Service control system and method, electronic device, storage medium, and program product

Abstract

The present application provides a service control system and method, an electronic device, a storage medium, and a program product. The service control system comprises a service orchestration module and a service control module. The service orchestration module is used for orchestrating artificial intelligence (AI) service requests and sending an orchestration result to the service control module, the orchestration result comprising an AI workflow, an execution result that needs to be achieved by the AI workflow, a plurality of AI tasks, and an execution result that needs to be achieved by each AI task. The service control module is used for executing the plurality of AI tasks, and, on the basis of actual execution results of the plurality of AI tasks, the execution result that needs to be achieved by the AI workflow, and the execution result that needs to be achieved by each AI task, adjusting an execution policy of the AI workflow and an execution policy of each AI task.

Description

Business control systems and methods, electronic devices, storage media and program products

[0001] Cross-references

[0002] This application claims priority to Chinese Patent Application No. 202410778533.6, filed on June 17, 2024, entitled "Business Control System and Method, Electronic Device, Storage Medium and Program Product", the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application relates to the field of intelligent business control, and more particularly to a business control system and method, electronic device, storage medium and program product. Background Technology

[0004] Endogenous intelligence typically refers to the deep integration of network connectivity with Artificial Intelligence (AI) elements (such as computing power, algorithms, data, and connectivity) at the architectural level to build a complete intelligent system within the network. Through intelligent distributed deployment and collaboration, it provides AI as-a-Service (AIaaS) services on demand, both internally and externally, ensuring efficient and high-quality intelligent services, facilitating intelligent autonomy of the network, and promoting the collaboration and integration of the intelligent ecosystem. In endogenous intelligence systems, the collaborative management and control of AI element resources is a key factor in the effectiveness of endogenous intelligence. However, current technologies typically only allow for the management and control of a single AI element resource, and do not yet support the collaborative management and control of multiple AI element resources. Summary of the Invention

[0005] This application provides a business control system and method, electronic device, storage medium, and program product.

[0006] This application is implemented as follows:

[0007] Firstly, a business control system is provided, including a business orchestration module and a business control module, wherein: the business orchestration module is used to orchestrate artificial intelligence (AI) service requests and send the orchestration results to the business control module; the orchestration results include an AI workflow, the execution results to be achieved by the AI ​​workflow, multiple AI tasks, and the execution results to be achieved by each AI task; the business control module is used to execute the multiple AI tasks and adjust the execution strategy of the AI ​​workflow and the execution strategy of each AI task based on the actual execution results of the multiple AI tasks, the execution results to be achieved by the AI ​​workflow, and the execution results to be achieved by each AI task.

[0008] Secondly, a business control method is provided, comprising: orchestrating AI service requests to obtain an orchestration result, the orchestration result including an AI workflow, the execution result to be achieved by the AI ​​workflow, multiple AI tasks, and the execution result to be achieved by each of the AI ​​tasks; executing the multiple AI tasks, and adjusting the execution strategy of the AI ​​workflow and the execution strategy of each AI task based on the actual execution result of the multiple AI tasks, the execution result to be achieved by the AI ​​workflow, and the execution result to be achieved by each of the AI ​​tasks.

[0009] Thirdly, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the method as described in the second aspect.

[0010] Fourthly, a computer-readable storage medium is provided, wherein when instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the method described in the second aspect.

[0011] Fifthly, a computer program product is provided, the computer program product including a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps of the method as described in the second aspect. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in this application or related technologies, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 is a schematic diagram of the structure of a business control system according to an embodiment of this application;

[0014] Figure 2 is a schematic diagram of the structure of a business control system according to another embodiment of this application;

[0015] Figure 3 is a schematic diagram of the structure of a business control system according to another embodiment of this application;

[0016] Figure 4 is a flowchart illustrating a business control method according to an embodiment of this application;

[0017] Figure 5 is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0018] In 4G or 5G networks, network resources and system resources are organized to provide data connections to users. When providing services to users, data can be transmitted to users through data streams, and the transmission status of data streams (such as transmission latency, reachable traffic, etc.) can be determined through Quality of Service (QoS). In turn, the resources used when transmitting data streams can be adjusted to achieve control over data transmission resources.

[0019] 6G networks introduce AI services, which encompass various AI elements. When providing these services to users, the system needs to offer resources corresponding to these AI elements, such as computing power, algorithm resources, data resources, and connectivity resources. These resources are used to implement the AI ​​service. To meet user needs, managing these multiple AI resources is necessary. However, current technologies typically only allow for the management of one type of AI resource, such as data transmission resources, and do not provide comprehensive, coordinated management of multiple AI resources.

[0020] This application provides a business control system and method, electronic device, storage medium, and program product. When receiving an AI service request, the AI ​​service request can be organized into an AI workflow, the execution result to be achieved by the AI ​​workflow, multiple AI tasks, and the execution result to be achieved by each AI task. Then, taking the AI ​​task as the smallest execution unit, the execution status of each AI task is monitored. Based on the actual execution results of multiple AI tasks, the execution result to be achieved by each AI task, and the execution result to be achieved by the AI ​​workflow, the execution strategy of each AI task and the execution strategy of the AI ​​workflow are adjusted. Therefore, by adjusting the strategies of AI tasks and AI workflows, the various AI element resources used to provide AI services can be adjusted. This allows for dynamic and flexible control and adaptive adjustment of various AI element resources used to provide AI services, achieving collaborative management and control of multiple AI element resources and effectively improving the quality of AI services.

[0021] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in this application will be clearly and completely described below with reference to the accompanying drawings of one or more embodiments. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this application.

[0022] The terms "first," "second," etc., used in this application and the claims are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that this application can be implemented in orders other than those illustrated or described herein. Furthermore, in this application and the claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0023] It should be noted that the technical solutions provided in this application can be applied to intelligent endogenous network systems. In one exemplary implementation, when an intelligent endogenous network system provides AI services to the outside world, it can achieve collaborative management and control of multiple AI element resources based on the technical solutions provided in this application, thereby improving the quality of AI services. Furthermore, the technical solutions provided in this application can also be applied to the design of communication network equipment, including system designs that use Radio Access Network (RAN) or Core Network (CN) network nodes.

[0024] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0025] Figure 1 is a schematic diagram of the structure of a business control system according to an embodiment of this application.

[0026] The business control system 10 shown in Figure 1 includes a business orchestration module 11 and a business control module 12. The business orchestration module 11 orchestrates AI service requests and sends the orchestration results to the business control module 12. The orchestration results of the business orchestration module 11 may include an AI workflow, the execution results that the AI ​​workflow needs to achieve, multiple AI tasks, and the execution results that each AI task needs to achieve. The business control module 12 executes multiple AI tasks and adjusts the execution strategies of the AI ​​workflow and each AI task based on the actual execution results of the multiple AI tasks, the execution results that the AI ​​workflow needs to achieve, and the execution results that each AI task needs to achieve.

[0027] In some implementations, the AI ​​workflow may include at least one of the following:

[0028] The ID of the AI ​​workflow identifies different types of AI workflows. Different types of AI workflows consist of different AI tasks and correspond to different execution results that the AI ​​workflow needs to achieve.

[0029] The business object types of AI workflows include the business object types of AI workflows within the system;

[0030] The type of AI workflow, such as whether it is an event-based workflow or a periodic workflow;

[0031] An AI workflow includes a set of AI task types, such as data acquisition, model training, model inference, model validation, and model deployment. In some implementations, the set of AI task types in an AI workflow can be represented as {data acquisition (yes / no), model training (yes / no), model inference (yes / no), model validation (yes / no), model deployment (yes / no)}, where (yes / no) indicates whether the AI ​​task is executed in the AI ​​workflow.

[0032] The desired execution result of an AI workflow can be the execution result that the AI ​​workflow is expected to achieve. In some implementations, the desired execution result of an AI workflow may include at least one of the following:

[0033] The work status that an AI workflow needs to achieve can be, such as successful execution of the AI ​​workflow and execution time.

[0034] The QoS requirements for AI workflows can include computing power QoS, data QoS, algorithm QoS, and connection QoS. Among them, computing power QoS can include computing resource type and computing resource level, data QoS can include data volume level, algorithm QoS can include training real-time performance and inference real-time performance, and connection QoS can include traffic level. It should be noted that the definitions of these QoS requirements should use the same type and level definitions as the descriptions of computing power, data, algorithms, and connections in the system.

[0035] By taking the work status and / or QoS that the AI ​​workflow needs to achieve as the execution result that the AI ​​workflow needs to achieve, it is easier to measure the execution of the AI ​​workflow and thus determine whether the execution result of the AI ​​workflow meets the actual needs.

[0036] Multiple AI tasks correspond to AI workflows, which can be obtained by breaking down or orchestrating the AI ​​workflow into tasks. The execution result required for each AI task corresponds to the execution result required for the AI ​​workflow, which can also be obtained by breaking down or orchestrating the execution result required for the AI ​​workflow. These multiple AI tasks can have a temporal relationship (i.e., a sequential execution order). When executing multiple AI tasks, the business control module 12 can execute them sequentially according to their temporal relationship.

[0037] In some implementations, each AI task may include at least one of the following:

[0038] The ID of an AI task can identify different types of AI tasks, and different types of AI tasks can correspond to different resource types and the execution results that the AI ​​task needs to achieve.

[0039] The business object of the AI ​​task includes the business object type and primary key of the AI ​​task in the system;

[0040] AI tasks can include data collection, model training, model inference, model validation, and model deployment.

[0041] The desired execution result of an AI task can be the expected execution result that the AI ​​task can achieve. In some implementations, the desired execution result for each AI task may include at least one of the following:

[0042] The execution status that an AI task needs to achieve, such as successful execution of the AI ​​task and execution time;

[0043] The QoS that AI tasks need to achieve can include some of the four QoS categories: computing power QoS, data QoS, algorithm QoS, and connection QoS. Among them, computing power QoS can include computing power resource type and computing power resource level, data QoS can include data volume level, algorithm QoS can include training real-time performance and inference real-time performance, and connection QoS can include traffic level. It should be noted that the definitions of these QoS categories should use the same type and level definitions as the descriptions of computing power, data, algorithms, and connections in the system.

[0044] By taking the working state and / or QoS that an AI task needs to achieve as the execution result, it is easier to measure the execution status of the AI ​​task and thus determine whether the execution result meets the actual requirements. Furthermore, since the QoS that an AI task needs to achieve includes a portion of the QoS from computing power QoS, data QoS, algorithm QoS, and connection QoS, effectively breaking down the QoS that the AI ​​workflow needs to achieve, allowing each AI task to achieve a portion of these QoS values, after obtaining the actual execution results of multiple AI tasks, the management of some AI resource elements can be achieved by adjusting the execution strategy of a single AI task, and the coordinated management of multiple AI resource elements can be achieved by adjusting the execution strategies of multiple AI tasks. Optionally, in some implementations, the QoS that each AI task needs to achieve may include one QoS from the QoS that the AI ​​workflow needs to achieve, with multiple AI tasks corresponding to different QoS values. This allows the QoS that the AI ​​workflow needs to achieve to be broken down into individual QoS values, thereby enabling the management of one AI resource element by adjusting the execution strategy of a single AI task, and the coordinated management of multiple AI resource elements by adjusting the execution strategies of multiple AI tasks.

[0045] In some implementations, the actual result of each AI task may include at least one of the following:

[0046] The actual execution status of an AI task (i.e., the actual execution status achieved by the AI ​​task), such as whether the AI ​​task was successfully executed or suspended, and the actual execution time, etc.

[0047] The actual QoS of an AI task (i.e., the QoS actually achieved by the AI ​​task) includes the actual computing power QoS, data QoS, algorithm QoS, and some QoS in connection QoS. Computing power QoS can include computing power resource type, computing power resource level, etc. Data QoS can include data volume level, etc. Algorithm QoS can include training real-time performance, inference real-time performance, etc. Connection QoS can include traffic level, etc. It should be noted that the definitions of these QoS should use the same type and level definitions as the descriptions of computing power, data, algorithms, and connections in the system.

[0048] It should be noted that for each AI task, the actual execution result of the AI ​​task must be consistent with the execution result that the AI ​​task needs to achieve, so that the actual execution result of the AI ​​task can be compared with the required execution result. For example, if the execution result that the AI ​​task needs to achieve includes the computing power QoS that the AI ​​task needs to achieve, then the actual execution result of the AI ​​task includes the actual computing power QoS of the AI ​​task.

[0049] In some implementations, the business orchestration module may include a workflow orchestration module and a task orchestration module. The workflow orchestration module orchestrates AI service requests to obtain workflow orchestration results, which include the AI ​​workflow and the execution results that the AI ​​workflow needs to achieve. The task orchestration module orchestrates the workflow orchestration results to obtain task orchestration results, which include multiple AI tasks and the execution results that each AI task needs to achieve. In other words, when orchestrating AI service requests, the workflow orchestration module can first orchestrate the AI ​​service requests to obtain the AI ​​workflow and the execution results that the AI ​​workflow needs to achieve, and then the task orchestration module can orchestrate the AI ​​workflow to obtain multiple AI tasks and orchestrate the execution results that the AI ​​workflow needs to achieve to obtain the execution results that each AI task needs to achieve. Therefore, by implementing AI workflow orchestration and AI task orchestration separately through two modules, orchestration efficiency can be improved.

[0050] In some implementations, the AI ​​service request may include business information and service level agreement (SLA) information for the AI ​​service, which can be input by the user through a predefined AI service interface provided by the system. The business information for the AI ​​service may include at least one of the following:

[0051] The business objects of AI services can include the types and instance information of business objects that perform AI business in the system;

[0052] AI services can include at least one of the following business types: data collection, model training, model inference, model validation, and model deployment.

[0053] The service level agreement information for AI services may include at least one of the following:

[0054] The service priority of AI services can be represented by a numerical value. The larger the value, the higher the service priority of the AI ​​service. Under resource constraints, AI services with higher service priority can preempt the resources of AI services with lower service priority.

[0055] The algorithm performance level of an AI service can be represented by a numerical value. The larger the value, the higher the algorithm performance requirement of the AI ​​service and the greater the corresponding resource demand.

[0056] The service latency level of AI services can be represented by a numerical value. The higher the value, the higher the real-time requirement of the AI ​​service.

[0057] When an AI service request includes business information and service level agreement (SLA) information for the AI ​​service, the workflow orchestration module orchestrates the AI ​​service request to obtain an AI workflow orchestration result. In an exemplary embodiment, the workflow orchestration module determines a target AI workflow template from multiple AI workflow templates based on the business information and SLA information of the AI ​​service, and determines the AI ​​workflow defined in the target AI workflow template and its execution result as the workflow orchestration result. The multiple AI workflow templates can be predefined templates in the system. Optionally, the system can maintain an AI workflow template library, which can store multiple AI workflow templates. Each AI workflow template defines an AI workflow supported by the system and its execution result. The AI ​​workflow defined in the target AI workflow template matches the business information of the AI ​​service, and the execution result of the AI ​​workflow defined in the target AI workflow template matches the SLA information of the AI ​​service.

[0058] Each AI workflow template may include at least one of the following:

[0059] The AI ​​workflow ID identifies different types of AI workflows, which consist of different AI tasks and correspond to different execution results.

[0060] The business object types of AI workflows include the business object types of AI workflows within the system;

[0061] The type of AI workflow, such as whether it is an event-based workflow or a periodic workflow;

[0062] An AI workflow includes a set of AI task types, such as data acquisition, model training, model inference, model validation, and model deployment. In some implementations, the set of AI task types in an AI workflow can be represented as {data acquisition (yes / no), model training (yes / no), model inference (yes / no), model validation (yes / no), model deployment (yes / no)}, where (yes / no) indicates whether the AI ​​task is executed in the AI ​​workflow.

[0063] The execution result of the AI ​​workflow defined in each AI workflow template may include at least one of the following:

[0064] The working status of the AI ​​workflow can be, for example, the successful execution of the AI ​​workflow and the execution time.

[0065] QoS in AI workflows can include computing power QoS, data QoS, algorithm QoS, and connection QoS. Among them, computing power QoS can include computing resource type and computing resource level, data QoS can include data volume level, algorithm QoS can include training real-time performance and inference real-time performance, and connection QoS can include traffic level.

[0066] When orchestrating AI service requests, the workflow orchestration module first matches the business information of the AI ​​service with the AI ​​workflows defined in each AI workflow template to determine one or more matching AI workflow templates. Then, it matches the execution results of the AI ​​workflows defined in these templates with the service level agreement (SLA) information of the AI ​​service to determine the matching AI workflow template. This matching AI workflow template is the target AI workflow template. The AI ​​workflows defined in the target AI workflow template match the business information of the AI ​​service, and the execution results of the AI ​​workflows defined in the target AI workflow template match the SLA information of the AI ​​service. Matching the business information of the AI ​​workflows defined in the target AI workflow template can mean that the information of the AI ​​workflows defined in the target AI workflow template includes the business information of the AI ​​service. Matching the execution results of the AI ​​workflows defined in the target AI workflow template with the SLA information can mean that the execution results of the AI ​​workflows defined in the target AI workflow template meet (equal to or exceed) the requirements of the SLA information of the AI ​​service.

[0067] It should be noted that, in one possible implementation, there may be one or more target AI workflow templates. If there are multiple target AI workflow templates, the template that is closest to the business information and service level agreement information of the AI ​​service can be determined from the multiple target AI workflow templates as the final template, and the AI ​​service request can be orchestrated according to the template.

[0068] After determining the target AI workflow template, the workflow orchestration module can use the AI ​​workflow defined in the target AI workflow template as the AI ​​workflow obtained after orchestrating the AI ​​service request, and use the execution result of the AI ​​workflow defined in the target AI workflow template as the execution result that the AI ​​workflow obtained after orchestrating the AI ​​service request needs to achieve. Thus, the workflow orchestration result can be obtained.

[0069] Since the workflow orchestration module can orchestrate AI service requests based on AI workflow templates, it can improve orchestration efficiency. In addition, since the AI ​​workflow templates are provided by the system and define the AI ​​workflows supported by the system and the execution results of the AI ​​workflows, the results obtained by orchestration based on the AI ​​workflow templates can be supported by the system, avoiding the problem of orchestration failure caused by the lack of system support for the orchestration results.

[0070] In some implementations, after obtaining the workflow orchestration result, the workflow orchestration module, to facilitate subsequent execution of the AI ​​workflow, further determines whether the remaining system resources meet the resource requirements of the target AI workflow template, assuming the target AI workflow template is identified. If the remaining system resources meet the resource requirements of the target AI workflow template, the required resources for the target AI workflow template are deducted from the remaining system resources. After deducting the system resources, the AI ​​workflow orchestration is considered successful, and the AI ​​workflow orchestration process can be terminated. Optionally, in some implementations, if the remaining system resources do not meet the resource requirements of the target AI workflow template, the AI ​​workflow orchestration is considered to have failed, and a failure message can be returned for the AI ​​service request. This embodiment uses the example of the remaining system resources meeting the resource requirements of the target AI workflow template for illustration.

[0071] It should be noted that the workflow orchestration module can also use other orchestration methods when orchestrating AI service requests, such as using intelligent algorithms to orchestrate AI service requests, etc. Examples of other orchestration methods for AI service requests will not be provided here.

[0072] After the workflow orchestration module obtains the workflow orchestration result, it can send the workflow orchestration result to the task orchestration module. The task orchestration module is used to orchestrate the workflow orchestration result to obtain the task orchestration result.

[0073] In some implementations, the task orchestration module is used to orchestrate the workflow orchestration results to obtain task orchestration results. Specifically, the task orchestration module is used to determine a target AI task template from multiple AI task templates based on the workflow orchestration results, and to determine the multiple AI tasks defined in the target AI task template and the execution results of each AI task as the task orchestration results. The multiple AI task templates can be predefined templates in the system. Optionally, the system can maintain an AI task template library, which can store multiple AI task templates. Each AI task template defines multiple AI tasks and the execution results of each AI task. The multiple AI tasks defined in the target AI task template match the AI ​​workflow in the workflow orchestration results, and the execution results of each AI task defined in the target AI task template match the execution results that the AI ​​workflow in the workflow orchestration results needs to achieve.

[0074] For each AI task template, the multiple AI tasks defined in the template may include at least one of the following:

[0075] Multiple AI task IDs; different IDs can identify different types of AI tasks, and different types of AI tasks can correspond to different resource types and AI task execution results.

[0076] The business objects for multiple AI tasks include the business object type and primary key of each AI task in the system;

[0077] There are multiple types of AI tasks, and each type of AI task can include data collection, model training, model inference, model validation, model deployment, etc.

[0078] For each AI task template, the execution result of each AI task defined in the template may include at least one of the following:

[0079] The execution status of an AI task, such as whether the AI ​​task was successfully executed or the execution time;

[0080] QoS for AI tasks can include some of the following four aspects: computing power QoS, data QoS, algorithm QoS, and connection QoS. Among them, computing power QoS can include computing resource type and computing resource level, data QoS can include data volume level, algorithm QoS can include training real-time performance and inference real-time performance, and connection QoS can include traffic level.

[0081] When orchestrating AI service requests, the task orchestration module first matches the AI ​​workflows in the workflow orchestration results with multiple AI tasks defined in each AI task template to determine one or more matching AI task templates. Then, it matches the execution results of each AI task defined in these templates with the execution results of the AI ​​workflows in the workflow orchestration results to determine the matching AI task templates. These matching AI task templates are the target AI task templates. The multiple AI tasks defined in the target AI task templates are matched with the AI ​​workflows in the workflow orchestration results, and the execution results of each AI task defined in the target AI task templates are matched with the execution results of the AI ​​workflows in the workflow orchestration results. In this context, multiple AI tasks defined in the target AI task template are matched with AI workflows in the workflow orchestration results. This can be an AI workflow composed of multiple AI tasks defined in the target AI task template, including AI workflows in the workflow orchestration results. The execution result of each AI task defined in the target AI task template is matched with the execution result of the AI ​​workflow in the workflow orchestration results. This can be an AI workflow whose execution result, determined based on the execution result of each AI task defined in the target AI task template, meets (is equal to or exceeds) the requirement of the execution result of the AI ​​workflow in the workflow orchestration results.

[0082] Optionally, in some implementations, corresponding AI task templates can be maintained for different AI workflow templates. Multiple AI tasks defined in an AI task template are matched with AI workflows defined in the AI ​​workflow template, and the execution result of each AI task defined in the AI ​​task template is matched with the execution result of the AI ​​workflow defined in the AI ​​workflow template. In this way, when the task orchestration module performs task orchestration, it can directly determine the AI ​​task template corresponding to the target AI workflow template from multiple AI task templates and use that AI task template as the target AI task template for AI task orchestration. This eliminates the need to sequentially match the workflow orchestration result with multiple AI task templates, thereby shortening the AI ​​task orchestration time and improving the efficiency of AI task orchestration.

[0083] After determining the target AI task template, the task orchestration module can use the multiple AI tasks defined in the target AI task template as multiple AI tasks obtained by orchestrating the AI ​​workflow. The execution result of each AI task defined in the target AI task template can be used as the execution result that each AI task needs to achieve after orchestrating the execution result that the AI ​​workflow needs to achieve. Thus, the task orchestration result can be obtained.

[0084] Because the task orchestration module can orchestrate workflows based on AI task templates, it can improve orchestration efficiency. Furthermore, since the AI ​​task templates are provided by the system and define the AI ​​tasks and their execution results supported by the system, the results orchestrated based on these templates are supported by the system, avoiding orchestration failures caused by unsupported results.

[0085] It should be noted that the task orchestration module can also use other orchestration methods when orchestrating workflow orchestration results, such as using intelligent algorithms to orchestrate the workflow orchestration results, etc. Examples of other orchestration methods for workflow orchestration results will not be provided here.

[0086] Once the workflow orchestration module receives the workflow orchestration results, it can send them to the task control module. Similarly, once the task orchestration module receives the task orchestration results, it can send them to the task control module. Upon receiving both the workflow orchestration and task orchestration results, the business control module can execute multiple AI tasks. Based on the actual execution results of these AI tasks, the desired execution results of the AI ​​workflow, and the desired execution results for each individual AI task, the business control module adjusts the execution strategies for both the AI ​​workflow and each individual AI task to achieve collaborative management and control of various AI resource elements.

[0087] In some implementations, the business control module may include a workflow control module, a task control module, and a task execution module. The task execution module executes multiple AI tasks. The task control module adjusts the execution strategy of each AI task based on its actual execution result and the desired execution result. The workflow control module determines the actual execution result of the AI ​​workflow based on the actual execution results of multiple AI tasks and adjusts the execution strategy of the AI ​​workflow based on its actual execution result and the desired execution result. In other words, the business control module can consist of three parts: one responsible for executing AI tasks (task execution module), one responsible for adjusting the execution strategy of AI tasks (task control module), and one responsible for adjusting the execution strategy of AI workflows (workflow control module). This allows for better control of AI tasks and AI workflows, thereby achieving collaborative management and control of various AI resource elements.

[0088] In some implementations, the actual execution result of an AI workflow may include at least one of the following:

[0089] The actual working status of the AI ​​workflow (i.e., the actual execution status achieved by the AI ​​workflow), such as whether the AI ​​workflow is successfully executed or suspended, and the execution duration;

[0090] The actual QoS of an AI workflow (i.e., the QoS actually achieved by the AI ​​workflow) includes actual computing power QoS, data QoS, algorithm QoS, and connection QoS. Computing power QoS can include computing resource type, computing resource level, etc. Data QoS can include data volume level, etc. Algorithm QoS can include training real-time performance, inference real-time performance, etc. Connection QoS can include traffic level, etc.

[0091] It's important to note that the actual execution result of the AI ​​workflow must match the intended execution result to allow for comparison. For example, if the intended execution result of the AI ​​workflow includes QoS (Quality of Service), then the actual execution result includes the actual QoS of the AI ​​workflow.

[0092] In some implementations, the workflow control module is also used to receive orchestration results sent by the business orchestration module. When the business orchestration module includes both a workflow orchestration module and a task orchestration module, the workflow control module can receive workflow orchestration results from the workflow orchestration module and task orchestration results from the task orchestration module. Upon receiving the orchestration results, the workflow control module can send multiple AI tasks and the execution results required for each AI task to the task control module. After receiving the multiple AI tasks and the execution results required for each AI task, the task control module further sends the multiple AI tasks to the task execution module for execution. Where the multiple AI tasks have a temporal relationship (execution order), to ensure that the multiple AI tasks are executed sequentially, the task control module can also send the multiple AI tasks to the task execution module according to their execution order, so that the task execution module can execute the multiple AI tasks sequentially.

[0093] The task execution module, while executing multiple AI tasks, also monitors these tasks and sends the monitoring results to the task control module. For each AI task, the monitoring results can include at least one of the following: the task's working status and its performance metrics. The task's working status could include, for example, whether the task was successfully executed or its execution duration. The performance metrics could include, for example, the number of successful executions or the number of failed executions; the specific metrics can be determined based on the QoS (Quality of Service) that the AI ​​task needs to achieve, and are not specifically limited here.

[0094] Upon receiving monitoring results from multiple AI tasks sent by the task execution module, the task control module further determines the actual execution results of these AI tasks based on these results. For example, if the monitoring results include the AI ​​task's working status, the task control module can determine that working status as the actual execution result. Alternatively, if the monitoring results include AI task metrics, the task control module can determine the actual QoS achieved by the AI ​​task based on the metrics data and use that QoS as the actual execution result. After determining the actual execution results of multiple AI tasks, the task control module can adjust the execution strategy for each AI task based on these results and the required execution results for each AI task issued by the workflow orchestration module. This includes adjusting the number of task execution nodes used to execute the AI ​​task and adjusting the amount of data related to the AI ​​task.

[0095] The task control module, after determining the actual execution results of multiple AI tasks, also sends these results to the workflow control module. Upon receiving these results, the workflow control module can determine the actual execution result of the AI ​​workflow. For example, if the results include the working status of each AI task, the workflow control module can determine the working status of the AI ​​workflow and use that status as the actual execution result. Similarly, if the results include the QoS achieved by each AI task, the workflow control module can determine the QoS achieved by the AI ​​workflow and use that QoS as the actual execution result. After determining the actual execution result, the workflow control module can adjust the execution strategy of the AI ​​workflow based on these results and the desired execution result issued by the business orchestration module. This includes adjusting the number of AI tasks, their execution order, and the available resources for each task.

[0096] Figure 2 is a schematic diagram of the structure of a business control system according to another embodiment of this application. The business control system 10 shown in Figure 2 includes a business orchestration module 11 and a business control module 12. The business orchestration module 11 includes a workflow orchestration module 111 and a task orchestration module 112. The business control module 12 includes a workflow control module 121, a task control module 122, and a task execution module 123. When the business control system 10 receives an AI service request, and coordinates and manages multiple AI element resources used to provide AI services, the operations performed by each module shown in Figure 2 are as follows:

[0097] Workflow orchestration module 111 orchestrates AI service requests to obtain workflow orchestration results, which include the AI ​​workflow and the execution results that the AI ​​workflow needs to achieve. Task orchestration module 112 orchestrates the workflow orchestration results to obtain task orchestration results, which include multiple AI tasks and the execution results that each AI task needs to achieve. Workflow orchestration module 111 sends the workflow orchestration results to workflow control module 121, and task orchestration module 112 sends the task orchestration results to workflow control module 121.

[0098] After receiving the workflow orchestration results and task orchestration results, the workflow control module 121 can send the task orchestration results (multiple AI tasks and the execution results that each AI task needs to achieve) to the task control module 122. Optionally, the workflow control module 121 can create an AI workflow control instance, which maintains the execution results that the AI ​​workflow needs to achieve. Furthermore, it can also maintain an AI task list of multiple AI tasks, where the execution results that each AI task needs to achieve are maintained.

[0099] After receiving the task orchestration results, the task control module 122 can send multiple AI tasks from the orchestration results to the task execution module 123 in execution order. Optionally, the task control module 122 can create AI task control instances, which maintain the execution results that each AI task needs to achieve.

[0100] After receiving multiple AI tasks, the task execution module 123 can execute them sequentially. During the execution of these tasks, the task execution module 123 can monitor at least one of the execution status and performance metrics for each AI task and report the monitoring results to the task control module 122. Based on the monitoring results reported by the task execution module 123, the task control module 122 determines the actual execution result of each AI task. Then, for each AI task, it compares the actual execution result with the required execution result and decides whether to adjust the execution strategy based on the comparison results. For example, it may adjust the number of task execution nodes or reduce the amount of data collected and stored.

[0101] After determining the actual execution result of each AI task, the task control module 122 also needs to report the actual execution result of each AI task to the workflow control module 121. Based on the actual execution results of multiple AI tasks reported by the task control module 122, the workflow control module 121 determines the actual execution result of the AI ​​workflow. Then, it compares the actual execution result of the AI ​​workflow with the execution result that the AI ​​workflow needs to achieve, and decides whether to adjust the execution strategy of the AI ​​workflow based on the comparison result. For example, it may adjust the number of tasks to be executed, the available resources for task execution, or the order of task execution.

[0102] In some implementations, multiple AI tasks may include multiple task types, such as data acquisition, model training, model inference, model validation, and model deployment. To facilitate the execution of multiple AI tasks, the business control module may include multiple task execution modules, each capable of executing AI tasks of different types. Thus, when the task control module sends multiple AI tasks to the task execution modules, it can assign them to different execution modules based on their task types.

[0103] An intelligent intrinsic network system can include multiple service domains (such as communication service domains, sensing service domains, etc.). Different service domains can provide different service functions, and each service domain can provide AI service requests. When a service control system is applied to an intelligent intrinsic network system, in actual deployment, multiple task execution modules can be set up in each service domain. Different task execution modules can be responsible for executing different types of AI tasks. Furthermore, each service domain can also have a separate task control module, which is responsible for distributing different types of AI tasks to the corresponding task execution modules. See Figure 3.

[0104] Figure 3 uses an intelligent endogenous network system comprising three business domains (business domain 1, business domain 2, and business domain 3) as an example. Each business domain includes one task control module (task control module 1, task control module 2, and task control module 3) and multiple task execution modules (task execution modules 1a-1n, task execution modules 2a-2n, and task execution modules 3a-3n). When collaboratively managing multiple AI resource elements, the workflow control module 121 can send multiple AI tasks to multiple task control modules based on their respective business domains. Each task control module, upon receiving an AI task, can send it to a different task execution module for execution based on its task type. For each task execution module within a business domain, the execution status of the AI ​​task can be monitored during execution, and the monitoring results can be sent to the task control module within that business domain. The task control module adjusts the execution strategy of the AI ​​task based on the monitoring results and simultaneously reports the actual execution result of the AI ​​task to the workflow control module 121. Workflow control module 121 can adjust the execution strategy of AI workflow based on the actual execution results of AI tasks reported by business control module 1, business control module 2 and business control module 3.

[0105] The business control system provided in this application embodiment, upon receiving an AI service request, can orchestrate the AI ​​service request into an AI workflow, the execution result to be achieved by the AI ​​workflow, multiple AI tasks, and the execution result to be achieved by each AI task by the business orchestration module in the business control system. Then, the business control module in the business control system monitors the execution status of each AI task, using the AI ​​task as the smallest execution unit, and adjusts the execution strategy of each AI task and the execution strategy of the AI ​​workflow based on the actual execution results of multiple AI tasks, the execution result to be achieved by each AI task, and the execution result to be achieved by the AI ​​workflow. Therefore, by adjusting the strategies of AI tasks and AI workflows, the various AI element resources used to provide AI services can be adjusted. This allows for dynamic and flexible control and adaptive adjustment of various AI element resources used to provide AI services, achieving collaborative management and control of multiple AI element resources and effectively improving the quality of AI services.

[0106] This application also provides a business control method. Figure 4 is a flowchart illustrating a business control method according to an embodiment of this application. The method shown in Figure 4 includes the following steps.

[0107] S402: Orchestrate AI service requests to obtain orchestration results, which include AI workflows, the execution results that the AI ​​workflows need to achieve, multiple AI tasks, and the execution results that each AI task needs to achieve.

[0108] Upon receiving an AI service request, the system can orchestrate the request to generate an AI workflow, the desired execution result of the AI ​​workflow, multiple AI tasks, and the desired execution result for each AI task. The AI ​​service request can be sent by the user through a predefined AI service interface.

[0109] In some implementations, the AI ​​workflow may include at least one of the following:

[0110] The ID of the AI ​​workflow identifies different types of AI workflows. Different types of AI workflows consist of different AI tasks and correspond to different execution results that the AI ​​workflow needs to achieve.

[0111] The business object types of AI workflows include the business object types of AI workflows within the system;

[0112] The type of AI workflow, such as whether it is an event-based workflow or a periodic workflow;

[0113] An AI workflow includes a set of AI task types, such as data acquisition, model training, model inference, model validation, and model deployment. In some implementations, the set of AI task types in an AI workflow can be represented as {data acquisition (yes / no), model training (yes / no), model inference (yes / no), model validation (yes / no), model deployment (yes / no)}, where (yes / no) indicates whether the AI ​​task is executed in the AI ​​workflow.

[0114] The desired execution result of an AI workflow can be considered as the expected execution result that the AI ​​workflow can achieve. In some implementations, the desired execution result of an AI workflow may include at least one of the following:

[0115] The work status that an AI workflow needs to achieve can be, such as successful execution of the AI ​​workflow and execution time.

[0116] The QoS requirements for AI workflows can include computing power QoS, data QoS, algorithm QoS, and connection QoS. Among them, computing power QoS can include computing resource type and computing resource level, data QoS can include data volume level, algorithm QoS can include training real-time performance and inference real-time performance, and connection QoS can include traffic level. It should be noted that the definitions of these QoS requirements should use the same type and level definitions as the descriptions of computing power, data, algorithms, and connections in the system.

[0117] Each AI task may include at least one of the following:

[0118] The ID of an AI task can identify different types of AI tasks, and different types of AI tasks can correspond to different resource types and the execution results that the AI ​​task needs to achieve.

[0119] The business object of the AI ​​task includes the business object type and primary key of the AI ​​task in the system;

[0120] AI tasks can include data collection, model training, model inference, model validation, and model deployment.

[0121] The desired execution result of an AI task can be the expected execution result that the AI ​​task can achieve. In some implementations, the desired execution result for each AI task may include at least one of the following:

[0122] The execution status that an AI task needs to achieve, such as successful execution of the AI ​​task and execution time;

[0123] The QoS that AI tasks need to achieve can include some of the four QoS categories: computing power QoS, data QoS, algorithm QoS, and connection QoS. Among them, computing power QoS can include computing power resource type and computing power resource level, data QoS can include data volume level, algorithm QoS can include training real-time performance and inference real-time performance, and connection QoS can include traffic level. It should be noted that the definitions of these QoS categories should use the same type and level definitions as the descriptions of computing power, data, algorithms, and connections in the system.

[0124] Since the QoS requirements for AI tasks include a subset of QoS from computing power, data, algorithm, and connectivity, the QoS requirements of the AI ​​workflow are broken down into smaller parts. Each AI task achieves only a portion of these QoS requirements. Therefore, after obtaining the actual execution results of multiple AI tasks, adjustments to the execution strategy of a single AI task can control some AI resource elements, and adjustments to the execution strategies of multiple AI tasks can achieve coordinated control of various AI resource elements. Optionally, in some implementations, the QoS requirements for each AI task may include one QoS from the QoS requirements of the AI ​​workflow. Multiple AI tasks correspond to different QoS levels. This allows the QoS requirements of the AI ​​workflow to be broken down into individual QoS levels. Thus, adjustments to the execution strategy of a single AI task can control one type of AI resource element, and adjustments to the execution strategies of multiple AI tasks can achieve coordinated control of various AI resource elements.

[0125] In some implementations, the actual result of each AI task may include at least one of the following:

[0126] The actual execution status of an AI task (i.e., the actual execution status achieved by the AI ​​task), such as whether the AI ​​task was successfully executed or suspended, and the actual execution time, etc.

[0127] The actual QoS of an AI task (i.e., the QoS actually achieved by the AI ​​task) includes the actual computing power QoS, data QoS, algorithm QoS, and some QoS in connection QoS. Computing power QoS can include computing power resource type, computing power resource level, etc. Data QoS can include data volume level, etc. Algorithm QoS can include training real-time performance, inference real-time performance, etc. Connection QoS can include traffic level, etc. It should be noted that the definitions of these QoS should use the same type and level definitions as the descriptions of computing power, data, algorithms, and connections in the system.

[0128] It should be noted that for each AI task, the actual execution result of the AI ​​task must be consistent with the execution result that the AI ​​task needs to achieve, so that the actual execution result of the AI ​​task can be compared with the required execution result. For example, if the execution result that the AI ​​task needs to achieve includes the computing power QoS that the AI ​​task needs to achieve, then the actual execution result of the AI ​​task includes the actual computing power QoS of the AI ​​task.

[0129] In some implementations, orchestrating AI service requests to obtain orchestration results may include the following steps:

[0130] The AI ​​service requests are orchestrated to obtain the workflow orchestration results, which include the AI ​​workflow and the execution results that the AI ​​workflow needs to achieve.

[0131] The workflow orchestration results are then arranged to obtain the task orchestration results, which include multiple AI tasks and the execution results that each AI task needs to achieve.

[0132] In other words, when orchestrating AI service requests, one can first orchestrate the AI ​​workflow and its required execution results based on the AI ​​service requests. Then, one can orchestrate the AI ​​workflow and its required execution results to obtain multiple AI tasks and their required execution results. This progressive orchestration approach improves orchestration efficiency.

[0133] In some implementations, the AI ​​service request may include business information and service level agreement (SLA) information for the AI ​​service, which can be input by the user through a predefined AI service interface provided by the system. The business information for the AI ​​service may include at least one of the following:

[0134] The business objects of AI services can include the types and instance information of business objects that perform AI business in the system;

[0135] AI services can include at least one of the following business types: data collection, model training, model inference, model validation, and model deployment.

[0136] The service level agreement information for AI services may include at least one of the following:

[0137] The service priority of AI services can be represented by a numerical value. The larger the value, the higher the service priority of the AI ​​service. Under resource constraints, AI services with higher service priority can preempt the resources of AI services with lower service priority.

[0138] The algorithm performance level of an AI service can be represented by a numerical value. The larger the value, the higher the algorithm performance requirement of the AI ​​service and the greater the corresponding resource demand.

[0139] The service latency level of AI services can be represented by a numerical value. The higher the value, the higher the real-time requirement of the AI ​​service.

[0140] In this way, orchestrating AI service requests to obtain workflow orchestration results can include the following steps:

[0141] Based on the business information and service level agreement information of the AI ​​service, a target AI workflow template is determined from multiple AI workflow templates. Each AI workflow template is used to define an AI workflow and the execution result of the AI ​​workflow. The AI ​​workflow defined in the target AI workflow template matches the business information of the AI ​​service, and the execution result of the AI ​​workflow defined in the target AI workflow template matches the service level agreement information of the AI ​​service.

[0142] The AI ​​workflows defined in the target AI workflow template and the execution results of the AI ​​workflows are identified as the workflow orchestration results.

[0143] Multiple AI workflow templates can be predefined templates in the system. Optionally, the system can maintain an AI workflow template library, which can store multiple AI workflow templates. Each AI workflow template can define an AI workflow that includes at least one of the following:

[0144] The AI ​​workflow ID identifies different types of AI workflows, which consist of different AI tasks and correspond to different execution results.

[0145] The business object types of AI workflows include the business object types of AI workflows within the system;

[0146] The type of AI workflow, such as whether it is an event-based workflow or a periodic workflow;

[0147] An AI workflow includes a set of AI task types, such as data acquisition, model training, model inference, model validation, and model deployment. In some implementations, the set of AI task types in an AI workflow can be represented as {data acquisition (yes / no), model training (yes / no), model inference (yes / no), model validation (yes / no), model deployment (yes / no)}, where (yes / no) indicates whether the AI ​​task is executed in the AI ​​workflow.

[0148] The execution result of the AI ​​workflow defined in each AI workflow template may include at least one of the following:

[0149] The working status of the AI ​​workflow can be, for example, the successful execution of the AI ​​workflow and the execution time.

[0150] QoS in AI workflows can include computing power QoS, data QoS, algorithm QoS, and connection QoS. Among them, computing power QoS can include computing resource type and computing resource level, data QoS can include data volume level, algorithm QoS can include training real-time performance and inference real-time performance, and connection QoS can include traffic level.

[0151] When orchestrating AI service requests, the business information of the AI ​​service can be matched with the AI ​​workflows defined in each AI workflow template to identify one or more matching AI workflow templates. Then, the execution results of the AI ​​workflows defined in these templates are matched with the service level agreement (SLA) information of the AI ​​service to determine the matching AI workflow template. This matching AI workflow template is the target AI workflow template. The AI ​​workflows defined in the target AI workflow template match the business information of the AI ​​service, and the execution results of the AI ​​workflows defined in the target AI workflow template match the SLA information of the AI ​​service. Matching the business information of the AI ​​workflows defined in the target AI workflow template can mean that the information of the AI ​​workflows defined in the target AI workflow template includes the business information of the AI ​​service. Matching the execution results of the AI ​​workflows defined in the target AI workflow template with the SLA information can mean that the execution results of the AI ​​workflows defined in the target AI workflow template meet (equal to or exceed) the requirements of the SLA information of the AI ​​service.

[0152] It should be noted that, in one possible implementation, there may be one or more target AI workflow templates. If there are multiple target AI workflow templates, the template that is closest to the business information and service level agreement information of the AI ​​service can be determined from the multiple target AI workflow templates as the final template, and the AI ​​service request can be orchestrated according to the template.

[0153] After determining the target AI workflow template, the AI ​​workflow defined in the target AI workflow template can be used as the AI ​​workflow obtained after orchestrating the AI ​​service requests. The execution result of the AI ​​workflow defined in the target AI workflow template can be used as the execution result that the AI ​​workflow obtained after orchestrating the AI ​​service requests needs to achieve. Thus, the workflow orchestration result can be obtained. Since the AI ​​workflow template is provided by the system and defines the AI ​​workflows and execution results supported by the system, the result obtained by orchestrating according to the AI ​​workflow template can be supported by the system, avoiding orchestration failures caused by situations where the orchestration result is not supported by the system.

[0154] In some implementations, to facilitate the subsequent execution of the AI ​​workflow, after determining the target AI workflow template, the following steps may also be included:

[0155] Determine whether the remaining system resources meet the resource requirements of the target AI workflow template;

[0156] If the remaining system resources meet the resource requirements of the target AI workflow template, deduct the resources required for the target AI workflow template from the remaining system resources.

[0157] After deducting system resources, the AI ​​workflow orchestration can be considered successful, and the orchestration process can then be terminated. Optionally, in some implementations, if the remaining system resources do not meet the resource requirements of the target AI workflow template, the AI ​​workflow orchestration can be considered a failure, and a failure message can be returned for the AI ​​service request. This embodiment uses the example of the remaining system resources meeting the resource requirements of the target AI workflow template for illustration.

[0158] After obtaining the workflow orchestration results, the workflow orchestration results can be further arranged to obtain task orchestration results. In some implementations, arranging the workflow orchestration results to obtain task orchestration results may include the following steps:

[0159] Based on the workflow orchestration results, a target AI task template is determined from multiple AI task templates. Each AI task template is used to define multiple AI tasks and the execution result of each AI task. The multiple AI tasks defined in the target AI task template are matched with the AI ​​workflows included in the workflow orchestration results. The execution result of each AI task defined in the target AI task template is matched with the execution result that the AI ​​workflows included in the workflow orchestration results need to achieve.

[0160] The task orchestration result is determined by defining multiple AI tasks in the target AI task template and the execution result of each AI task.

[0161] Multiple AI task templates can be predefined templates in the system. Optionally, the system can maintain a template library for AI tasks, which can store multiple AI task templates. For each AI task template, the multiple AI tasks defined in the template can include at least one of the following:

[0162] Multiple AI task IDs; different IDs can identify different types of AI tasks, and different types of AI tasks can correspond to different resource types and AI task execution results.

[0163] The business objects for multiple AI tasks include the business object type and primary key of each AI task in the system;

[0164] There are multiple types of AI tasks, and each type of AI task can include data collection, model training, model inference, model validation, model deployment, etc.

[0165] For each AI task template, the execution result of each AI task defined in the template may include at least one of the following:

[0166] The execution status of an AI task, such as whether the AI ​​task was successfully executed or the execution time;

[0167] QoS for AI tasks can include some of the following four aspects: computing power QoS, data QoS, algorithm QoS, and connection QoS. Among them, computing power QoS can include computing resource type and computing resource level, data QoS can include data volume level, algorithm QoS can include training real-time performance and inference real-time performance, and connection QoS can include traffic level.

[0168] When orchestrating AI service requests, the AI ​​workflow in the workflow orchestration result can be matched with multiple AI tasks defined in each AI task template to determine one or more matching AI task templates. Then, the execution results of each AI task defined in these templates can be matched with the execution results of the AI ​​workflow in the workflow orchestration result to determine the matching AI task template. This matching AI task template is the target AI task template. The multiple AI tasks defined in the target AI task template are matched with the AI ​​workflow in the workflow orchestration result, and the execution results of each AI task defined in the target AI task template are matched with the execution results of the AI ​​workflow in the workflow orchestration result. In this context, multiple AI tasks defined in the target AI task template are matched with AI workflows in the workflow orchestration results. This can be an AI workflow composed of multiple AI tasks defined in the target AI task template, including AI workflows in the workflow orchestration results. The execution result of each AI task defined in the target AI task template is matched with the execution result of the AI ​​workflow in the workflow orchestration results. This can be an AI workflow whose execution result, determined based on the execution result of each AI task defined in the target AI task template, meets (is equal to or exceeds) the requirement of the execution result of the AI ​​workflow in the workflow orchestration results.

[0169] Optionally, in some implementations, corresponding AI task templates can be maintained for different AI workflow templates. Multiple AI tasks defined in an AI task template are matched with AI workflows defined in the AI ​​workflow template, and the execution result of each AI task defined in the AI ​​task template is matched with the execution result of the AI ​​workflow defined in the AI ​​workflow template. In this way, when the task orchestration module performs task orchestration, it can directly determine the AI ​​task template corresponding to the target AI workflow template from multiple AI task templates and use that AI task template as the target AI task template for AI task orchestration. This eliminates the need to sequentially match the workflow orchestration result with multiple AI task templates, thereby shortening the AI ​​task orchestration time and improving the efficiency of AI task orchestration.

[0170] After determining the target AI task template, the multiple AI tasks defined in the target AI task template can be used as multiple AI tasks obtained by orchestrating the AI ​​workflow. The execution result of each AI task defined in the target AI task template can be used as the execution result that each AI task needs to achieve after orchestrating the execution result that the AI ​​workflow needs to achieve. Thus, the task orchestration result can be obtained.

[0171] Since AI service requests can be orchestrated based on AI workflow templates and AI task templates, orchestration efficiency can be improved. Furthermore, because both AI workflow templates and AI task templates are provided by the system and define the system-supported AI workflows, their execution results, AI tasks, and their execution results, the results orchestrated based on these templates are supported by the system, avoiding orchestration failures caused by unsupported results.

[0172] It should be noted that in other possible implementations, other orchestration methods can also be used when orchestrating AI service requests. For example, intelligent algorithms can be used to orchestrate AI service requests to obtain workflow orchestration results, and then intelligent algorithms can be used to orchestrate workflow orchestration results to obtain task orchestration results, etc. Examples of other orchestration methods will not be provided here.

[0173] S404: Execute multiple AI tasks. Based on the actual execution results of the multiple AI tasks, the execution results that the AI ​​workflow needs to achieve, and the execution results that each AI task needs to achieve, adjust the execution strategy of the AI ​​workflow and the execution strategy of each AI task.

[0174] During the execution of multiple AI tasks, the execution status of each AI task can be monitored to obtain monitoring results for multiple AI tasks. Based on these results, the actual execution outcome of the multiple AI tasks can be determined. For each AI task, the monitoring results can include at least one of the following: the task's working status and its performance metrics. The task's working status could include, for example, whether the task was successfully executed or its execution duration. The performance metrics could include, for example, the number of successful executions or the number of failed executions, which can be determined based on the QoS (Quality of Service) required by the AI ​​task; no specific limitations are imposed here. When determining the actual execution outcome of multiple AI tasks based on their monitoring results, for example, if the monitoring results include the task's working status, that working status can be considered the actual execution outcome. Alternatively, if the monitoring results include the task's performance metrics, the actual QoS achieved by the AI ​​task can be determined based on the performance metrics, and this achieved QoS can be considered the actual execution outcome of the AI ​​task.

[0175] After obtaining the actual execution results of multiple AI tasks, the execution strategies of the AI ​​workflow and each AI task can be adjusted based on the actual execution results of the multiple AI tasks, the execution results that the AI ​​workflow arranged in S402 needs to achieve, and the execution results that each AI task needs to achieve.

[0176] In some implementations, adjusting the execution strategy of the AI ​​workflow and the execution strategy of each AI task based on the actual execution results of multiple AI tasks, the execution results that the AI ​​workflow needs to achieve, and the execution results that each AI task needs to achieve may include the following steps:

[0177] The actual execution results of the AI ​​workflow are determined based on the actual execution results of multiple AI tasks.

[0178] Adjust the execution strategy of the AI ​​workflow based on the actual execution results and the execution results that the AI ​​workflow needs to achieve.

[0179] For each AI task, the execution strategy is adjusted based on the actual execution result and the desired execution result.

[0180] Multiple AI tasks are generated by orchestrating an AI workflow. Given the actual execution results of these multiple AI tasks, the actual execution result of the AI ​​workflow can be obtained through computational analysis. In some implementations, the actual execution result of the AI ​​workflow may include at least one of the following:

[0181] The actual working status of the AI ​​workflow (i.e., the actual execution status achieved by the AI ​​workflow), such as whether the AI ​​workflow is successfully executed or suspended, and the execution duration;

[0182] The actual QoS of an AI workflow (i.e., the QoS actually achieved by the AI ​​workflow) includes actual computing power QoS, data QoS, algorithm QoS, and connection QoS. Computing power QoS can include computing resource type, computing resource level, etc. Data QoS can include data volume level, etc. Algorithm QoS can include training real-time performance, inference real-time performance, etc. Connection QoS can include traffic level, etc.

[0183] It's important to note that the actual execution result of the AI ​​workflow must match the intended execution result to allow for comparison. For example, if the intended execution result of the AI ​​workflow includes QoS (Quality of Service), then the actual execution result includes the actual QoS of the AI ​​workflow.

[0184] After determining the actual execution result of the AI ​​workflow, the actual result can be compared with the desired result, and the execution strategy can be adjusted based on the comparison. In some implementations, adjusting the execution strategy may include, but is not limited to, at least one of the following:

[0185] Adjust the number of AI tasks executed, such as increasing or decreasing the number of AI tasks executed;

[0186] Adjust the execution order of multiple AI tasks, such as advancing or delaying the execution order of AI tasks;

[0187] Adjust the available resources for multiple AI tasks, such as increasing or decreasing the available resources for AI tasks.

[0188] For each AI task, given the actual execution result, the actual result can be compared with the desired result. The execution strategy can then be adjusted based on this comparison. In some implementations, adjusting the execution strategy for each AI task may include, but is not limited to, at least one of the following:

[0189] Adjust the number of task execution nodes used to perform AI tasks, such as increasing or decreasing the number of task execution nodes;

[0190] Adjust the amount of data related to the AI ​​task, such as increasing or decreasing the amount of data collected or stored for the AI ​​task.

[0191] The business control method provided in this application, upon receiving an AI service request, can orchestrate the AI ​​service request into an AI workflow, the execution result to be achieved by the AI ​​workflow, multiple AI tasks, and the execution result to be achieved by each AI task. Then, taking the AI ​​task as the smallest execution unit, the execution status of each AI task is monitored, and the execution strategies of each AI task and the AI ​​workflow are adjusted based on the actual execution results of multiple AI tasks, the execution result to be achieved by each AI task, and the execution result to be achieved by the AI ​​workflow. Therefore, by adjusting the strategies of AI tasks and AI workflows, the various AI element resources used to provide AI services can be adjusted. This allows for dynamic and flexible control and adaptive adjustment of various AI element resources used to provide AI services, achieving collaborative management and control of multiple AI element resources and effectively improving the quality of AI services.

[0192] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0193] Figure 5 is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Referring to Figure 5, at the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. The memory may include RAM, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk storage device. Of course, the electronic device may also include other hardware required for other services.

[0194] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. This bus can be categorized as an address bus, data bus, control bus, etc. For ease of illustration, only a single bidirectional arrow is used in Figure 5, but this does not imply that there is only one bus or one type of bus.

[0195] Memory is used to store programs. Specifically, programs may include program code, which includes computer operation instructions. Memory may include main memory and non-volatile memory, and provides instructions and data to the processor.

[0196] The processor reads the corresponding computer program from non-volatile memory into main memory and then executes it, forming a business control unit at the logical level. The processor executes the program stored in memory and specifically performs the following operations:

[0197] The AI ​​service requests are orchestrated to obtain an orchestration result, which includes an AI workflow, the execution result that the AI ​​workflow needs to achieve, multiple AI tasks, and the execution result that each AI task needs to achieve.

[0198] Execute the multiple AI tasks, and adjust the execution strategy of the AI ​​workflow and the execution strategy of each AI task based on the actual execution results of the multiple AI tasks, the execution results that the AI ​​workflow needs to achieve, and the execution results that each AI task needs to achieve.

[0199] The method executed by the service control device disclosed in the embodiment shown in Figure 5 of this application can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0200] The electronic device can also execute the method of FIG4 and implement the functions of the business control device in the embodiment shown in FIG4, which will not be described in detail here.

[0201] Of course, in addition to software implementation, the electronic device of this application does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. In other words, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0202] This application also proposes a computer-readable storage medium that stores one or more programs, the programs including instructions that, when executed by a portable electronic device including multiple applications, enable the portable electronic device to perform the method of the embodiment shown in FIG4, and specifically to perform the following operations:

[0203] The AI ​​service requests are orchestrated to obtain an orchestration result, which includes an AI workflow, the execution result that the AI ​​workflow needs to achieve, multiple AI tasks, and the execution result that each AI task needs to achieve.

[0204] Execute the multiple AI tasks, and adjust the execution strategy of the AI ​​workflow and the execution strategy of each AI task based on the actual execution results of the multiple AI tasks, the execution results that the AI ​​workflow needs to achieve, and the execution results that each AI task needs to achieve.

[0205] This application also proposes a computer program product comprising a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps described in the above-described business control method embodiments.

[0206] In summary, the above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

[0207] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0208] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0209] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0210] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

Claims

1. A business control system, comprising a business orchestration module and a business control module, wherein: The business orchestration module is used to orchestrate artificial intelligence (AI) service requests and send the orchestration results to the business control module. The orchestration results include AI workflow, the execution result that the AI ​​workflow needs to achieve, multiple AI tasks, and the execution result that each AI task needs to achieve. The business control module is used to execute the multiple AI tasks, and adjust the execution strategy of the AI ​​workflow and the execution strategy of each AI task based on the actual execution results of the multiple AI tasks, the execution results that the AI ​​workflow needs to achieve, and the execution results that each AI task needs to achieve.

2. The system as described in claim 1, wherein the business orchestration module includes a workflow orchestration module and a task orchestration module; The workflow orchestration module is used to orchestrate the AI ​​service requests to obtain workflow orchestration results, which include the AI ​​workflow and the execution results that the AI ​​workflow needs to achieve. The task orchestration module is used to orchestrate the workflow orchestration results to obtain task orchestration results, which include the multiple AI tasks and the execution results that each AI task needs to achieve.

3. The system as described in claim 2, wherein the AI ​​service request includes business information and service level agreement information of the AI ​​service, and the workflow orchestration module is used to determine a target AI workflow template from multiple AI workflow templates based on the business information and the service level agreement information, and to determine the AI ​​workflow defined in the target AI workflow template and the execution result of the AI ​​workflow as the workflow orchestration result; in, Each AI workflow template defines an AI workflow and its execution result. The AI ​​workflow defined in the target AI workflow template matches the business information, and the execution result of the AI ​​workflow defined in the target AI workflow template matches the service level agreement information.

4. The system as described in claim 3, wherein the business information includes at least one of the following: the business object of the AI ​​service, and the business type of the AI ​​service; The service level agreement information includes at least one of the following: the service priority of the AI ​​service, the algorithm performance level of the AI ​​service, and the service latency level of the AI ​​service.

5. The system as described in claim 3, wherein the workflow orchestration module is further configured to, when determining the target AI workflow template, determine whether the remaining system resources meet the resource requirements of the target AI workflow template, and, if the remaining system resources meet the resource requirements of the target AI workflow template, deduct the resources required by the target AI workflow template from the remaining system resources.

6. The system as described in claim 2, wherein the task orchestration module is configured to determine a target AI task template from multiple AI task templates based on the workflow orchestration result, and to determine the multiple AI tasks defined in the target AI task template and the execution result of each AI task as the task orchestration result; in, Each AI task template is used to define multiple AI tasks and the execution result of each AI task. The multiple AI tasks defined in the target AI task template are matched with the AI ​​workflows included in the workflow orchestration result. The execution result of each AI task defined in the target AI task template is matched with the execution result that the AI ​​workflow included in the workflow orchestration result needs to achieve.

7. The system as described in claim 1, wherein the business control module includes a workflow control module, a task control module, and a task execution module; The task execution module is used to execute the multiple AI tasks; For each AI task, the task control module is used to adjust the execution strategy of the AI ​​task based on the actual execution result of the AI ​​task and the execution result that the AI ​​task needs to achieve. The workflow control module is used to determine the actual execution result of the AI ​​workflow based on the actual execution results of the multiple AI tasks, and to adjust the execution strategy of the AI ​​workflow based on the actual execution result of the AI ​​workflow and the execution result that the AI ​​workflow needs to achieve.

8. The system as described in claim 7, wherein the workflow control module is further configured to receive the orchestration result sent by the business orchestration module, send the plurality of AI tasks included in the orchestration result and the execution result to be achieved by each AI task to the task control module, and receive the actual execution result of the plurality of AI tasks sent by the task control module; The task control module is also used to send the multiple AI tasks to the task execution module according to the execution order of the multiple AI tasks, receive the monitoring results of the multiple AI tasks sent by the task execution module, determine the actual execution results of the multiple AI tasks according to the monitoring results, and send the actual execution results of the multiple AI tasks to the workflow control module. The task execution module is also used to monitor the multiple AI tasks and send the monitoring results to the task control module. For each AI task, the monitoring results include at least one of the working status and indicator data of the AI ​​task.

9. The system of claim 8, wherein the plurality of AI tasks includes a plurality of task types; The number of task execution modules is multiple, and different task execution modules are used to execute AI tasks of different task types. The task control module is also used to send the multiple AI tasks to different task execution modules according to the task types of the multiple AI tasks.

10. The system as described in any one of claims 1 to 9, The AI ​​workflow includes at least one of the following: the ID of the AI ​​workflow, the business object type of the AI ​​workflow, the type of the AI ​​workflow, and the set of AI task types included in the AI ​​workflow; The execution result that the AI ​​workflow needs to achieve includes at least one of the following: the working state that the AI ​​workflow needs to achieve, and the Quality of Service (QoS) that the AI ​​workflow needs to achieve; wherein, The QoS requirements for the AI ​​workflow include computing power QoS, data QoS, algorithm QoS, and connection QoS. The actual execution result of the AI ​​workflow includes at least one of the following: the actual working state of the AI ​​workflow and the actual QoS of the AI ​​workflow; wherein, the actual QoS of the AI ​​workflow includes computing power QoS, data QoS, algorithm QoS and connection QoS.

11. The system as described in any one of claims 1 to 9, The AI ​​task includes at least one of the following: the AI ​​task ID, the business object of the AI ​​task, and the type of the AI ​​task; The execution result that the AI ​​task needs to achieve includes at least one of the following: the execution state that the AI ​​task needs to achieve, and the QoS that the AI ​​task needs to achieve; wherein, The QoS requirements for the AI ​​task include some of the following: computing power QoS, data QoS, algorithm QoS, and connection QoS. The actual execution result of the AI ​​task includes at least one of the following: the actual execution status of the AI ​​task, and the actual QoS of the AI ​​task; wherein, the actual QoS of the AI ​​task includes a portion of the QoS of computing power, data, algorithm, and connection.

12. A business control method, comprising: The AI ​​service requests are orchestrated to obtain an orchestration result, which includes an AI workflow, the execution result that the AI ​​workflow needs to achieve, multiple AI tasks, and the execution result that each AI task needs to achieve. Execute the multiple AI tasks, and adjust the execution strategy of the AI ​​workflow and the execution strategy of each AI task based on the actual execution results of the multiple AI tasks, the execution results that the AI ​​workflow needs to achieve, and the execution results that each AI task needs to achieve.

13. The method of claim 12, wherein orchestrating the AI ​​service requests to obtain an orchestration result includes: The AI ​​service requests are orchestrated to obtain workflow orchestration results, which include the AI ​​workflow and the execution results that the AI ​​workflow needs to achieve. The workflow orchestration results are arranged to obtain task orchestration results, which include the multiple AI tasks and the execution results that each AI task needs to achieve.

14. The method of claim 13, wherein the AI ​​service request includes business information and service level agreement information of the AI ​​service; and the orchestration of the AI ​​service request to obtain a workflow orchestration result includes: Based on the business information and the service level agreement information, a target AI workflow template is determined from multiple AI workflow templates. Each AI workflow template is used to define an AI workflow and the execution result of the AI ​​workflow. The AI ​​workflow defined in the target AI workflow template matches the business information, and the execution result of the AI ​​workflow defined in the target AI workflow template matches the service level agreement information. The AI ​​workflow defined in the target AI workflow template and the execution result of the AI ​​workflow are determined as the workflow orchestration result.

15. The method of claim 14, further comprising, after determining the target AI workflow template: Determine whether the remaining system resources meet the resource requirements of the target AI workflow template; If the remaining system resources meet the resource requirements of the target AI workflow template, the resources required for the target AI workflow template are deducted from the remaining system resources.

16. The method of claim 13, wherein arranging the workflow orchestration result to obtain the task orchestration result includes: Based on the workflow orchestration results, a target AI task template is determined from multiple AI task templates. Each AI task template defines multiple AI tasks and the execution result of each AI task. The multiple AI tasks defined in the target AI task template are matched with the AI ​​workflows included in the workflow orchestration results. The execution result of each AI task defined in the target AI task template is matched with the execution result that the AI ​​workflows included in the workflow orchestration results need to achieve. The task orchestration result is determined by defining multiple AI tasks in the target AI task template and the execution result of each AI task.

17. The method of claim 12, wherein adjusting the execution strategy of the AI ​​workflow and the execution strategy of each AI task based on the actual execution results of the plurality of AI tasks, the execution results that the AI ​​workflow needs to achieve, and the execution results that each AI task needs to achieve, comprises: The actual execution result of the AI ​​workflow is determined based on the actual execution results of the multiple AI tasks; The execution strategy of the AI ​​workflow is adjusted based on the actual execution results and the execution results that the AI ​​workflow needs to achieve. For each AI task, the execution strategy of the AI ​​task is adjusted based on the actual execution result of the AI ​​task and the execution result that the AI ​​task needs to achieve.

18. The method of claim 12 or 17, wherein adjusting the execution strategy of the AI ​​workflow includes at least one of the following: Adjust the number of AI tasks to be executed; Adjust the execution order of the multiple AI tasks; Adjust the available resources for the multiple AI tasks; For each AI task, the execution strategy of the AI ​​task is adjusted, including at least one of the following: Adjust the number of task execution nodes used to perform the AI ​​task; Adjust the amount of data related to the AI ​​task.

19. An electronic device comprising: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the method as described in any one of claims 12 to 18.

20. A computer-readable storage medium, wherein instructions in the storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the method as described in any one of claims 12 to 18.

21. A computer program product comprising a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps of the method as claimed in any one of claims 12 to 18.

Images