 .

[0063] In a possible implementation, the access management module 111 may be used to support authentication and authorization for docking with at least one docking system, and support multiple resource data access methods.

[0064] In one possible implementation, the computing power management module 112 may be used to support the management and monitoring of computing resources of at least one docking system.

[0065] Exemplarily, the computing power management module 112 may be used to support the management and monitoring of computing resources such as general computing power, intelligent computing power, and super computing power of at least one docking system.

[0066] In a possible implementation, the network management module 113 may be used to obtain network resource data of at least one docking system.

[0067] Exemplarily, the above-mentioned network management module 113 can be used to support obtaining network resource data from a network management platform.

[0068] In a possible implementation, the sample management module 114 may be used to support acquisition of sample data information of at least one docking system.

[0069] In a possible implementation, the model management module 115 may be used to support obtaining model information of at least one docking system.

[0070] In a possible implementation, the task management module 116 may be used to support obtaining computing task information of at least one docking system.

[0071] In the embodiment of the present application, the above-mentioned scheduling management module 12 can be used to determine the scheduling strategy for executing the scheduling task according to the resources required for executing the scheduling task and the available resource data, and execute the scheduling task according to the scheduling strategy.

[0072] In one possible implementation, the scheduling management module 12 can be used to support the selection of a scheduling task running environment based on the resources required for the scheduling task execution and the available resources of each cluster, ensuring that each scheduling task can run on a node with sufficient resources, or scheduling samples and models on demand.

[0073] In the present application embodiment, Figure 3 As shown, the above-mentioned scheduling management module 12 may include: a scheduling decision module 121 , a scheduling memory module 122 and a scheduling tool module 123 .

[0074] In a possible implementation, the scheduling management module 12 may be an agent responsible for generating scheduling instructions and executing actions.

[0075] In an embodiment of the present application, the above-mentioned scheduling decision module 121 can be used to match and allocate resources for scheduling tasks based on scheduling task information, determine the scheduling strategy for executing the scheduling tasks, and guide the acceptance and execution of scheduling tasks according to the scheduling strategy.

[0076] In a possible implementation, the scheduling decision module 121 may guide detailed scheduling operations through the thinking and planning capabilities of a large model.

[0077] In one possible implementation, the scheduling decision module 121 can parse the scheduling task information submitted by the user through the semantic understanding ability of the big model, match and allocate resources for the scheduling task through thinking and planning capabilities, and guide the acceptance and execution of detailed scheduling tasks.

[0078] In a possible implementation, the scheduling decision module 121 may include a parsing module, a resource matching and allocation module, and a task acceptance and execution module;

[0079] In a possible implementation, the above-mentioned parsing module is used to perform at least one of the following processing on the scheduling task information: intent recognition, entity recognition, parameter parsing and constraint condition parsing.

[0080] In one possible implementation, the above-mentioned intent recognition can analyze the keywords, phrases and sentence structures in the scheduling task information submitted by the user through a pre-trained large language model to determine the specific operation that the user wants to perform.

[0081] In one possible implementation, the entity recognition can be performed by identifying and parsing specific parameters or configuration options of the scheduling task. For example, the entity recognition can identify the number of iterations, batch size, accuracy requirements, etc. of the model training.

[0082] In a possible implementation, the constraint analysis may include but is not limited to at least one of the following: resource budget, computing power distribution, priority level, and expected execution time of the scheduling task.

[0083] In a possible implementation, the resource matching and allocation module is used to perform availability check, scheduling strategy matching and scheduling algorithm matching on the scheduling task information, and determine the scheduling strategy for executing the scheduling task.

[0084] In a possible implementation, the availability check may include real-time tracking of the status of all computing nodes, including load level, availability, etc.

[0085] In a possible implementation, the above-mentioned scheduling strategy matching may include configuring appropriate scheduling strategies on demand, such as cost-aware scheduling strategies, load-aware scheduling strategies, energy efficiency-aware scheduling strategies, and service level agreement (SLA)-aware scheduling strategies.

[0086] In a possible implementation, the above-mentioned scheduling algorithm matching may include configuring a suitable scheduling algorithm on demand, such as multi-objective optimization, scheduling rules based on learning knowledge, constraint satisfaction problems, and heuristic scheduling algorithms.

[0087] In a possible implementation, the task acceptance and execution module is used for at least one of the following: scheduling service activation, scheduling work order acceptance, and scheduling task acceptance.

[0088] In a possible implementation, the above-mentioned scheduling service activation may refer to selecting a suitable image based on the analyzed business requirements, configuring corresponding service parameters, and completing the creation of a scheduling business work order.

[0089] In a possible implementation, the above-mentioned scheduling work order acceptance may refer to processing the work order generated during the service activation, which needs to be strictly reviewed. After passing the review, a corresponding scheduling task order is generated.

[0090] In one possible implementation, the above-mentioned scheduling task acceptance may refer to the task order generated in the processing of work order acceptance. After the work order review is completed, a scheduling task order is generated for the execution of the actual scheduling business, and the corresponding task order is tracked to facilitate users to understand the progress of the scheduling task at any time.

[0091] In the embodiment of the present application, the scheduling memory module 122 may be used to provide historical scheduling information.

[0092] In a possible implementation, the scheduling memory module 122 may be used to provide historical information for reference for scheduling strategies.

[0093] In a possible implementation, the scheduling memory module 122 may provide the scheduling decision module 121 with historical or external related scheduling experience information for reference in scheduling decision making.

[0094] For example, Figure 3 As shown, the scheduling memory module 122 can provide relevant scheduling experience information such as user preferences, domain knowledge and scheduling history.

[0095] In a possible implementation, the user preference may refer to context learning based on information from user interactions, which may be implemented through prompt engineering.

[0096] In one possible implementation, the above domain knowledge can provide the ability to retain and recall long-term information, and provide relevant knowledge accumulated in the domain through Retrieval-Augmented Generation (RAG).

[0097] In one possible implementation, the above-mentioned scheduling history can be used to provide recorded historical scheduling task information, including task requirements, scheduling strategies, scheduling algorithms and allocated resource information, which is stored and retrieved through RAG.

[0098] In the embodiment of the present application, the above-mentioned scheduling tool module 123 can be used to provide a scheduling tool.

[0099] In a possible implementation, the scheduling tool module 123 may include a series of optional scheduling-related tools.

[0100] For example, Figure 3 As shown, the scheduling tool module 123 may include resource retrieval, scheduling algorithm, scheduling strategy, scheduling service activation, scheduling work order acceptance and scheduling task acceptance, etc.

[0101] In one possible implementation, the above-mentioned resource retrieval may include the retrieval of scheduling-related resources such as computing power, models, samples, and the like, as well as the retrieval of scheduling memory.

[0102] In a possible implementation, the above-mentioned scheduling algorithm may include specific scheduling algorithms.

[0103] In a possible implementation, the above scheduling strategy may include specific scheduling strategies.

[0104] In a possible implementation, the above-mentioned scheduling service activation may refer to creating a scheduling work order based on the business demand information. Among them, the scheduling service may include various types, such as computing power scheduling, sample scheduling, model scheduling, and task scheduling.

[0105] In a possible implementation, the above-mentioned scheduling work order acceptance may refer to being able to check the status of the scheduling business work order and review it, and automatically generate a scheduling task order after the review is confirmed.

[0106] In a possible implementation, the above-mentioned scheduling task acceptance may refer to being able to check the status of the scheduling task list and complete the execution of the scheduling task.

[0107] In a possible implementation, the scheduling management module 12 may further include a tool execution module.

[0108] In a possible implementation, the tool execution module can be specifically executed by the scheduling tool selected by the scheduling decision. It can be understood that the tool execution is the scheduling decision module 121 parsing the scheduling task, matching and allocating resources, and selecting the corresponding scheduling tool to execute the specific scheduling task.

[0109] In a possible implementation, the scheduling management module 12 may further include: a scheduling agent module.

[0110] The scheduling agent module is used to maintain the operation of the scheduling management module 12 .

[0111] In a possible implementation, the scheduling agent module can be an agent program with a large language model (LLM) as the core, responsible for generating scheduling instructions and executing actions, aiming to achieve some scheduling tasks set by users. It can be understood that LLM can obtain user interaction information (i.e., scheduling task information submitted by users), choose to use preset or newly created scheduling tools, and complete scheduling tasks in an iterative operation mode.

[0112] In a possible implementation, the scheduling agent module is used to maintain the operation of the scheduling management module 12, and may include:

[0113] The scheduling decision module 121 parses the scheduling task information of the scheduling task, and based on the historical scheduling information provided by the scheduling memory module 122, the parsed scheduling task information is used to match and allocate resources for the scheduling task, and the scheduling strategy for executing the scheduling task is determined. Based on the scheduling strategy, a scheduling instruction is generated, and the scheduling tool and available resource data provided by the scheduling tool are used to execute the scheduling task corresponding to the scheduling instruction.

[0114] In one possible implementation, combining Figure 1 ,like Figure 4 As shown, the scheduling system 10 may further include: a system management module 13 , a visual management module 14 and a resource large screen module 15 .

[0115] Among them, the above-mentioned system management module can be used to maintain the operation of the scheduling system.

[0116] In one possible implementation, the system management module 13 can cover a series of basic functions of the scheduling system 10, which is the key to ensuring the normal operation of the electric power artificial intelligence computing power scheduler.

[0117] In a possible implementation, the system management module 13 may include a role management module.

[0118] In a possible implementation, the role management module is used to manage the resource access rights of tenants and users in the scheduling system, and the resource access rights are resource access rights across interconnected systems.

[0119] Exemplarily, the role management module can be used to support tenants and users in the scheduling system to isolate resource access rights across docking systems.

[0120] In a possible implementation, the system management module 13 may further include a tenant management module, a user management module, a dictionary management module, a log management module and other modules.

[0121] In a possible implementation, the visual management module 14 may be used to support visual management of resource data obtained from the docking system, and provide operations such as resource information list display and query.

[0122] In a possible implementation, the visual management module 14 may include functions such as computing power visualization, sample visualization, model visualization, and task visualization.

[0123] In one possible implementation, computing power visualization can display the connected platform, cluster, and node resource information within the cluster.

[0124] In one possible implementation, sample visualization can display sample information required for model training and testing.

[0125] In one possible implementation, the model visualization may display all available model image resources.

[0126] In one possible implementation, task visualization may display computing task information of all docked systems.

[0127] In a possible implementation, the resource large screen module 15 can be used to support the presentation of resource data statistics obtained from the docking system in different dimensions such as region and time in the form of a large screen.

[0128] In a possible implementation, the resource large screen module 15 may include functions such as resource overview, central resources, provincial cloud resources and model services.

[0129] In one possible implementation, the resource overview may provide a topological visualization of the global docking system and the Internet, and a graphical representation of the statistical analysis results of global resource data.

[0130] In a possible implementation, the central resources may provide topological visualization of the interconnected central systems and graphical representation of statistical analysis results of resource data of the central systems.

[0131] In one possible implementation, the provincial cloud resources may provide regional location visualization of the provincial system and a graphical representation of the statistical analysis results of the provincial cloud system's resource data.

[0132] In one possible implementation, the model service may provide graphical representations of statistical analysis results of global models, samples, and scheduling tasks.

[0133] like Figure 5 The above-mentioned at least one docking system may include at least one of the following: a local artificial intelligence training platform, a local artificial intelligence reasoning platform, a network management platform, a sample library, and a model library.

[0134] In one possible implementation, Figure 5 The local system management of the local artificial intelligence training platform shown in can cover a series of basic functions to ensure the normal operation of the platform; resource management can be responsible for the unified management and monitoring of resource data such as computing, storage, samples, and models; model training can be responsible for local model training tasks.

[0135] It should be noted that the above-mentioned scheduling system 10 can perform interface authentication and authorization with the local artificial intelligence training platform to obtain local computing, storage, sample, model and other resource data and model training task data.

[0136] In one possible implementation, Figure 5 The local system management of the local artificial intelligence reasoning platform shown in can cover a series of basic functions to ensure the normal operation of the platform; resource management can be responsible for the unified management and monitoring of computing, storage, samples, models and other resources; model reasoning can be responsible for local model reasoning tasks.

[0137] It should be noted that the above-mentioned scheduling system 10 can perform interface authentication and authorization with the local artificial intelligence reasoning platform to obtain local computing, storage, sample, model and other resource data and model reasoning task data.

[0138] In one possible implementation, Figure 5 The local system management of the network management platform shown in can cover a series of basic functions to ensure the normal operation of the platform; network management can be responsible for the unified management and monitoring of network resources.

[0139] It should be noted that the above-mentioned scheduling system 10 performs interface authentication and authorization with the network management platform to obtain network resource data between various systems.

[0140] In one possible implementation, Figure 5 The local system management of the sample library shown in can cover a series of basic functions to ensure the normal operation of the platform; sample management can be responsible for the unified management and monitoring of sample data.

[0141] It should be noted that the scheduling system 10 performs interface authentication and authorization with the sample library to obtain sample data information of the sample library.

[0142] In one possible implementation, Figure 5 The system management of the model library shown in can cover a series of basic functions to ensure the normal operation of the platform; model management can be responsible for the unified management and monitoring of model resources.

[0143] It should be noted that the above-mentioned scheduling system 10 performs interface authentication and authorization with the model library to obtain the model data information of the model library.

[0144] It should be noted that the docking system corresponding to the above-mentioned scheduling system 10 includes but is not limited to the systems listed above. Others such as an integrated development platform for artificial intelligence training and reasoning, or a platform for integrated management of model samples can also be docked. In addition, there can be multiple docking systems of the same type, such as multiple local artificial intelligence training platforms and multiple local artificial intelligence reasoning platforms scattered across the country. This application embodiment is not specifically limited.

[0145] The embodiment of the present application provides a scheduling system that can use the available resource data in at least one docking system to perform resource scheduling for scheduling tasks through a resource access module and a scheduling management module, break the current phenomenon of computing power islands, and solve the problem of resource scheduling and coordination, thereby achieving supply and demand matching between computing power resources, improving resource utilization, and realizing global scheduling optimization of resources.

[0146] The present application embodiment provides a scheduling method, which can be applied to a scheduler, and the scheduler can include the above-mentioned scheduling system. Figure 6 As shown, the method may include the following steps S101-S104.

[0147] S101. A scheduler obtains available resource data from at least one docking system.

[0148] In a possible implementation, the above scheduler can also be called an electric power artificial intelligence computing power scheduler. The two have the same meaning and can be used interchangeably in the text.

[0149] In a possible implementation, the at least one docking system may synchronize local resource data to a scheduling system in the scheduler. In other words, the scheduler may obtain local resource data of the at least one docking system.

[0150] In a possible implementation, the above step S101 can be specifically implemented through the following step S101a.

[0151] S101a. The scheduler obtains at least one of the following from at least one docking system:

[0152] The computing resources of at least one docking system, the network resource data of at least one docking system, the sample data information of at least one docking system, the model information of at least one docking system, and the computing task information of at least one docking system.

[0153] In a possible implementation, after obtaining the available resource data in the at least one docking system, the scheduler may also perform global visual management of the obtained local resource data of each docking system.

[0154] In a possible implementation, a user of the scheduler can log in to the scheduling system in the scheduler and view global resource data.

[0155] S102: The scheduler parses the scheduling task information of the scheduling task.

[0156] In a possible implementation, the scheduling task information may be a scheduling service requirement submitted by a user in a natural language.

[0157] It is understandable that the scheduler can formulate a resource scheduling solution based on the scheduling business requirements submitted by the user, so that the user can view and confirm the resource scheduling solution given by the scheduler.

[0158] In one possible implementation, the scheduling services corresponding to the scheduling service requirements submitted by the above-mentioned users in natural language may include multiple types such as computing power scheduling, sample scheduling, model scheduling and task scheduling.

[0159] In one possible implementation, the computing power scheduling may support automatic allocation on demand or allocation of available computing power in a specified area or platform.

[0160] In a possible implementation, the sample scheduling may support sample set transmission between docking systems according to user-specified sample sets and scheduling paths.

[0161] In a possible implementation, the above-mentioned model scheduling can support the transmission of sample sets between docking systems according to the user-specified model and scheduling path.

[0162] In a possible implementation, the above task scheduling can support the configuration of computing task information and scheduling policies, and automatically allocate required resources to computing tasks.

[0163] In a possible implementation, the scheduling agent module of the scheduling management module in the scheduler may be responsible for interacting with the user, receiving the business requirements submitted by the user, and sending them to the scheduling decision module.

[0164] In one possible implementation, combining Figure 6 ,like Figure 7 As shown, the above step S102 can be specifically implemented through the following step S102a.

[0165] S102a. The scheduler performs at least one of the following processes on the scheduling task information: intent recognition, entity recognition, parameter parsing, and constraint condition parsing.

[0166] In a possible implementation, the scheduling decision module in the scheduler can parse the scheduling task information submitted by the user through the task request resource parsing capability.

[0167] In a possible implementation, the above-mentioned analysis of the scheduling task information may include at least one of intent recognition, entity recognition, parameter analysis and constraint analysis.

[0168] In one possible implementation, the above-mentioned intent recognition may refer to the scheduler analyzing the keywords, phrases and sentence structures in the user interaction information through a pre-trained large language model LLM (such as Llama, Qwen, GPT, Bert, etc.) to determine the specific operation the user wants to perform.

[0169] For example, a user can enter "I want to train an image classification model" in the scheduler, so that LLM can recognize that this is a deep learning task about image classification.

[0170] In a possible implementation, the above entity recognition may refer to the scheduler extracting key entities from the scheduling task information, such as the computing framework used, the data set, the expected output, etc.

[0171] For example, the user mentioned the use of PyTorch framework and CIFAR-10 dataset in the scheduling task information submitted.

[0172] In one possible implementation, the above parameter parsing may refer to identifying and parsing task-specific parameters or configuration options, such as resource configuration for model training, and training hyperparameters such as number of iterations, batch size, accuracy requirements, etc.

[0173] For example, the user mentioned the type and number of graphics processing units (GPUs), batch size = 64, and the number of epochs for training the entire training dataset through the neural network = 50 in the scheduling task information submitted. It can be understood that training hyperparameters help to more accurately assess resource requirements.

[0174] In a possible implementation, the constraint analysis may include but is not limited to analysis of the task's resource budget, computing power distribution, priority level, and expected execution time.

[0175] For example, a user may mention in the submitted scheduling task information that he / she prefers to use a specific type of hardware or has specific performance goals, or that some tasks need to be completed within a specific time period.

[0176] S103: The scheduler matches and allocates resources for the scheduling tasks based on the parsed scheduling task information based on the historical scheduling information, determines a scheduling strategy for executing the scheduling tasks, and generates a scheduling instruction based on the scheduling strategy.

[0177] In a possible implementation, the scheduling decision module in the above scheduler can make decisions and plans for scheduling tasks through resource matching and allocation capabilities.

[0178] In one possible implementation, combining Figure 6 ,like Figure 8 As shown, the above step S103 can be specifically implemented through the following steps S103a and S103b.

[0179] S103a: The scheduler matches and allocates resources of the scheduling tasks for the parsed scheduling task information based on the historical scheduling information to obtain at least one scheduling strategy.

[0180] S103b: The scheduler determines a scheduling strategy for executing the scheduling task based on the user's selection of at least one scheduling strategy, and generates a scheduling instruction based on the scheduling strategy.

[0181] It is understandable that the scheduler can match and allocate resources of the scheduling task based on the parsed scheduling task information based on the historical scheduling information, and obtain at least one scheduling strategy for the user to select. Thus, the user can select the scheduling strategy to be executed from the at least one scheduling strategy according to his or her own preferences or needs.

[0182] In a possible implementation, the "matching and allocating resources of the scheduling task for the parsed scheduling task information" in the above step S103 can be specifically implemented by the following step A.

[0183] A. The scheduler performs availability check, scheduling strategy matching and scheduling algorithm matching on the scheduling task information.

[0184] In a possible implementation, the availability check may include real-time tracking of the status of all computing nodes, such as the load level and availability of the nodes.

[0185] It is understandable that resources are only considered for allocation to new scheduling tasks when they are in an idle or low-load state.

[0186] In a possible implementation, the above-mentioned scheduling strategy matching may include configuring appropriate scheduling strategies on demand, such as cost-aware scheduling strategies, load-aware scheduling strategies, energy efficiency-aware scheduling strategies, and SLA-aware scheduling strategies.

[0187] In a possible implementation, the above-mentioned scheduling algorithm matching may include configuring a suitable scheduling algorithm on demand, such as multi-objective optimization, scheduling rules based on learning knowledge, constraint satisfaction problems, and heuristic scheduling algorithms.

[0188] In a possible implementation, the above multi-objective optimization can be achieved by designing a scoring function that comprehensively considers multiple factors (such as performance, cost, energy consumption, etc.) to score each potential resource option, and then select the one with the highest score as the final allocation.

[0189] In a possible implementation, the above heuristic search can use heuristic methods (such as genetic algorithms, simulated annealing, etc.) to find approximate optimal solutions for complex scheduling problems.

[0190] In a possible implementation, the above-mentioned Constraint Satisfaction Problem (CSP) can model the resource allocation problem as a CSP, and find the best resource combination that satisfies all constraints through a solver.

[0191] S104: The scheduler uses the scheduling tool and available resource data to execute the scheduling task corresponding to the scheduling instruction.

[0192] In a possible implementation, the above step S104 may be specifically implemented through the following step S104a.

[0193] S104a. The scheduler sends the scheduling task to the docking system indicated by the scheduling strategy according to the scheduling instruction.

[0194] In a possible implementation, the scheduler may send the resource scheduling policy to the corresponding docking system.

[0195] In a possible implementation, after receiving the resource scheduling policy sent by the scheduler, each docking system may execute the resource scheduling policy sent by the scheduler according to the scheduling instruction.

[0196] In a possible implementation, each docking system may feed back the execution status of the resource scheduling strategy and the updated local resource data to the scheduler.

[0197] In a possible implementation, the scheduler may also obtain the resource scheduling policy execution status of each docking system and updated local resource data.

[0198] In a possible implementation, the scheduler feeds back the execution status of its scheduling service to the user.

[0199] In one possible implementation, combining Figure 6 ,like Fig. 9 The above step S104 can be specifically performed through the following step S104b.

[0200] S104b: The scheduler uses the scheduling tool and the available resource data to perform at least one of the following: scheduling service activation, scheduling work order acceptance, and scheduling task acceptance.

[0201] In one possible implementation, the scheduling tool module in the scheduler can call resource retrieval, scheduling strategy, scheduling algorithm and other tools to complete the planning of scheduling tasks based on the decisions and plans made by the scheduling decision module on the scheduling tasks.

[0202] In a possible implementation, the tool execution module in the scheduler can execute corresponding resource retrieval, scheduling strategy, scheduling algorithm and other tools.

[0203] In one possible implementation, the above-mentioned resource retrieval may include the retrieval of scheduling-related resources such as computing power, models, samples, and the retrieval of scheduling memory.

[0204] In a possible implementation, the above-mentioned scheduling algorithm may include specific scheduling algorithms; the above-mentioned scheduling strategy includes specific scheduling strategies.

[0205] In one possible implementation, the scheduling memory module in the scheduler can provide retrieval of user preferences, domain knowledge, and scheduling history, and provide the scheduling agent with historical or external relevant scheduling experience information for reference in scheduling decisions.

[0206] In one possible implementation, user preferences can be contextually learned based on information from user interactions, which can be achieved through prompt engineering.

[0207] In one possible implementation, domain knowledge can provide the ability to retain and recall long-term information, providing relevant knowledge accumulated in the domain through RAG.

[0208] In one possible implementation, the scheduling history may be used to provide recorded historical scheduling task information, including task requirements, scheduling strategies, scheduling algorithms, and allocated resource information, and stored and retrieved through the RAG.

[0209] In one possible implementation, the scheduling tool module in the scheduler can call the scheduling service activation tool to complete the creation of the scheduling service work order. The scheduling service activation tool can complete the creation of the scheduling service work order based on the analyzed business requirements, such as the required sample set and model, as well as the computing power requirements, and select the corresponding scheduling algorithm and scheduling strategy.

[0210] In a possible implementation, a tool execution module in the scheduler may execute a corresponding scheduling service activation tool.

[0211] In a possible implementation, the scheduling agent module in the scheduler can receive the created scheduling business work order and provide feedback to the user.

[0212] In one possible implementation, when the review fails, the user modifies the created scheduling business work order, and then receives the confirmed scheduling business work order through the scheduling agent module, such as changing the resource configuration; when the review passes, the confirmed scheduling business work order can be directly received through the scheduling agent module.

[0213] In a possible implementation, the scheduling tool module in the scheduler can call the scheduling work order acceptance tool and the scheduling task acceptance tool to complete the creation and execution of the scheduling task order. Among them, the work order generated in the scheduling work order acceptance processing service activation can be the corresponding scheduling task order generated after the user reviews and approves the work order; the scheduling task acceptance processing work order acceptance task order can be used for the scheduling task order for the actual execution of the scheduling business, and track the corresponding task order, so that the user can understand the progress of the scheduling task at any time.

[0214] In a possible implementation, the tool execution module in the scheduler can execute the corresponding scheduling work order acceptance tool and scheduling task acceptance tool.

[0215] In a possible implementation, the scheduling agent module in the scheduler can receive the scheduling task order information and provide feedback on its execution status to the user.

[0216] In a possible implementation, a scheduling memory module in the scheduler stores the current scheduling task information in the scheduling history.

[0217] Exemplarily, the scheduling memory module may use a vector storage method to store the current scheduling task information.

[0218] In a possible implementation, the user can view the information related to the scheduling task order in the scheduler to understand the execution status.

[0219] In a possible implementation, the scheduling method provided in the embodiment of the present application may further include the following step S105.

[0220] S105: The scheduler manages the resource access rights of tenants and users in the scheduling system.

[0221] The resource access permission mentioned above may be a resource access permission across docking systems.

[0222] In a possible implementation, the role management module in the scheduler can support the isolation of resource access rights for tenants and users in the scheduling system across interconnected systems.

[0223] The scheduling method provided in the embodiment of the present application can use the available resource data in at least one docking system to perform resource scheduling for scheduling tasks through a resource access module and a scheduling management module, break the current phenomenon of computing power islands, and solve the problem of resource scheduling and coordination, thereby achieving supply and demand matching between computing power resources, improving resource utilization, and realizing global scheduling optimization of resources.

[0224] The following is an illustrative example of the scheduling method provided in the embodiments of the present application.

[0225] like Fig.10 As shown, the scheduling method may include the following steps S201 to S209.

[0226] S201. A scheduler synchronizes resource data of at least one docking system.

[0227] The scheduler can obtain local resource data of multiple local artificial intelligence training platforms and local artificial intelligence reasoning platforms, which may include computing resource data, storage resource data, sample data information, model data information, and computing task information. In addition, the scheduler can also obtain network resource data of the network management platform, sample data information of the sample library, and model data information of the model library.

[0228] In one possible implementation, the computing power resource data may include computing power type, identification and geographic information, processor model, configuration, scale and status; the storage resource data may include storage type, storage capacity, storage performance, identification and geographic information; the network resource data may include network type, network nodes, routing topology and network performance; the computing task information may include task type, creation time, task status, task introduction, task usage resources and other information; the sample data information may include sample type, version information, creation time, sample set introduction, scale, sample link, identification and regional information and other information; the model data information may include model type, version information, creation time, model introduction, scale, model link, identification and regional information and other information.

[0229] S202: The scheduler performs visual management of resource data.

[0230] In one possible implementation, the scheduler can perform global visual management of the local resource data of each docking system obtained, have a list query function for resource data, and support viewing detailed information.

[0231] S203. The scheduler supports viewing resource data.

[0232] In a possible implementation, a user may log in to a scheduling system in a scheduler to view global resource data.

[0233] For example, the user can filter and view resource data by type, region, platform, etc.

[0234] S204: The scheduler receives the scheduling service requirement submitted by the user.

[0235] In a possible implementation, the user may submit a scheduling service in a natural language. It is understood that the scheduler obtains the scheduling task to be executed based on the scheduling service requirement.

[0236] For example, users can use a task scheduling service for resource allocation of model fine-tuning training tasks, use sample set X and model W, and require 8*A100 GPUs for intelligent computing power. Please pay attention to reducing resource fragmentation during the allocation process and submit business demand information.

[0237] S205: The scheduler provides a resource scheduling strategy.

[0238] In a possible implementation, the scheduling management module of the scheduler may formulate a resource scheduling strategy based on the scheduling service requirements submitted by the user.

[0239] In one possible implementation, the scheduling management module can parse the business demand information, make decisions and plans for scheduling tasks through resource matching and allocation capabilities, call resource retrieval, scheduling strategy, scheduling algorithm and other tools to complete the planning of scheduling tasks, and call the scheduling business activation tool to complete the creation of scheduling business work orders.

[0240] In one possible implementation, the scheduling work order information may include the sample set X and the model W, the computing power requirement of 8*A100 GPU, the resource fragmentation minimization scheduling strategy, and the best fit algorithm (Best Fit) scheduling algorithm, that is, among all the optional idle resource blocks, the resource block closest to the required size is selected for allocation.

[0241] In one possible implementation, after the user reviews and confirms the scheduling work order, the scheduling management module can call the scheduling work order acceptance and scheduling task acceptance tools to complete the creation and execution of the scheduling task order. In other words, the execution plan is: the scheduler selects the local artificial intelligence training platform P with the highest resource utilization rate and meeting the computing power requirements. 1 Allocate computing resources and store the 1 The sample set X and the model library M 1 The model W is dispatched to the local artificial intelligence training platform P 1 .

[0242] S206: The scheduler confirms the resource scheduling strategy.

[0243] In a possible implementation, the user can view and confirm the resource scheduling policy given by the scheduler.

[0244] S207: The scheduler sends a resource scheduling strategy.

[0245] In a possible implementation, the scheduler may send the resource scheduling policy to the corresponding docking system.

[0246] Specifically, the above may include one sample data flow: the scheduler sends the sample data to the sample library D 1 Get the download address of sample set X, and then send the download address of sample set X to the local artificial intelligence training platform P 1 ; 1st model data flow: the scheduler sends data to the model library M 1 Get the model W download address, and then send the model M download address to the local artificial intelligence training platform P 1 ; 1st model fine-tuning training task resource allocation: The scheduler allocates resources to the local artificial intelligence training platform P 1 Obtain computing resources; 1 model fine-tuning training task instruction: The scheduler sends a command to the local artificial intelligence training platform P 1 Submit a model fine-tuning training task.

[0247] S208. The scheduler records the execution of the resource scheduling policy.

[0248] In a possible implementation, each of the above docking systems may execute the resource scheduling plan corresponding to the resource scheduling policy issued by the scheduler, and feedback the execution status of the resource scheduling plan and the updated local resource data to the scheduler.

[0249] Specifically, the execution of the above resource scheduling policy may include: Sample library D 1 Send the download address of sample set X to the scheduler; Local artificial intelligence training platform P 1 Obtain the download address of sample set X from the scheduler; The scheduler adds a new sample scheduling record; Model library M 1 Send the download address of model W to the scheduler; Local artificial intelligence training platform P 1 Obtain the download address of model W from the scheduler; The scheduler adds a new model scheduling record; The scheduler sends to the local artificial intelligence training platform P 1 Submit the computing resource allocation requirement; Local artificial intelligence training platform P 1 Allocate the required computing resources for the fine-tuning training task of the model; The scheduler adds a new computing power scheduling record; The scheduler sends to the local artificial intelligence training platform P 1 Submit the model fine-tuning training task; Local artificial intelligence training platform P 1 Update its resource data information to the scheduler, including the update of computing power resource data, storage resource data, sample data information, model data information, and computing task information; The scheduler adds a new task scheduling record, and jointly constitutes the log of this task scheduling service together with the previously added sample scheduling record, model scheduling record, and algorithm scheduling record.

[0250] S209. The scheduler feeds back the scheduling service situation.

[0251] In a possible implementation, the scheduler may feed back the execution status of its scheduling service to the user, and the user can view the log of this task scheduling service.

[0252] In this way, through the resource access module and the scheduling management module, the available resource data in at least one docking system can be used for resource scheduling of the scheduling task, breaking the existing computing power island phenomenon, solving the problems of resource scheduling and coordination, thereby realizing the supply-demand matching between computing power resources, improving resource utilization, and achieving global scheduling optimization of resources.

[0253] In the above embodiments of the present application, each of the various solutions can be combined on the premise of not being contradictory.

[0254] The embodiment of the present application provides a scheduling device, which can be a scheduler or a chip or system on chip in the scheduler. The scheduling device can implement the functions performed by the scheduler in the above aspects or possible designs, and the functions can be implemented by hardware, such as: in one possible design, the scheduling device may include: a processor and a communication interface, and the processor can be used to support the scheduling device to implement the functions involved in the above second aspect or any possible design of the second aspect, for example: the scheduling device obtains available resource data from at least one docking system through the communication interface; and parses the scheduling task information of the scheduling task through the processor; based on the historical scheduling information, the parsed scheduling task information is used to match and allocate the resources of the scheduling task, determine the scheduling strategy for executing the scheduling task, and generate the scheduling instruction based on the scheduling strategy; use the scheduling tool and the available resource data to execute the scheduling task corresponding to the scheduling instruction.

[0255] An embodiment of the present application provides a computer program product including instructions. When the computer program product is run on a computer, the computer is enabled to execute the scheduling method in the above method embodiment.

[0256] An embodiment of the present application also provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a computer, the computer executes the scheduling method in the method flow shown in the above method embodiment.

[0257] Among them, the computer readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a register, a hard disk, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above, or any other form of computer readable storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application-specific integrated circuit (ASIC). In the embodiments of the present application, a computer-readable storage medium may be any tangible medium that contains or stores a program, which may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0258] An embodiment of the present application provides a computer program product including instructions, which, when executed on a computer, enables the computer to execute the scheduling method as described above.

[0259] Since the scheduling system, computer-readable storage medium, and computer program product in the embodiments of the present application can be applied to the above-mentioned scheduling method, the technical effects that can be obtained can also refer to the above-mentioned method embodiments, and the embodiments of the present application will not be repeated here.

[0260] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0261] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0262] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0263] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0264] It should be noted that the terms "first" and "second" in the specification, claims and drawings of the present application are used to distinguish different objects rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or devices.

[0265] It should be understood that in the present application, "at least one (item)" means one or more, "more than one" means two or more, "at least two (items)" means two or three and more than three, and "and / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0266] Through the description of the above implementation methods, technical personnel in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0267] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0268] The units described as separate components may or may not be physically separated, and the components shown as units may be one physical unit or multiple physical units, that is, they may be located in one place or distributed in multiple different places. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0269] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0270] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium, including several instructions to enable a device (which can be a single-chip microcomputer, chip, etc.) or a processor (processor) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard drives, ROM, RAM, magnetic disks or optical disks.

[0271] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present application shall be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A scheduling system, characterized in that: include: Resource access module and scheduling management module; Wherein, the resource access module is used to obtain available resource data from at least one docking system; The scheduling management module is used to determine the scheduling strategy for executing the scheduling task according to the resources required for the scheduling task execution and the available resource data, and execute the scheduling task according to the scheduling strategy. The scheduling management module includes: a scheduling decision module, a scheduling memory module and a scheduling tool module; The scheduling decision module is used to match and allocate resources for the scheduling task based on the scheduling task information, determine the scheduling strategy for executing the scheduling task, and guide the acceptance and execution of the scheduling task according to the scheduling strategy; The scheduling memory module is used to provide historical scheduling information; The scheduling tool module is used to provide a scheduling tool.

2. The system according to claim 1, characterized in that The scheduling system also includes: a system management module; Wherein, the system management module is used to maintain the operation of the scheduling system.

3. The system according to claim 2, characterized in that The system management module includes a role management module; The role management module is used to manage the resource access rights of tenants and users in the scheduling system, and the resource access rights are cross-docking system resource access rights.

4. The system according to claim 1, characterized in that The resource access module includes at least one of the following: an access management module, a computing power management module, a network management module, a sample management module, a model management module and a task management module; The access management module is used to support authentication and authorization for docking with the at least one docking system, and supports multiple resource data access methods; The computing power management module is used to support the management and monitoring of computing resources of the at least one docking system; The network management module is used to obtain network resource data of the at least one docking system; The sample management module is used to support the acquisition of sample data information of the at least one docking system; The model management module is used to support obtaining model information of the at least one docking system; The task management module is used to support obtaining computing task information of the at least one docking system.

5. The system according to claim 1, characterized in that The scheduling management module also includes: a scheduling agent module; Wherein, the scheduling agent module is used to maintain the operation of the scheduling management module.

6. The system according to claim 5, characterized in that The scheduling agent module is used to maintain the operation of the scheduling management module, including: The scheduling decision module parses the scheduling task information of the scheduling task, and based on the historical scheduling information provided by the scheduling memory module, the parsed scheduling task information is used to match and allocate resources for the scheduling task, determine the scheduling strategy for executing the scheduling task, and generate a scheduling instruction based on the scheduling strategy, and use the scheduling tool provided by the scheduling tool and the available resource data to execute the scheduling task corresponding to the scheduling instruction.

7. The system according to claim 1 or 6, characterized in that: The scheduling decision module includes a parsing module, a resource matching and allocation module, and a task acceptance and execution module; The parsing module is used to perform at least one of the following processing on the scheduling task information: intent recognition, entity recognition, parameter parsing and constraint condition parsing; The resource matching and allocation module is used to perform availability check, scheduling strategy matching and scheduling algorithm matching on the scheduling task information, and determine the scheduling strategy for executing the scheduling task; The task acceptance and execution module is used for at least one of the following: scheduling service activation, scheduling work order acceptance and scheduling task acceptance.

8. A scheduling method, characterized in that: The dispatching system according to claims 1 to 7 comprises: Obtaining available resource data from at least one docking system; Parse the scheduling task information of the scheduling task; Based on the historical scheduling information, the parsed scheduling task information is used to match and allocate resources for the scheduling task, a scheduling strategy for executing the scheduling task is determined, and a scheduling instruction is generated based on the scheduling strategy; The scheduling task corresponding to the scheduling instruction is executed using a scheduling tool and the available resource data.

9. The method according to claim 8, characterized in that The method of matching and allocating resources of the scheduling task based on the parsed scheduling task information based on the historical scheduling information and determining a scheduling strategy for executing the scheduling task includes: Based on the historical scheduling information, performing resource matching and allocation of the scheduling task on the parsed scheduling task information to obtain at least one scheduling strategy; Based on the user's selection of the at least one scheduling strategy, a scheduling strategy for executing the scheduling task is determined.

10. The method according to claim 8, characterized in that The method further comprises: The resource access rights of tenants and users in the scheduling system are managed, and the resource access rights are cross-docking system resource access rights.

11. The method according to claim 8, characterized in that The obtaining of available resource data from at least one docking system comprises: Obtain at least one of the following from at least one docking system: The computing resources of the at least one docking system, the network resource data of the at least one docking system, the sample data information of the at least one docking system, the model information of the at least one docking system, and the computing task information of the at least one docking system.

12. The method according to claim 8, characterized in that The parsing of the scheduling task information includes: The scheduling task information is processed by at least one of the following: intent recognition, entity recognition, parameter parsing, and constraint condition parsing.

13. The method according to claim 8, characterized in that The performing resource matching and allocation of the scheduling task on the parsed scheduling task information includes: Availability check, scheduling strategy matching and scheduling algorithm matching are performed on the scheduling task information.

14. The method according to claim 8, characterized in that The executing of the scheduling task corresponding to the scheduling instruction includes at least one of the following: scheduling service activation, scheduling work order acceptance and scheduling task acceptance.

15. The method according to claim 8, characterized in that The executing the scheduling task corresponding to the scheduling instruction includes: According to the scheduling instruction, the scheduling task is sent to the docking system indicated by the scheduling strategy.

16. A scheduling device, characterized in that: include: A processor, a memory and a communication interface; wherein the communication interface is used to obtain available resource data from at least one docking system; The memory is used to store one or more programs, and the one or more programs include computer-executable instructions. When the scheduling device is running, the processor executes the computer-executable instructions stored in the memory to enable the scheduling device to perform the method described in any one of claims 8-15.

17. A computer-readable storage medium, characterized in that: The readable storage medium stores instructions, and when the instructions are executed, the method according to any one of claims 8 to 15 is implemented.

Images