Network computing power scheduling methods, devices, equipment, media and program products
By classifying computing resources according to execution time and using methods such as intersection calculation and overall mode, the problem of inconsistent completion times of sub-computation tasks in industrial parks was solved, achieving equal-delay scheduling and improving the continuity and efficiency of production lines.
Patent Information
- Application Number
- CN202611458198.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-18
- Publication Date
- 2026-11-17
AI Technical Summary
Existing network computing scheduling methods fail to effectively incorporate the dynamic changes in production intervals within industrial parks, resulting in inconsistent completion times for sub-computation tasks after the computational task is split, which affects the continuity and efficiency of production lines.
By classifying computing network resources according to the execution time of completing sub-computation tasks, and grouping resources with the same execution time into the same level, resource allocation is determined using methods such as intersection calculation and overall mode, ensuring that each sub-computation task is completed at a consistent execution time.
It achieves equal-latency scheduling of computing tasks, reduces the occurrence of computing resources being preempted, and improves the continuity and efficiency of production line production.
Smart Images

Figure CN123123345A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer communication technology, and in particular to a network computing power scheduling method, apparatus, device, medium and program product. Background Technology
[0002] With the development of the Industrial Internet, the number of computing tasks in industrial parks is constantly increasing. Network computing scheduling methods have been applied to the management and control of computing resources in industrial parks to balance the load of computing nodes and improve the utilization rate of computing resources. Existing network computing scheduling methods usually use the load prediction data of computing nodes, task priority, node health and other factors as the basis for allocating computing resources, or use static latency, energy consumption and other factors as the triggering conditions for computing task migration.
[0003] However, the aforementioned network computing power scheduling method does not take into account the dynamic changes in production intervals in industrial parks, and the allocation criteria do not reflect the time required for network resources to complete tasks. In industrial parks, the production interval of task requesters directly affects production line scheduling and production efficiency. For computing tasks that are broken down into multiple sub-computing tasks, if the execution time of the network resources allocated to each sub-computing task is different, the completion time of each sub-computing task will be inconsistent. Task requesters will need to wait for the slowest sub-computing task to be completed, and the computing resources occupied by completed sub-computing tasks will be idle and easily preempted by other tasks, thereby reducing the continuity of production line operations. Summary of the Invention
[0004] This application provides a network computing power scheduling method, apparatus, equipment, medium, and program product, which are used to reduce the preemption of computing power resources and improve the continuity and efficiency of production line production.
[0005] In a first aspect, embodiments of this application provide a network computing power scheduling method, which includes: a computing network management center dividing a computing task into N sub-computing tasks, where N is an integer greater than 1; and based on the execution time of each computing network resource in multiple computing network resources corresponding to a target industrial park to complete the target sub-computing task, dividing computing network resources with the same execution time into computing network resources of the same level, obtaining M-level computing network resources, where the target sub-computing task is one of the N sub-computing tasks, and M is an integer greater than 1; and obtaining the execution time set corresponding to each sub-computing task, the execution time set corresponding to each sub-computing task including the execution time of each level of computing network resources to complete each sub-computing task; and performing intersection calculation on the execution time sets corresponding to the N sub-computing tasks, and if there is an intersection element, determining the computing network resource level corresponding to the intersection element in the execution time set corresponding to each sub-computing task, and allocating the computing network resources of the computing network resource level to the corresponding sub-computing task.
[0006] The technical solution provided in this application brings at least the following beneficial effects: Since computing network resources are classified based on the execution time of completing sub-computation tasks, and computing network resources with the same execution time are classified into the same level, multiple computing network resources have a unified comparison benchmark for the time required to complete sub-computation tasks. Therefore, the execution time of each sub-computation task under each level of computing network resource can be recorded in the corresponding execution time set and compared with each other. Furthermore, the intersection element obtained by performing an intersection calculation on the execution time sets corresponding to each sub-computation task represents the execution time that each sub-computation task can achieve. Thus, based on the intersection element... Each sub-computing task is assigned a corresponding computing network resource level and allocated computing network resources. The execution time of the computing network resources allocated to each sub-computing task is consistent with each other. As a result, N sub-computing tasks can be completed at the same execution time. The task requester does not need to wait for the slowest sub-computing task to be completed. Therefore, it is possible to achieve equal-latency scheduling of computing tasks, so that computing power scheduling can be combined with the dynamic changes in the production interval of the industrial park. Instead of allocating based on criteria that do not reflect execution time, as in the existing technology, the completion times of the sub-computing tasks after the computing task is split are inconsistent. This reduces the situation of computing power resources being preempted and improves the continuity and efficiency of production line production.
[0007] One possible implementation is that, after calculating the intersection of the execution time sets corresponding to the N sub-computing tasks, the method further includes: if there is no intersection element, the computing network management center merges the execution time sets corresponding to the N sub-computing tasks to obtain a merged execution time set; and determines the overall mode of the merged execution time set, and allocates resources to the N sub-computing tasks based on the overall mode.
[0008] Another possible implementation, the above-mentioned resource allocation for N sub-computation tasks based on the overall mode, includes: if the execution time set corresponding to the first sub-computation task contains the overall mode, then the computing network management center determines the computing network resource level corresponding to the overall mode in the execution time set as the computing network resource level corresponding to the first sub-computation task; if the execution time set corresponding to the first sub-computation task does not contain the overall mode, then the computing network management center determines the computing network resource level corresponding to the element in the execution time set closest to the overall mode as the computing network resource level corresponding to the first sub-computation task; and allocates the computing network resources corresponding to the computing network resource level of each sub-computation task to the corresponding sub-computation task; wherein, the first sub-computation task is one of the N sub-computation tasks.
[0009] Another possible implementation involves allocating network resources at the network resource level to corresponding sub-computing tasks, including: When the network resource level corresponding to the intersection element of the execution time sets of at least two sub-computing tasks in the N sub-computing tasks is the same, the network management center modifies the intersection element based on a preset disturbance time to obtain a modified execution time set; and determines multiple candidate allocation schemes based on the modified execution time set, where the candidate allocation schemes indicate the network resources allocated to each of the at least two sub-computing tasks; and determines the candidate allocation scheme with the smallest variance of the execution times of the at least two sub-computing tasks as the target allocation scheme; and allocates resources to the at least two sub-computing tasks based on the target allocation scheme, where different sub-computing tasks within the at least two sub-computing tasks are allocated different network resources.
[0010] Another possible implementation is that the hierarchical interval between adjacent levels in the aforementioned M-level computing network resources is determined based on at least one of the following: the minimum order of magnitude of the execution time interval between the aforementioned multiple computing network resources; the production interval time of the task requester; wherein, the production interval time is the waiting time of the task requester before executing the next production instruction.
[0011] Another possible implementation is that the above-mentioned allocation of computing network resources at the computing network resource level to the corresponding sub-computing tasks includes: the computing network management center selecting computing network resources from the computing network resources at the computing network resource level that have not been allocated to other sub-computing tasks among the N sub-computing tasks, and using them as computing network resources to allocate to the corresponding sub-computing tasks.
[0012] Another possible implementation, before classifying computing network resources with the same execution time into computing network resources of the same level, the above method further includes: the computing network management center obtaining the communication time between the first computing network resource and the computing network management center, and the calculation time for the first computing network resource to complete the second sub-computation task; and determining the sum of the communication time and calculation time corresponding to the first computing network resource as the execution time for the first computing network resource to complete the second sub-computation task; wherein, the first computing network resource is one of multiple computing network resources; and the second sub-computation task is one of N sub-computation tasks.
[0013] Secondly, embodiments of this application provide a network computing power scheduling device, which includes a splitting module, an acquisition module, a determination module, and an allocation module. The splitting module is used to split a computing task into N sub-computing tasks, where N is an integer greater than 1. The splitting module is also used to classify computing resources with the same execution time into computing resources of the same level based on the execution time of each computing network resource among multiple computing network resources corresponding to the target industrial park, resulting in M-level computing network resources. The target sub-computing task is one of the N sub-computing tasks, where M is an integer greater than 1. The acquisition module is further used to acquire the execution time set corresponding to each sub-computing task, where the execution time set includes the execution time of each level of computing network resource for each sub-computing task. The determination module is used to perform intersection calculation on the execution time sets corresponding to the N sub-computing tasks, and, if an intersection element exists, determine the computing network resource level corresponding to the intersection element in the execution time set corresponding to each sub-computing task. The allocation module is used to allocate computing network resources of the computing network resource level to the corresponding sub-computing task.
[0014] One possible implementation is that the above-mentioned acquisition module is used to merge the execution time sets corresponding to the N sub-computation tasks if there are no intersecting elements, to obtain a merged execution time set; the above-mentioned determination module is used to determine the overall mode of the merged execution time set; and the above-mentioned allocation module is used to allocate resources to the N sub-computation tasks based on the overall mode.
[0015] In another possible implementation, the first sub-computation task is one of N sub-computation tasks. The determining module is configured to: if the execution time set corresponding to the first sub-computation task contains the overall mode, then determine the network resource level corresponding to the overall mode in the execution time set as the network resource level corresponding to the first sub-computation task; and if the execution time set corresponding to the first sub-computation task does not contain the overall mode, then determine the network resource level corresponding to the element in the execution time set closest to the overall mode as the network resource level corresponding to the first sub-computation task. The allocation module is configured to allocate the network resources corresponding to the network resource level of each sub-computation task to the corresponding sub-computation task.
[0016] In another possible implementation, the aforementioned candidate allocation scheme is used to indicate the computing network resources allocated to each of the at least two sub-computing tasks. The aforementioned acquisition module is used to modify the intersection elements of the execution time sets corresponding to at least two sub-computing tasks within the N sub-computing tasks, where the computing network resource levels corresponding to these intersection elements are at the same level, based on a preset perturbation time, to obtain a modified execution time set. The aforementioned determination module is used to determine multiple candidate allocation schemes based on the modified execution time set; it is also used to determine the candidate allocation scheme with the smallest variance of the execution times corresponding to at least two sub-computing tasks among the multiple candidate allocation schemes as the target allocation scheme. The aforementioned allocation module is used to allocate resources to the at least two sub-computing tasks based on the target allocation scheme, wherein different sub-computing tasks within the at least two sub-computing tasks are allocated different computing network resources.
[0017] Another possible implementation is that the hierarchical interval between adjacent levels in the above-mentioned M-level computing network resources is determined based on at least one of the following: the minimum order of magnitude of the execution time interval between multiple computing network resources; the production interval time of the task requester; wherein the production interval time is the waiting time of the task requester before executing the next production instruction.
[0018] Another possible implementation is that the above allocation module is used to select computing network resources that have not been allocated to other sub-computing tasks among the N sub-computing tasks from the computing network resource level, and use them as computing network resources to allocate to the corresponding sub-computing tasks.
[0019] In another possible implementation, the first computing network resource is one of multiple computing network resources; the second sub-computing task is one of N sub-computing tasks. The acquisition module is used to acquire the communication time between the first computing network resource and the computing network management center, and the computation time for the first computing network resource to complete the second sub-computing task; the determination module is used to determine the sum of the communication time and computation time corresponding to the first computing network resource as the execution time for the first computing network resource to complete the second sub-computing task.
[0020] Thirdly, this application provides an electronic device comprising: a processor and a memory; the memory stores a program or instructions executable on the processor, wherein the program or instructions, when executed by the processor, implement the method of the first aspect described above.
[0021] Fourthly, this application provides a readable storage medium on which a program or instructions are stored, which, when executed by a computer, implement the method of the first aspect described above.
[0022] Fifthly, this application provides a computer program product stored in a storage medium, which, when executed by a computer, implements the method described in the first aspect.
[0023] In a sixth aspect, embodiments of this application provide a chip including a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the method described in the first aspect.
[0024] The beneficial effects of the second to sixth aspects mentioned above are described in the corresponding description of the first aspect and will not be repeated here. Attached Figure Description
[0025] Figure 1 A schematic diagram of the network architecture for an application of a network computing power scheduling method provided in this application embodiment;
[0026] Figure 2 A flowchart illustrating a network computing power scheduling method provided in an embodiment of this application;
[0027] Figure 3 A flowchart illustrating another network computing power scheduling method provided in this application embodiment;
[0028] Figure 4 A flowchart illustrating another network computing power scheduling method provided in this application embodiment;
[0029] Figure 5 A flowchart illustrating another network computing power scheduling method provided in this application embodiment;
[0030] Figure 6 A flowchart illustrating another network computing power scheduling method provided in this application embodiment;
[0031] Figure 7 A flowchart illustrating another network computing power scheduling method provided in this application embodiment;
[0032] Figure 8 A flowchart illustrating the implementation process of a network computing power scheduling method provided in this application embodiment;
[0033] Figure 9 A schematic diagram of the structure of a network computing power scheduling device provided in an embodiment of this application;
[0034] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0035] The network computing power scheduling method, apparatus, equipment, media, and program products provided in this application will now be described in detail with reference to the accompanying drawings.
[0036] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.
[0037] The terms "first," "second," etc., used in the specification and claims of this application 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 embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and 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.
[0038] The terms "at least one," "at least one of," etc., used in the specification and claims of this application refer to any one, any two, or a combination of two or more of the included items. For example, at least one of a, b, and c can mean: "a," "b," "c," "a and b," "a and c," "b and c," and "a, b, and c," where a, b, and c can be single or multiple. Similarly, "at least two" refers to two or more items, and its meaning is similar to that of "at least one."
[0039] In the description of this application, unless otherwise stated, "a plurality of" means two or more.
[0040] The embodiments of this application provide a network computing power scheduling method, apparatus, device, medium, and program product that can be applied to network computing power dynamic scheduling scenarios.
[0041] With the development of the Industrial Internet, the number of computing tasks in industrial parks is constantly increasing. Network computing scheduling methods have been applied to the management and control of computing resources in industrial parks to balance the load of computing nodes and improve the utilization rate of computing resources. Existing network computing scheduling methods usually use the load prediction data of computing nodes, task priority, node health and other factors as the basis for allocating computing resources, or use static latency, energy consumption and other factors as the triggering conditions for computing task migration.
[0042] However, the aforementioned network computing power scheduling method does not take into account the dynamic changes in production intervals in industrial parks, and the allocation criteria do not reflect the time required for network resources to complete tasks. In industrial parks, the production interval of task requesters directly affects production line scheduling and production efficiency. For computing tasks that are broken down into multiple sub-computing tasks, if the execution time of the network resources allocated to each sub-computing task is different, the completion time of each sub-computing task will be inconsistent. Task requesters will need to wait for the slowest sub-computing task to be completed, and the computing resources occupied by completed sub-computing tasks will be idle and easily preempted by other tasks, thereby reducing the continuity of production line operations.
[0043] To address the aforementioned technical problems, embodiments of this application provide a network computing power scheduling method, apparatus, device, medium, and program product. Since network resources are categorized based on the execution time of completing sub-computation tasks, and network resources with the same execution time are grouped into the same level, multiple network resources have a unified comparison benchmark for the time required to complete sub-computation tasks. Consequently, the execution time of each sub-computation task under each level of network resource can be recorded in the corresponding execution time set and compared with each other. Furthermore, the intersection element obtained by performing an intersection calculation on the execution time sets corresponding to each sub-computation task represents the execution time that each sub-computation task can achieve. Based on the intersection elements, the corresponding computing network resource level is determined for each sub-computing task and computing network resources are allocated. The execution time of the computing network resources allocated to each sub-computing task is consistent with each other. As a result, N sub-computing tasks can be completed at the same execution time. The task requester does not need to wait for the slowest sub-computing task to be completed. Therefore, it is possible to achieve equal-latency scheduling of computing tasks, so that computing power scheduling can be combined with the dynamic changes in the production interval time of the industrial park. Instead of allocating based on a basis that does not reflect the execution time, as in the existing technology, the completion time of each sub-computing task after the computing task is split is inconsistent. This reduces the situation of computing power resources being preempted and improves the continuity and efficiency of production line production.
[0044] The network computing power scheduling method, apparatus, device, medium, and program products provided in this application will be described in detail below with reference to the accompanying drawings.
[0045] Figure 1 This illustration shows a network architecture for a network computing power scheduling method provided in an embodiment of this application. For example... Figure 1 As shown, the network architecture includes a network computing power scheduling device 101 and a terminal device 102. The network computing power scheduling device 101 and the terminal device 102 are interconnected.
[0046] In some embodiments, the network computing power scheduling device 101 may be a server, a computer, or a processor or processing unit within a server or computer. The server may be a single server or a server cluster consisting of multiple servers. It should be noted that the embodiments of this application do not limit the specific device form of the network computing power scheduling device 101. Figure 1 The example shown is a single server using the network computing power scheduling device 101.
[0047] In some embodiments, the terminal device may be a mobile phone, tablet computer, laptop computer, handheld computer, in-vehicle electronic device, mobile internet device (MID), augmented reality (AR) / virtual reality (VR) device, robot, wearable device, personal computer (PC), ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc., and the embodiments of this application do not specifically limit it. Figure 1 The example shown is a mobile phone, with terminal device 102 as an example.
[0048] It should be noted that the network architecture described in the embodiments of this application is for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and does not constitute a limitation on the technical solutions provided in the embodiments of this application. As network architectures evolve, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.
[0049] See Figure 2 This is a flowchart illustrating a network computing power scheduling method provided in an embodiment of this application. Figure 2 As shown, the network computing power scheduling method provided in this application embodiment can be implemented by the above-mentioned network computing power scheduling device, specifically including the following steps 101 to 105.
[0050] Step 101: The computing network management center splits the computing task into N sub-computing tasks, where N is an integer greater than 1.
[0051] In some embodiments of this application, the aforementioned computing network management center is a system that uniformly manages, schedules, and orchestrates computing and network resources under the "computing power network" architecture. It can be the "intelligent brain" of the entire computing power network system.
[0052] In some examples, the aforementioned computing resources are hardware and supporting logical resources used to perform computing tasks, including CPU, GPU, memory, storage, etc., which can provide the business with data computing, model inference and data processing capabilities.
[0053] For example, large model inference computation tasks require the use of GPU computing resources. The user's query data is input, the GPU performs model calculations, and the inference results are output. The GPU and memory are the computing resources used in this task.
[0054] In some examples, the aforementioned network resources are communication resources that support service transmission, including link bandwidth, network slicing, transmission latency, forwarding channels, etc., used to ensure data interaction and data transmission for computing tasks.
[0055] For example, in large model inference tasks, the link bandwidth and network slices used by users to request data transmission to computing nodes and to return inference results to users are the network resources used by this service.
[0056] In some examples, the aforementioned computing power network uses the network as a carrier to uniformly perceive, identify, and schedule heterogeneous computing power and storage resources of intelligent computing, cloud, and edge nodes in various regions. Relying on network forwarding and control capabilities, it enables on-demand scheduling of computing power resources, completes business deployment and operation under computing network collaboration, and achieves routable computing power and schedulable business.
[0057] Alternatively, the aforementioned computing power network can also be understood as turning computing power into a network resource that can be scheduled.
[0058] Optionally, the aforementioned heterogeneous computing power refers to computing resources composed of various types of computing hardware, including CPUs, GPUs, NPUs, ASICs, and other computing units, each of which is adept at handling different computing tasks.
[0059] Optionally, the aforementioned storage resources are resources used for persistent data storage and temporary caching, including disks, memory caches, storage arrays, etc., providing data read / write and data storage capabilities for computing tasks.
[0060] In some embodiments of this application, the aforementioned computing tasks refer to data processing, computational reasoning, model calculation, and business processing tasks generated by users or services that require computing resources to perform.
[0061] In some examples, the type of computational task described above is a computational task that can be broken down.
[0062] For example, a decomposable computational task could be an overall AI image recognition computational task for 10,000 images, which can be broken down into multiple groups of image recognition sub-computation tasks. Each group of sub-computation tasks is independent of each other, has no execution dependency, and can be scheduled to different heterogeneous computing power nodes for parallel operation to complete the image recognition work synchronously.
[0063] Optionally, the aforementioned divisible computing task refers to a computing task that can be divided into N sub-computing tasks that can be executed in parallel. These N sub-computing tasks are the smallest unit for the schedulable computing network resources of the industrial park.
[0064] Optionally, the above N sub-computing tasks are the smallest unit for the schedulable computing network resources of the industrial park. This means that for the schedulable computing network resources of the industrial park, any sub-computing task can no longer be decomposed into sub-computing tasks that can be executed in parallel. It can only be completed by a single schedulable computing network resource of the industrial park, and cannot be distributed to multiple computing network resources for parallel execution.
[0065] In some embodiments of this application, the aforementioned N sub-computation tasks refer to the smallest sub-computation tasks that can be executed in parallel, which are divided into a divisible computation task.
[0066] In some examples, the above N sub-computation tasks are at least 2 sub-computation tasks.
[0067] For example, the above sub-computation task could be: transcode a 60-minute original high-definition video to generate video files of various resolutions;
[0068] The complete video can be sliced by duration and broken down into multiple independent sub-computation tasks: Sub-computation task 1: processes video segments from 0 to 15 minutes, completing decoding and encoding / transcoding; Sub-computation task 2: processes video segments from 15 to 30 minutes, completing decoding and encoding / transcoding; Sub-computation task 3: processes video segments from 30 to 45 minutes, completing decoding and encoding / transcoding; Sub-computation task 4: processes video segments from 45 to 60 minutes, completing decoding and encoding / transcoding.
[0069] In some embodiments of this application, the computing network management center divides the computing task into N sub-computing tasks, where N is an integer greater than 1. This can be understood as follows: the computing network management center identifies all computing tasks, identifies computing tasks that can be divided, and divides the computing task into multiple smallest sub-computing tasks that can be executed in parallel, wherein there are at least 2 sub-computing tasks.
[0070] Step 102: Based on the execution time of each computing network resource in the multiple computing network resources corresponding to the target industrial park to complete the target sub-computation task, the computing network management center divides the computing network resources with the same execution time into computing network resources of the same level, thus obtaining M-level computing network resources.
[0071] In some embodiments of this application, the aforementioned target industrial park refers to a specific industrial cluster area that has been uniformly planned, gathers multiple industrial enterprises, and deploys industrial production equipment, computing power, and network infrastructure. A large number of industrial computing tasks can be generated within the park.
[0072] In some examples, the aforementioned industrial computing tasks refer to data processing and computation tasks generated during industrial production processes within industrial parks, serving industrial operations such as production control, product testing, and simulation analysis.
[0073] For example, industrial computing tasks could involve detecting defects in the appearance of parts by collecting image data of a batch of parts. The overall task can be broken down into multiple sub-tasks for image defect recognition. Each sub-task is independent of the others and can be scheduled to run in parallel by heterogeneous computing nodes to complete the recognition of defects such as scratches and damage to parts.
[0074] In some embodiments of this application, the aforementioned computing network resources are a collective term for computing power resources and network resources, including heterogeneous computing power, storage resources, and various communication network resources, providing computing power, data storage capabilities, and data transmission capabilities for computing tasks.
[0075] For example, in the task of detecting defects in industrial parts, the computing power of GPUs used for image recognition inference, the storage resources for storing part images, and the bandwidth resources for transmitting image data together constitute the computing network resources used by this service.
[0076] In some embodiments of this application, the above execution time is the time for the computing network management center to communicate with the i-th computing network resource and the time for the i-th computing network resource to complete a computing task.
[0077] In the embodiments of this application, the target sub-computation task is one of N sub-computation tasks, where M is an integer greater than 1.
[0078] In some embodiments of this application, the above-mentioned N sub-computation tasks are all the sub-computation tasks after the above-mentioned divisible computation task has been divided.
[0079] In some embodiments of the application, the hierarchical interval between adjacent levels in the above-mentioned M-level computing network resources is determined based on at least one of the following: the minimum order of magnitude of the execution time interval between multiple computing network resources; the production interval time of the task requester.
[0080] In some examples, the task requester is the entity that initiates the computing task, submits the task request, and needs to call computing network resources to complete the business processing.
[0081] In some examples, the above production interval is the waiting time for the task requester before executing the next production instruction.
[0082] In some examples, the aforementioned grading interval refers to the time interval at which network resources that have already been graded based on the same execution time are graded again.
[0083] In some examples, the minimum order of magnitude of the execution time interval between the multiple computing network resources can be understood as: the minimum value of the interval between the execution time between the multiple computing network resources, that is, the time for the computing network management center to communicate with the i-th computing network resource, and the time for the i-th computing network resource to complete a computing task.
[0084] In some examples, the hierarchical interval between adjacent levels in the aforementioned M-level computing network resources can also be determined by collecting a large amount of historical indicator data of computing network resources and determining the hierarchical interval based on the distribution characteristics of the indicators.
[0085] Thus, by first calculating the execution time of each sub-computation task using computing network resources to initially classify the computing network resources, and then grouping those resources into the same level if the execution times of each sub-computation task are the same, the computing network resources are further classified according to the classification intervals between adjacent levels within the multi-level computing network resources. This facilitates the subsequent acquisition of the execution time set of N sub-computation tasks.
[0086] In some embodiments of this application, combined with Figure 2 ,like Figure 7 As shown, before classifying computing network resources with the same execution time into computing network resources of the same level, step 102 can be implemented through steps 301 to 302 as follows.
[0087] Step 301: The computing network management center obtains the communication time between the first computing network resource and the computing network management center, as well as the computing time for the first computing network resource to complete the second sub-computing task.
[0088] In some embodiments of this application, the first computing network resource is one of a plurality of computing network resources.
[0089] In some embodiments of this application, the aforementioned second sub-computation task is one of N sub-computation tasks.
[0090] In some examples, the computation time for the first computing network resource to complete the second sub-computation task is related to the amount of computation task and the performance of the computing network resource. It can be an estimated quantity, such as: number of task instructions / CPU clock frequency / average clock cycles required for the computing network resource to execute one instruction.
[0091] In some examples, the aforementioned communication time can be understood as the time when the computing network management center notifies the first computing network resource that it will execute a computing task.
[0092] Step 302: The computing network management center determines the sum of the communication time and computing time corresponding to the first computing network resource as the execution time for the first computing network resource to complete the second sub-computation task.
[0093] In some examples, the above execution time is used to classify computing network resources by level.
[0094] In some examples, the first computing network resource mentioned above is one of multiple computing network resources.
[0095] In some examples, the second subcomputation task mentioned above is one of N subcomputation tasks.
[0096] In some examples, the aforementioned computing network management center determines the sum of the communication time and computation time corresponding to the first computing network resource as the execution time for the first computing network resource to complete the second sub-computation task. This can be understood as follows: the computing network management center performs calculations based on the communication time between each computing network resource and the computing network management center and the computation time of the second sub-computation task. The calculation can be performed by adding them together, and the result is used as the execution time of each computing network resource.
[0097] Thus, by classifying computing network resources by the sum of the communication time between the first computing network resource and the computing network management center, and the computing time for the first computing network resource to complete the second sub-computing task, the latency scheduling of computing tasks can be further improved.
[0098] Step 103: The computing network management center obtains the set of execution times corresponding to each sub-computing task.
[0099] In the embodiments of this application, the execution time set corresponding to each sub-computing task includes the execution time of each sub-computing task completed by each level of computing network resources.
[0100] In some embodiments of this application, the above-mentioned execution time set refers to the set of execution times of each sub-computing task under different levels of computing network resources.
[0101] For example, the execution time set can be
[0102] Q1[1] Q1[2] Q1[3] Q1[4] Q1[5] = 5 10 15 20 25
[0103] Q2[1] Q2[2] Q2[3] Q2[4] Q2[5] = 10 15 20 25 30
[0104] Q3[1] Q3[2] Q3[3] Q3[4] Q3[5] = 10 40 80 100 120;
[0105] Where Qi[J] refers to the execution time of the i-th sub-computation task using J-level computing network resources for computing network management.
[0106] In some embodiments of this application, the above-mentioned computing network management center obtains the set of execution times corresponding to each sub-computing task, which can be understood as: the computing network management center determines a set based on the execution time of each sub-computing task under different levels of computing network resources.
[0107] Step 104: The computing network management center performs intersection calculation on the execution time sets corresponding to the N sub-computing tasks, and if there are intersection elements, determines the computing network resource level corresponding to the intersection element in the execution time set corresponding to each sub-computing task.
[0108] In some embodiments of this application, the intersection mentioned above refers to the set consisting of elements that coexist in multiple sets.
[0109] For example, if set A = {apple, banana, orange, grape} and set B = {banana, grape, watermelon, mango}, then the intersection of sets A and B consists of bananas and grapes.
[0110] In some embodiments of this application, the aforementioned computing network resource level refers to one of the multiple computing network levels obtained based on the fact that the execution times of the N sub-computing tasks are the same.
[0111] In some embodiments of this application, the aforementioned computing network management center performs an intersection calculation on the execution time sets corresponding to N sub-computing tasks. If an intersection element exists, it determines the computing network resource level corresponding to the intersection element in the execution time set of each sub-computing task. This can be understood as follows: the computing network management center performs an intersection operation on the execution time sets of each of the multiple sub-computing tasks split from the computing task, finding the intersection element corresponding to the execution time set of each sub-computing task. The premise is that the computing network resources corresponding to the time level indicated by the intersection element in the execution time set of each sub-computing task do not overlap. Based on the time indicated by the intersection element, it finds the computing network resources of the corresponding level for executing the sub-computing task.
[0112] In some embodiments of this application, combined with Figure 2 ,like Figure 3 As shown, in step 104 above, after performing the intersection calculation on the execution time sets corresponding to the N sub-computation tasks, it can be further implemented through the following steps 201 to 202.
[0113] Step 201: If there are no intersecting elements, the computing network management center will merge the execution time sets corresponding to the N sub-computing tasks to obtain the merged execution time set.
[0114] In some examples, the above-mentioned merged execution time set is obtained by merging the execution time sets of N sub-computation tasks that have no overlapping elements.
[0115] In some examples, if there is no intersection element, the computing network management center will merge the execution time sets corresponding to the N sub-computing tasks to obtain a merged execution time set. This can be understood as follows: the computing network management center processes the execution time sets of the N sub-computing tasks based on the result of the intersection calculation of the execution time sets corresponding to the N sub-computing tasks. First, it performs an intersection operation. If, after the intersection operation, there is no intersection element in the execution time sets of the N sub-computing tasks, the above sets are merged to obtain a merged execution time set.
[0116] Step 202: The computing network management center determines the overall mode of the merged execution time set and allocates resources to the N sub-computing tasks based on the overall mode.
[0117] In some examples, the mode refers to the element that appears most frequently across all sets.
[0118] In some examples, the aforementioned computing network management center determines the overall mode of the merged execution time set and allocates resources to N sub-computing tasks based on the overall mode. This can be understood as follows: the computing network management center calculates the mode of the merged execution time set by first counting all elements in the merged execution time set, counting the frequency of each element, taking the element with the most occurrences as the overall mode, and allocating computing network resources to the corresponding sub-computing tasks based on the overall mode value. The allocated resource is one of the multiple levels of computing network resources corresponding to the time value of the overall mode.
[0119] Thus, by first calculating the intersection of the execution time sets of N sub-computation tasks, and then calculating the mode if there is no intersection element, equal-latency scheduling of computation tasks can be achieved.
[0120] In some embodiments of this application, combined with Figure 3 ,like Figure 4 As shown, step 202 above can be specifically implemented through steps 202a to 202c below.
[0121] Step 202a: If the execution time set corresponding to the first sub-computation task contains the overall mode, then the computing network management center determines the computing network resource level corresponding to the overall mode in the execution time set as the computing network resource level corresponding to the first sub-computation task.
[0122] In some embodiments of this application, the first sub-computation task mentioned above is one of N sub-computation tasks.
[0123] In some examples, the overall mode mentioned above is obtained by merging the execution time sets that have no overlapping elements through intersection calculation and then performing mode calculation on the merged execution time set.
[0124] Step 202b: If the execution time set corresponding to the first sub-computation task does not contain the overall mode, the computing network management center will determine the computing network resource level corresponding to the element in the execution time set that is closest to the overall mode as the computing network resource level corresponding to the first sub-computation task.
[0125] In some embodiments of this application, the first sub-computation task mentioned above is one of N sub-computation tasks.
[0126] In some embodiments of the application, if the execution time set corresponding to the first sub-computation task does not contain the overall mode, the computing network management center determines the computing network resource level corresponding to the element in the execution time set that is closest to the overall mode as the computing network resource level corresponding to the first sub-computation task. This can be understood as follows: if, after performing intersection calculation and mode calculation, some sub-computation tasks still do not have the elements of the above calculation results in their execution time sets, then the computing network level corresponding to the element in the execution time set of the sub-computation task that is closest to the element value corresponding to the overall mode calculation result obtained by the above mode calculation is found, and the computing network resource of this level is determined as the computing network resource executed by the sub-computation task.
[0127] Step 202c: The computing network management center allocates the computing network resources at the computing network resource level corresponding to each sub-computing task to the corresponding sub-computing task.
[0128] In some examples, each of the sub-computation tasks mentioned above refers to each of all the sub-computation tasks that a computation task is broken down into.
[0129] In some examples, the aforementioned computing network management center allocates the computing network resources at the computing network resource level corresponding to each sub-computing task to the corresponding sub-computing task. This can be understood as follows: the computing network management center determines the computing network resources at the computing network level corresponding to each sub-computing task based on the methods described above, such as intersection, mode, and finding the closest value of the mode, and then allocates them to the corresponding sub-computing task.
[0130] Thus, since the network resources corresponding to the N sub-computing tasks cannot be determined based on the intersection elements, the network resources corresponding to the N sub-computing tasks are determined by selecting the mode. If some sub-computing tasks still do not have corresponding network resources, the network resources of the sub-computing tasks are determined by finding the value closest to the mode determined above. It can be seen that the network resources of the sub-computing tasks are determined by a step-by-step method, which can significantly reduce the situation of computing resources being preempted and achieve equal-latency scheduling of computing tasks.
[0131] Step 105: The computing network management center allocates computing network resources at the computing network resource level to the corresponding sub-computing tasks.
[0132] In some embodiments of this application, the aforementioned sub-computation task refers to each sub-computation task of a subdivisible computation task.
[0133] In some embodiments of this application, the computing network management center allocates computing network resources at the computing network resource level to the corresponding sub-computing tasks. This can be understood as follows: the computing network management center finds the computing network resources at the computing network resource level corresponding to each sub-computing task based on the above-mentioned intersection, mode, and the method of finding the closest mode value, and allocates the corresponding computing network resources to each sub-computing task for execution.
[0134] In some embodiments of this application, combined with Figure 2 ,like Figure 5 As shown, step 105 above can be implemented through steps 105a to 105d below.
[0135] Step 105a: If the network management center finds that the network resource level corresponding to the intersection element of the execution time set of at least two sub-computing tasks is the same, it modifies the intersection element of the execution time set of at least two sub-computing tasks based on a preset disturbance time to obtain the modified execution time set.
[0136] In some examples, the aforementioned preset disturbance time can be determined by methods such as the minimum order of magnitude of the interval between the communication time between the first computing network resource and the computing network management center and the interval between the execution time of the second sub-computing task by the first computing network resource, or by the production rhythm of the task requester.
[0137] In some examples, the above-mentioned modification of the intersection elements may refer to changing the values of the obtained intersection elements by means of the preset perturbation time.
[0138] In some examples, when the intersection element of the execution time sets corresponding to at least two sub-computing tasks in N sub-computing tasks corresponds to the same computing network resource level, the computing network management center modifies the intersection element of the execution time sets corresponding to at least two sub-computing tasks based on a preset disturbance time, obtaining a modified execution time set. This can be understood as follows: the computing network management center allocates computing network resources for executing each sub-computing task based on the result of the intersection element calculated from the execution time set of each sub-computing task; when at least two sub-computing tasks have the same computing network level corresponding to the intersection element of their execution time sets, the intersection element of the execution time sets of these sub-computing tasks with the same computing network resource level is disturbed, and the value of the intersection element is changed according to a preset disturbance time, resulting in a modified execution time set of sub-computing tasks with the same computing network resource level.
[0139] Step 105b: The computing network management center determines multiple candidate allocation schemes based on the changed execution time set.
[0140] In some embodiments of this application, the above-described candidate allocation scheme is used to indicate the computing network resources allocated to each of at least two sub-computing tasks.
[0141] In some examples, the aforementioned computing network management center determines multiple candidate allocation schemes based on the changed execution time set. This can be understood as follows: the computing network management center selects different methods to calculate different candidate allocation schemes based on the obtained new execution time set, and selects the optimal candidate scheme from among them.
[0142] Alternatively, the above optimal candidate scheme can be understood as: the time interval between the completion of each sub-computation task is the shortest, that is, the latency of executing each sub-computation task is similar.
[0143] Step 105c: The computing network management center determines the candidate allocation scheme with the smallest variance in execution time for at least two sub-computation tasks among multiple candidate allocation schemes as the target allocation scheme.
[0144] In some examples, the variance described above describes the degree of dispersion of the individual values of a set of data relative to the mean; the larger the variance, the greater the data fluctuation; the smaller the variance, the more concentrated the data.
[0145] For example, in the dataset {1,3,5}, the mean is (1+3+5) ÷3 = 3. Subtracting the mean from each number and squaring them gives 4, 0, and 4 respectively. Squaring them again gives 4+0+4=8. The variance is the sum of squares divided by the number of data points, so the variance is 8÷3≈2.67.
[0146] In some examples, the target allocation scheme mentioned above refers to the optimal allocation scheme among the candidate allocation schemes.
[0147] In some examples, the aforementioned computing network management center determines the candidate allocation scheme with the smallest variance in execution time among at least two sub-computation tasks from multiple candidate allocation schemes as the target allocation scheme. This can be understood as follows: the computing network management center selects the optimal allocation scheme as the target allocation scheme based on the results of all candidate allocation schemes. The optimal target allocation scheme can be obtained by considering the computing network resource allocation scheme corresponding to the variance in execution time of at least two sub-computation tasks.
[0148] Step 105d: The computing network management center allocates resources to at least two sub-computing tasks based on the target allocation scheme. Different sub-computing tasks within the at least two sub-computing tasks are allocated different computing network resources.
[0149] In some examples, the above target allocation scheme can be understood as follows: the sub-computation execution time value calculated according to this allocation scheme is closest to the execution time value of the sub-computation task for which the computing network resources have been determined. In other words, the execution time of the computing network resources for each sub-computation task is approximately equal to the latency of the computing network management center.
[0150] Thus, when a conflict occurs in computing network resources, the intersection elements are first updated based on a preset disturbance time. Then, based on the updated set of execution times of sub-computation tasks of computing network resources with the same level, the optimal computing network resource allocation scheme is selected from different candidate allocation schemes to allocate the sub-computation tasks. Therefore, the situation of computing resources being preempted can be reduced, and each sub-computation task can achieve near-equal latency for the computing network management center.
[0151] In some embodiments of this application, combined with Figure 2 ,like Figure 6 As shown, step 105 above can also be implemented through step 105A below.
[0152] Step 105A: The computing network management center selects computing network resources from the computing network resources at the computing network resource level that have not been allocated to other sub-computing tasks in the N sub-computing tasks, and uses them as computing network resources to allocate to the corresponding sub-computing tasks.
[0153] In some examples, the aforementioned unallocated computing network resources refer to the unused computing network resources other than those corresponding to the computing network resource level for each sub-computing task as determined in the examples above.
[0154] In some examples, the aforementioned computing network management center selects computing network resources from the computing network resources at the computing network resource level that have not been allocated to other sub-computing tasks in the N sub-computing tasks, and uses these resources as computing network resources to allocate to the corresponding sub-computing tasks. This can be understood as follows: the computing network management center queries the remaining sub-computing tasks that have not been allocated computing network resources, queries the unused computing network resources, and selects the computing network resources that have not been allocated to other sub-computing tasks in the N sub-computing tasks, and uses these resources as computing network resources to allocate to the sub-computing tasks that have not been allocated computing network resources.
[0155] In this way, by allocating sub-computing tasks based on unallocated computing network resources, the situation of computing resources being preempted can be greatly reduced, and the continuity and efficiency of production lines in industrial parks can be improved.
[0156] In the network computing power scheduling method provided in this application, network resources are classified according to the execution time of completing sub-computation tasks, and network resources with the same execution time are classified into network resources of the same level. This provides a unified comparison benchmark for the time required to complete sub-computation tasks across multiple network resources. Consequently, the execution time of each sub-computation task under each level of network resource can be recorded in the corresponding execution time set and compared with each other. Furthermore, the intersection element obtained by performing an intersection calculation on the execution time sets corresponding to each sub-computation task represents the execution time that each sub-computation task can achieve. Thus, each sub-computation task is scheduled based on the intersection element. The computing task determines the corresponding computing network resource level and allocates computing network resources. The execution time of the computing network resources allocated to each sub-computing task is consistent with each other. As a result, N sub-computing tasks can be completed at the same execution time. The task requester does not need to wait for the slowest sub-computing task to be completed. Therefore, it is possible to achieve equal-latency scheduling of computing tasks, so that computing power scheduling can be combined with the dynamic changes in the production interval of the industrial park. Instead of allocating based on criteria that do not reflect execution time, as in the existing technology, the completion time of each sub-computing task after the computing task is split is inconsistent. This reduces the situation of computing power resources being preempted and improves the continuity and efficiency of production line production.
[0157] The network computing power scheduling method of this application will be described below through specific embodiments.
[0158] like Figure 8 As shown, the implementation process of the network computing power scheduling method provided in this application embodiment includes the following S1 to S3:
[0159] S1, Decompose the calculation task.
[0160] For example, the requester in the industrial park sends a computing task, and the computing network management center analyzes the computing task to find a computing task that can be broken down into multiple sub-computing tasks.
[0161] For example, the decomposability of the aforementioned computational task Q means that the computational task Q can be broken down into sub-computational tasks Q1, Q2, Q3...Qn that can be executed in parallel. These sub-computational tasks Qi are the smallest unit for the schedulable computing network resources of the industrial park. The fact that Qi is the smallest unit for the schedulable computing network resources of the industrial park means that, given the schedulable computing network resources of the industrial park, the computational task Qi can no longer be broken down into sub-computational tasks that can be executed in parallel. Qi can only be completed by a single schedulable computing network resource of the industrial park, and cannot be distributed to multiple computing network resources for parallel execution.
[0162] For example, the above Q = Q1 + Q2 + Q3 + ... + Qn.
[0163] For example, the default computing network management center (or a similar platform, system, or device with the same function) connects to all schedulable computing network resources in the industrial park through an operational network, receives computing tasks (requests), and schedules computing network resources to complete the task. Meanwhile, the connection between the industrial park's computing network resources and the computing network management center remains stable. A requester of a computing task publishes a computing request to the computing network management center and then waits for the center to provide the result.
[0164] For example, the computing network management center breaks down the computing task into N sub-computing tasks according to the above example.
[0165] S2. Create a calculation timetable.
[0166] For example, for the party issuing (requesting) a computing task, the time required to obtain the result of the computing task can be divided into the following parts: the communication time between the requester and the computing network management center, the allocation time of the computing network management center in assigning the computing task, the communication time between the computing network management center and the allocated computing network resources, and the computing time of the computing network resources. Among these, the communication time between the requester and the computing network management center, and the allocation time of the computing task by the computing network management center, do not change due to different scheduling results. Therefore, for the party issuing the computing task, the perceived latency is determined by the communication time between the computing network management center and the allocated computing network resources, and the computing time of the computing network resources.
[0167] For example, since the computing network resources are located in an industrial park, the connection between the computing network management center and the computing network is assumed to be stable, and their communication time is denoted by Ti. This represents the communication time between the computing network management center and the i-th computing network resource.
[0168] For example, TEXEi represents the time it takes for the i-th computing network resource to complete a computing task. This time is related to the amount of computing tasks and the performance of the computing network resource. For the computing network management center, TEXEi is an estimate, and a possible use case is: number of task instructions / CPU clock frequency / average number of clock cycles required for the computing network resource to execute one instruction.
[0169] For example, for an indivisible computational task Qexp, the process time for the i-th computational network resource to complete the task and obtain the result by the computational network management center is Ti+TEXEi. The computational network resources that the computational network center can allocate for this computational task are classified into levels from low to high based on the value of Ti+TEXEi, with computational network resources of the same value grouped into the same level, resulting in C levels of classification.
[0170] For example, the selection of the above-mentioned hierarchical interval can be determined by the minimum order of magnitude of the interval between Ti and TEXEi, or by the production tick of the task requester, such as ms (milliseconds). The production tick of the task requester refers to the waiting time before the task requester executes the next production instruction.
[0171] For example, for a decomposable computing task Q = Q1 + Q2 + Q3 + ... + Qn, the computing network management center can obtain the following schedule.
[0172] Q1[1] Q1[2] Q1[3] … Q1[C]
[0173] Q2[1] Q2[2] Q2[3] … Q2[C]
[0174] …
[0175] Qn[1] Qn[2] Qn[3] … Qn[C]
[0176] For example, Qi[J] above represents the process time for the computing network management center to complete the Qi task and obtain the result using J-level computing network resources.
[0177] S3. Schedule network resources according to the timetable.
[0178] For example, each row's Qi[J] value forms a set with C elements. For a computation task Q, n sets with C elements can be formed. The intersection of these n sets yields a common element T. Using T to infer the computing network resource level in each set indicates that the computing network management center will use computing network resources at that level to complete the corresponding task.
[0179] For example, an example of a timetable:
[0180] Q1[1] Q1[2] Q1[3] Q1[4] Q1[5] = 5 10 15 20 25
[0181] Q2[1] Q2[2] Q2[3] Q2[4] Q2[5] = 10 15 20 25 30
[0182] Q3[1] Q3[2] Q3[3] Q3[4] Q3[5] = 10 40 80 100 120.
[0183] For example, the computation task Q can be broken down into three sub-tasks Q1, Q2, and Q3. The computing network management center can then divide the available computing network resources for these three tasks into five levels. The union of the three intersections is 10. Conversely, for task Q1, computing network resources at level 2 are selected for computation; for task Q2, computing network resources at level 1 are selected for computation; and for task Q3, computing network resources at level 1 are selected for computation.
[0184] For example, in the case where the intersection is empty, the mode of the elements is selected as T. For sets that do not contain T, the value closest to T is used as T for reverse calculation. In this case, each Qi task has nearly equal latency to the computing network management center.
[0185] For example, in the case of insufficient computing network resources, if the same computing network resources are selected after reverse deduction, it does not meet the requirements of parallelism because the tasks are non-indivisible and are executed in parallel.
[0186] For example, the scheduling method in the above case can be:
[0187] The first step is to determine the number of computing network resources to select for conflict resolution.
[0188] The second step is to select non-conflicting computing network resources;
[0189] The third step involves perturbing the time required for the Qi task corresponding to the conflicting network resources, starting with T. This leads to the derivation of non-conflicting network resource selection methods. Among different selection methods, the optimal one can be chosen, such as the method with the lowest variance.
[0190] For example, the perturbation step size can be on the order of the minimum interval between Ti and TEXEi or determined by the production cycle of the task requester, such as ms (milliseconds).
[0191] For example, at this point, each Qi task has near-equal latency for the computing network management center. If there is no solution, it indicates that computing network resources are being reused, which means that the campus computing network resources are insufficient, and the computing network management center should notify campus users.
[0192] Thus, since computing network resources are categorized based on the execution time of completing sub-computation tasks, and computing network resources with the same execution time are grouped into the same level, multiple computing network resources have a unified comparison benchmark for the time required to complete sub-computation tasks. Consequently, the execution time of each sub-computation task under each level of computing network resource can be recorded in the corresponding execution time set and compared with each other. Furthermore, the intersection element obtained by calculating the intersection of the execution time sets corresponding to each sub-computation task represents the execution time that each sub-computation task can achieve. Therefore, the execution time of each sub-computation task is determined based on the intersection element. The corresponding computing network resource level is determined and computing network resources are allocated accordingly. The execution time of the computing network resources allocated to each sub-computing task is consistent with each other. As a result, N sub-computing tasks can be completed at the same execution time. The task requester does not need to wait for the slowest sub-computing task to be completed. Therefore, it is possible to achieve equal-latency scheduling of computing tasks, so that computing power scheduling can be combined with the dynamic changes in the production interval of the industrial park. Instead of allocating based on a basis that does not reflect the execution time, as in the existing technology, the completion time of each sub-computing task after the computing task is split is inconsistent. This reduces the situation of computing power resources being preempted and improves the continuity and efficiency of production line production.
[0193] It should be noted that the descriptions of each step S1 to S3 in this embodiment can be found in the descriptions in the above embodiments, and will not be repeated here.
[0194] It should be noted that the above-described method embodiments, or the various possible implementations of the method embodiments, can be executed individually, or, provided there is no conflict, they can be combined with each other. The specific implementation can be determined according to actual usage requirements, and this application embodiment does not impose any restrictions on this.
[0195] As can be seen, the above mainly describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, the embodiments of this application provide corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the modules and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0196] This application embodiment can divide the network computing power scheduling device into functional modules according to the above method example. For example, each function can be divided into a separate functional module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. Optionally, the module division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.
[0197] In some embodiments, this application also provides a network computing power scheduling device. This network computing power scheduling device may include one or more functional modules for implementing the network computing power scheduling method of the above method embodiments.
[0198] For example, Figure 9 This is a schematic diagram of a network computing power scheduling device provided in an embodiment of this application. Figure 9 As shown, the network computing power scheduling device 900 includes: a splitting module 901, an acquisition module 902, a determination module 903, and an allocation module 904.
[0199] The aforementioned device includes: a splitting module 901, an acquisition module 902, a determination module 903, and an allocation module 904. The splitting module 901 is used to split a computational task into N sub-computational tasks, where N is an integer greater than 1. Based on the execution time of each computing network resource among multiple computing network resources corresponding to the target industrial park in completing the target sub-computational task, the splitting module 901 classifies computing network resources with the same execution time into computing network resources of the same level, resulting in M-level computing network resources. The target sub-computational task is one of the N sub-computational tasks, where M is an integer greater than 1. The acquisition module 902 is further used to acquire the execution time set corresponding to each sub-computational task, which includes the execution time of each level of computing network resource in completing each sub-computational task. The determination module 903 is used to perform intersection calculation on the execution time sets corresponding to the N sub-computational tasks, and, if an intersection element exists, determines the computing network resource level corresponding to the intersection element in the execution time set corresponding to each sub-computational task. The allocation module 904 is used to allocate computing network resources at the computing network resource level to the corresponding sub-computational tasks.
[0200] In the network computing power scheduling device provided in this application, network resources are classified according to the execution time of completing sub-computation tasks, and network resources with the same execution time are classified into network resources of the same level. This provides a unified comparison benchmark for the time required to complete sub-computation tasks across multiple network resources. Consequently, the execution time of each sub-computation task under each level of network resource can be recorded in the corresponding execution time set and compared with each other. Furthermore, the intersection element obtained by performing an intersection calculation on the execution time sets corresponding to each sub-computation task represents the execution time that each sub-computation task can achieve. Thus, each sub-computation task is scheduled based on the intersection element. The computing task determines the corresponding computing network resource level and allocates computing network resources. The execution time of the computing network resources allocated to each sub-computing task is consistent with each other. As a result, N sub-computing tasks can be completed at the same execution time. The task requester does not need to wait for the slowest sub-computing task to be completed. Therefore, it is possible to achieve equal-latency scheduling of computing tasks, so that computing power scheduling can be combined with the dynamic changes in the production interval of the industrial park. Instead of allocating based on criteria that do not reflect execution time, as in the existing technology, the completion time of each sub-computing task after the computing task is split is inconsistent. This reduces the situation of computing power resources being preempted and improves the continuity and efficiency of production line production.
[0201] In some embodiments, the acquisition module 902 is used to merge the execution time sets corresponding to the N sub-computation tasks if there are no intersection elements, to obtain a merged execution time set; the determination module 903 is used to determine the overall mode of the merged execution time set; and the allocation module 904 is used to allocate resources to the N sub-computation tasks based on the overall mode.
[0202] In other embodiments, the first sub-computation task is one of N sub-computation tasks. The determining module 903 is configured to, if the execution time set corresponding to the first sub-computation task contains the overall mode, determine the computing network resource level corresponding to the overall mode in the execution time set as the computing network resource level corresponding to the first sub-computation task; and is further configured to, if the execution time set corresponding to the first sub-computation task does not contain the overall mode, determine the computing network resource level corresponding to the element in the execution time set closest to the overall mode as the computing network resource level corresponding to the first sub-computation task; the allocation module 904 is configured to allocate the computing network resources corresponding to the computing network resource level of each sub-computation task to the corresponding sub-computation task.
[0203] In some other embodiments, the aforementioned candidate allocation scheme is used to indicate the computing network resources allocated to each of the at least two sub-computing tasks. The acquisition module 902 is used to modify the intersection elements of the execution time sets corresponding to at least two sub-computing tasks within the N sub-computing tasks, based on a preset disturbance time, when the computing network resource levels corresponding to the intersection elements are the same, to obtain a modified execution time set. The determination module 903 is used to determine multiple candidate allocation schemes based on the modified execution time set; it is also used to determine the candidate allocation scheme with the smallest variance of the execution times corresponding to at least two sub-computing tasks among the multiple candidate allocation schemes as the target allocation scheme. The allocation module 904 is used to allocate resources to the at least two sub-computing tasks based on the target allocation scheme, wherein different sub-computing tasks within the at least two sub-computing tasks are allocated different computing network resources.
[0204] In some other embodiments, the hierarchical interval between adjacent levels in the aforementioned M-level computing network resources is determined based on at least one of the following: the minimum order of magnitude of the execution time interval between multiple computing network resources; the production interval time of the task requester; wherein the production interval time is the waiting time of the task requester before executing the next production instruction.
[0205] In some other embodiments, the allocation module 904 is used to select computing network resources that have not been allocated to other sub-computing tasks among the N sub-computing tasks from the computing network resources at the computing network resource level, and use them as computing network resources allocated to the corresponding sub-computing tasks.
[0206] In some other embodiments, the first computing network resource is one of multiple computing network resources; the second sub-computing task is one of N sub-computing tasks. The acquisition module 902 is used to acquire the communication time between the first computing network resource and the computing network management center, and the computation time for the first computing network resource to complete the second sub-computing task; the determination module 903 is used to determine the sum of the communication time and computation time corresponding to the first computing network resource as the execution time for the first computing network resource to complete the second sub-computing task.
[0207] It should be noted that the network computing power scheduling device can implement all the processes implemented in the above method embodiments and achieve the same beneficial effects. To avoid repetition, it will not be described again here.
[0208] In the case where the functions of the integrated modules described above are implemented in hardware, this application provides a possible structural schematic diagram of the electronic device involved in the above embodiments. For example... Figure 10 As shown, the electronic device 90 includes: a processor 92, a communication interface 93, and a bus 94. Optionally, the electronic device 90 may also include a memory 91.
[0209] Processor 92 may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 92 may be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 92 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.
[0210] Communication interface 93 is used to connect with other devices via a communication network. This communication network can be Ethernet, wireless access network, wireless local area network (WLAN), etc.
[0211] The memory 91 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.
[0212] As one possible implementation, the memory 91 can exist independently of the processor 92. The memory 91 can be connected to the processor 92 via a bus 94 and is used to store instructions or program code. When the processor 92 calls and executes the instructions or program code stored in the memory 91, it can implement the network computing power scheduling method provided in the embodiments of this application.
[0213] In another possible implementation, memory 91 can also be integrated with processor 92.
[0214] Bus 94 can be an Extended Industry Standard Architecture (EISA) bus, etc. Bus 94 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 10 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.
[0215] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the service calling device can be divided into different functional modules to complete all or part of the functions described above.
[0216] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface and the processor are coupled. The processor is used to run programs or instructions to implement the various processes of the above network computing power scheduling method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.
[0217] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.
[0218] This application also provides a readable storage medium storing a program or instructions. When executed by a computer, the program or instructions implement the network computing power scheduling method provided in the above embodiments. It is understood that all or part of the processes in the above method embodiments can be executed by computer instructions instructing related hardware. The readable storage medium can be any of the foregoing embodiments or memory. The readable storage medium can also be an external storage device of the service invocation device, such as a plug-in hard drive, SmartMedia Card (SMC), Secure Digital (SD) card, flash card, etc., equipped on the service invocation device. Further, the readable storage medium can include both internal storage units of the service invocation device and external storage devices. The readable storage medium is used to store the computer program and other programs and data required by the service invocation device. The readable storage medium can also be used to temporarily store data that has been output or will be output.
[0219] This application also provides a computer program product, which is stored in a storage medium and implements the network computing power scheduling method provided in the above embodiments when the computer program product is executed by a computer.
[0220] It should be noted that, in this document, 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 limitations, 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 that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.
[0221] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.
[0222] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.
Claims
1. A network computing power scheduling method, applied to a network management center, characterized in that, The method includes: The computation task is divided into N sub-computation tasks, where N is an integer greater than 1; Based on the execution time of each computing network resource in the multiple computing network resources corresponding to the target industrial park to complete the target sub-computation task, computing network resources with the same execution time are divided into computing network resources of the same level to obtain M-level computing network resources. The target sub-computation task is one of the N sub-computation tasks, and M is an integer greater than 1. Obtain the execution time set corresponding to each sub-computation task. The execution time set corresponding to each sub-computation task includes the execution time of each level of computing network resources to complete each sub-computation task. The execution time sets corresponding to the N sub-computing tasks are intersected. If there is an intersection element, the computing network resource level corresponding to the intersection element in the execution time set of each sub-computing task is determined, and the computing network resource of the computing network resource level is allocated to the corresponding sub-computing task.
2. The network computing power scheduling method according to claim 1, characterized in that, After performing the intersection calculation on the execution time sets corresponding to the N sub-computation tasks, the method further includes: If there is no intersection element, the execution time sets corresponding to the N sub-computation tasks are merged to obtain a merged execution time set; Determine the overall mode of the merged execution time set, and allocate resources to the N sub-computation tasks based on the overall mode.
3. The network computing power scheduling method according to claim 2, characterized in that, The resource allocation for the N sub-computation tasks based on the overall mode includes: If the set of execution times corresponding to the first sub-computing task contains the overall mode, then the computing network resource level corresponding to the overall mode in the set of execution times is determined as the computing network resource level corresponding to the first sub-computing task. If the execution time set corresponding to the first sub-computation task does not contain the overall mode, then the computing network resource level corresponding to the element in the execution time set that is closest to the overall mode is determined as the computing network resource level corresponding to the first sub-computation task. Allocate the computing network resources at the computing network resource level corresponding to each sub-computing task to the corresponding sub-computing task; The first sub-computation task is one of the N sub-computation tasks.
4. The network computing power scheduling method according to claim 1, characterized in that, The step of allocating the computing network resources at the computing network resource level to the corresponding sub-computing tasks includes: If the network resource level corresponding to the intersection element of the execution time set corresponding to at least two of the N sub-computing tasks is the same, the intersection element of the execution time set corresponding to the at least two sub-computing tasks is changed based on a preset disturbance time to obtain the changed execution time set. Based on the modified execution time set, multiple candidate allocation schemes are determined, and the candidate allocation schemes are used to indicate the computing network resources allocated to each of the at least two sub-computing tasks. The candidate allocation scheme with the smallest variance in execution time for at least two sub-computation tasks among the plurality of candidate allocation schemes is determined as the target allocation scheme; Based on the target allocation scheme, resources are allocated to the at least two sub-computing tasks, and different sub-computing tasks are allocated different computing network resources.
5. The network computing power scheduling method according to claim 1, characterized in that, The hierarchical interval between adjacent levels in the M-level computing network resources is determined based on at least one of the following: The minimum order of magnitude of the execution time interval between the multiple computing network resources; The production interval of the task requester; The production interval time is the waiting time for the task requester before executing the next production instruction.
6. The network computing power scheduling method according to claim 1, characterized in that, The step of allocating the computing network resources at the computing network resource level to the corresponding sub-computing tasks includes: From the computing network resources at the computing network resource level, select computing network resources that have not been allocated to other sub-computing tasks among the N sub-computing tasks, and use them as computing network resources allocated to the corresponding sub-computing tasks.
7. The network computing power scheduling method according to claim 1, characterized in that, Before classifying computing network resources with the same execution time into computing network resources of the same level, the method further includes: The communication time between the first computing network resource and the computing network management center, and the computing time for the first computing network resource to complete the second sub-computing task are obtained. The sum of the communication time and the computation time corresponding to the first computing network resource is determined as the execution time for the first computing network resource to complete the second sub-computation task; Wherein, the first computing network resource is one of the plurality of computing network resources; the second sub-computing task is one of the N sub-computing tasks.
8. A device for network computing power scheduling, characterized in that, The device includes: a splitting module, an acquisition module, a determination module, and an allocation module. The splitting module is used to split the computing task into N sub-computing tasks, where N is an integer greater than 1; The splitting module is used to divide computing network resources with the same execution time into computing network resources of the same level based on the execution time of each computing network resource in the multiple computing network resources corresponding to the target industrial park to obtain M-level computing network resources. The target sub-computing task is one of the N sub-computing tasks, and M is an integer greater than 1. The acquisition module is also used to acquire the execution time set corresponding to each sub-computation task, and the execution time set corresponding to each sub-computation task includes the execution time of each level of computing network resources to complete each sub-computation task; The determining module is used to perform intersection calculation on the execution time sets corresponding to the N sub-computing tasks, and if there are intersection elements, determine the computing network resource level corresponding to the intersection element in the execution time set corresponding to each sub-computing task. The allocation module is used to allocate the computing network resources at the computing network resource level to the corresponding sub-computing tasks.
9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing programs or instructions that can run on the processor, the programs or instructions being executed by the processor to implement the network computing power scheduling method as described in any one of claims 1-7.
10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions, which, when executed by a computer, implement the network computing power scheduling method as described in any one of claims 1-7.
11. A computer program product, characterized in that, The computer program product is stored in a storage medium, and when the computer program product is executed by a computer, it implements the network computing power scheduling method as described in any one of claims 1-7.
