Task scheduling method and device, storage medium, system and program product

By using an intelligent sorting model in the machine learning platform, the task execution order is reasonably arranged according to the task execution efficiency characteristics and submission time, the problems of low resource utilization and long task waiting time in the existing technology are solved, and the overall task scheduling is improved efficiency.

CN118981360APending Publication Date: 2024-11-19THE FOURTH PARADIGM BEIJING TECH CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410993560.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

The existing machine learning platform adopts fixed strategies in task scheduling, resulting in long waits for tasks with low resource utilization, low resource occupancy and short execution time, affecting the work efficiency of algorithm engineers and causing resource waste.

Method used

By obtaining the execution efficiency characteristics and submission time of the tasks to be scheduled, an intelligent sorting model is used to determine the execution order of the tasks, combining hardware requirements and submission time, reasonably balance the task execution order, and improving resource utilization and task scheduling efficiency.

Benefits of technology

It improves resource utilization, shortens task waiting time, improves the overall efficiency of task scheduling, and helps improve the work efficiency of algorithm engineers and model tuning process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118981360A_ABST
    Figure CN118981360A_ABST
Patent Text Reader

Abstract

The invention provides a task scheduling method and device, a storage medium, a system and a program product, and the method comprises the steps: obtaining a task set which comprises a plurality of to-be-scheduled tasks; the execution efficiency feature and the submission time of each to-be-scheduled task in the task set are obtained, and the execution efficiency feature is used for reflecting the efficiency of executing the corresponding task by the computer equipment; based on the execution efficiency feature and the submission time of each to-be-scheduled task in the task set, determining an execution sequence of each to-be-scheduled task in the task set by using a sorting model; the to-be-scheduled tasks in the task set are sent to a task execution system according to the execution sequence, and the task execution system comprises at least one computer device. According to the method, the execution sequence of the tasks can be reasonably balanced, the resource utilization rate is improved, the waiting time of the tasks is shortened as much as possible, and the comprehensive efficiency of task scheduling is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure generally relates to the field of computer software, and more specifically, to a task scheduling method, device, storage medium, system and program product. Background Art

[0002] With the development of artificial intelligence technology, machine learning models have been widely used, either alone or embedded in applications. In addition to model reasoning (i.e., using the model directly to complete calculations), model updates and iterations are also required. These computing tasks require a lot of computing resources.

[0003] There are machine learning platforms in related technologies, which usually use container technology to automatically implement the deployment, expansion and management of applications, and can centrally execute the reasoning and update tasks of machine learning models. However, current machine learning platforms usually adopt a fixed number of task scheduling strategies, and strictly schedule tasks according to the task submission time, which easily causes tasks that occupy less resources and have a short execution time to wait for a long time, which not only affects the work efficiency of algorithm engineers and delays the process of model tuning, but also wastes resources and leads to low resource utilization. Summary of the invention

[0004] The present disclosure provides a task scheduling method, device, storage medium, system and program product, which are used to solve at least one of the above problems.

[0005] According to one aspect of the present disclosure, a task scheduling method is provided, the task scheduling method comprising: obtaining a task set, the task set comprising a plurality of tasks to be scheduled; obtaining execution efficiency characteristics and submission time of each task to be scheduled in the task set, wherein the execution efficiency characteristics are used to reflect the efficiency of a computer device in executing corresponding tasks; based on the execution efficiency characteristics and submission time of each task to be scheduled in the task set, using a sorting model, determining the execution order of each task to be scheduled in the task set; and sending each task to be scheduled in the task set to a task execution system according to the execution order, wherein the task execution system comprises at least one computer device.

[0006] Optionally, the task scheduling method also includes: obtaining an efficiency fairness parameter, wherein the efficiency fairness parameter is used to indicate the degree of influence of the execution efficiency characteristics or the submission time on the execution order; and determining the execution order of each task to be scheduled in the task set using a sorting model based on the execution efficiency characteristics and submission time of each task to be scheduled in the task set, including: determining the execution order using the sorting model based on the efficiency fairness parameter and the execution efficiency characteristics and submission time of each task to be scheduled in the task set.

[0007] Optionally, based on the efficiency fairness parameters and the execution efficiency characteristics and submission time of each task to be scheduled in the task set, the sorting model is used to determine the execution order of each task to be scheduled in the task set, including: inputting the efficiency fairness parameters and the execution efficiency characteristics and submission time of each task to be scheduled in the task set into the sorting model to determine the execution order.

[0008] Optionally, the execution order of each task to be scheduled in the task set is determined using the sorting model based on the efficiency fairness parameter and the execution efficiency characteristics and submission time of each task to be scheduled in the task set, including: when the efficiency fairness parameter indicates that the influence of the execution efficiency characteristics on the execution order is greater than 0, the execution efficiency characteristics of each task to be scheduled in the task set are input into the sorting model to obtain a candidate execution order of each task to be scheduled in the task set; based on the efficiency fairness parameter and the submission time of each task to be scheduled in the task set, using preset rules, the candidate execution order is adjusted to obtain the execution order.

[0009] Optionally, the task scheduling method also includes: removing the tasks to be scheduled that have been sent to the task execution system from the task collection; in response to receiving new tasks to be scheduled and the execution efficiency characteristics and submission time of the new tasks to be scheduled, adding the new tasks to be scheduled to the task collection to obtain an updated task collection.

[0010] Optionally, the task scheduling method also includes: continuously acquiring computing resource data of the task execution system; judging whether there is a target task among the remaining tasks to be scheduled in the task set based on the computing resource data, wherein the target task is a task whose execution needs cannot be met by the current computing resources of the task execution system; if the judgment result is yes, removing the target task from the task set until the computing resources of the task execution system meet the execution needs of the removed target task, and re-adding the removed target task to the task set to obtain an updated task set.

[0011] Optionally, the task scheduling method also includes: for the updated task set, executing the step of determining the execution order of each task to be scheduled in the task set based on the execution efficiency characteristics and submission time of each task to be scheduled in the task set using a sorting model, and the step of sending each task to be scheduled in the task set to the task execution system in the execution order.

[0012] Optionally, the execution efficiency feature includes at least one of the following: resource occupancy data, reference running time, the resource occupancy data includes at least one of the following: central processing unit resource occupancy data, memory resource occupancy data, graphics processing unit resource occupancy data, and the reference running time is a predicted value of the time required to execute the corresponding task.

[0013] Optionally, the task scheduling method further includes: receiving the to-be-scheduled task execution data sent by the task execution system; and outputting the to-be-scheduled task execution data.

[0014] Optionally, the task scheduling method further includes: receiving feedback information from a user on the execution data of the task to be scheduled; and updating the sorting model according to the feedback information.

[0015] According to another aspect of the present disclosure, a task scheduling device is provided, which includes: a first acquisition unit, configured to acquire a task set, wherein the task set includes a plurality of tasks to be scheduled; a second acquisition unit, configured to acquire execution efficiency characteristics and submission time of each task to be scheduled in the task set, wherein the execution efficiency characteristics are used to reflect the efficiency of a computer device in executing corresponding tasks; a sorting unit, configured to determine the execution order of each task to be scheduled in the task set based on the execution efficiency characteristics and submission time of each task to be scheduled in the task set, using a sorting model; and a sending unit, configured to send each task to be scheduled in the task set to a task execution system in the execution order, wherein the task execution system includes at least one computer device.

[0016] Optionally, the task scheduling device also includes a third acquisition unit, configured to obtain an efficiency fairness parameter, wherein the efficiency fairness parameter is used to indicate the degree of influence of the execution efficiency characteristics or the submission time on the execution order; the sorting unit is also configured to determine the execution order using the sorting model based on the efficiency fairness parameter and the execution efficiency characteristics and submission time of each task to be scheduled in the task set.

[0017] Optionally, the sorting unit is further configured to input the efficiency fairness parameter and the execution efficiency characteristics and submission time of each to-be-scheduled task in the task set into the sorting model to determine the execution order.

[0018] Optionally, the sorting unit is also configured to: when the efficiency fairness parameter indicates that the degree of influence of the execution efficiency characteristics on the execution order is greater than 0, input the execution efficiency characteristics of each task to be scheduled in the task set into the sorting model to obtain a candidate execution order for each task to be scheduled in the task set; based on the efficiency fairness parameter and the submission time of each task to be scheduled in the task set, use preset rules to adjust the candidate execution order to obtain the execution order.

[0019] Optionally, the task scheduling device also includes: a removal unit, configured to remove the to-be-scheduled tasks that have been sent to the task execution system from the task set; an adding unit, configured to add the new to-be-scheduled tasks to the task set in response to receiving the new to-be-scheduled tasks and the execution efficiency characteristics and submission time of the new to-be-scheduled tasks, so as to obtain an updated task set.

[0020] Optionally, the task scheduling device also includes: a fourth acquisition unit, configured to continuously acquire computing resource data of the task execution system; a judgment unit, configured to judge whether there is a target task among the remaining tasks to be scheduled in the task set based on the computing resource data, wherein the target task is a task whose execution needs cannot be met by the current computing resources of the task execution system; an update unit, configured to, if the judgment result is yes, remove the target task from the task set until the computing resources of the task execution system meet the execution needs of the removed target task, and then re-add the removed target task to the task set to obtain an updated task set.

[0021] Optionally, the sorting unit and the sending unit are further configured to run the updated task set.

[0022] Optionally, the execution efficiency feature includes at least one of the following: resource occupancy data, reference running time, the resource occupancy data includes at least one of the following: central processing unit resource occupancy data, memory resource occupancy data, graphics processing unit resource occupancy data, and the reference running time is a predicted value of the time required to execute the corresponding task.

[0023] Optionally, the task scheduling device further includes: a receiving unit configured to receive the to-be-scheduled task execution data sent by the task execution system; and an output unit configured to output the to-be-scheduled task execution data.

[0024] Optionally, the task scheduling device further includes: a feedback unit configured to receive feedback information from a user on the execution data of the task to be scheduled; and a model updating unit configured to update the sorting model according to the feedback information.

[0025] According to another aspect of the present disclosure, a computer-readable storage medium storing instructions is provided, wherein when the instructions are executed by at least one computing device, the at least one computing device is prompted to perform the task scheduling method as described above.

[0026] According to another aspect of the present disclosure, a system is provided including at least one computing device and at least one storage device storing instructions, wherein when the instructions are executed by the at least one computing device, the at least one computing device is prompted to perform the task scheduling method as described above.

[0027] According to another aspect of the present disclosure, a computer program product is provided, comprising instructions, wherein when the instructions are executed by at least one computing device, the at least one computing device is prompted to perform the task scheduling method as described above.

[0028] According to the task scheduling method, device, storage medium, system and program product of the exemplary embodiments of the present disclosure, by obtaining the execution efficiency characteristics and submission time of the tasks to be scheduled and using an intelligent sorting model to determine the execution order of each task to be scheduled, it is possible to reasonably balance the execution order of each task to be scheduled based on the hardware requirements of the tasks to be scheduled for the computer equipment and the submission time of the tasks to be scheduled, thereby improving resource utilization and shortening the waiting time of tasks as much as possible, which helps to improve the overall efficiency of task scheduling.

[0029] Additional aspects and / or advantages of the present general inventive concept will be set forth in part in the following description and in part will be apparent from the description or may be learned through practice of the present general inventive concept. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] These and / or other aspects and advantages of the present disclosure will become clear and more easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0031] Figure 1 is a flowchart illustrating a task scheduling method according to an exemplary embodiment of the present disclosure;

[0032] Figure 2 is an architectural diagram showing a task scheduling system according to an exemplary embodiment of the present disclosure;

[0033] Figure 3 is a schematic diagram showing the workflow of a task scheduling system according to an exemplary embodiment of the present disclosure;

[0034] Figure 4 is a block diagram illustrating a task scheduling apparatus according to an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0035] The following description with reference to the accompanying drawings is provided to assist in a comprehensive understanding of the embodiments of the present invention as defined by the claims and their equivalents. Various specific details are included to assist in understanding, but these details are considered to be exemplary only. Therefore, one of ordinary skill in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. In addition, descriptions of well-known functions and structures are omitted for clarity and brevity.

[0036] It should be noted that the phrase "at least one of the items" in the present disclosure includes three types of parallel situations: "any one of the items", "a combination of any number of the items", and "all of the items". For example, "including at least one of A and B" includes the following three parallel situations: (1) including A; (2) including B; (3) including A and B. Another example is "executing at least one of step 1 and step 2" which means the following three parallel situations: (1) executing step 1; (2) executing step 2; (3) executing step 1 and step 2.

[0037] Refer to the following Figures 1 to 4 A task scheduling method, apparatus, storage medium, system, and program product according to exemplary embodiments of the present disclosure are described in detail.

[0038] Figure 1 is a flow chart showing a task scheduling method according to an exemplary embodiment of the present disclosure. The task scheduling method according to an exemplary embodiment of the present disclosure can be implemented in a computing device with sufficient computing power.

[0039] Reference Figure 1 In step S101, a task set is obtained, and the task set includes multiple tasks to be scheduled.

[0040] A task to be scheduled refers to a specific computing task that is to be scheduled to the task execution system so that the task execution system can perform the corresponding computing task. Tasks to be scheduled include, for example, model training tasks, model update tasks, and model reasoning tasks. It is usually not necessary to distinguish between different types of specific tasks when scheduling. The task set represents the whole of all tasks to be scheduled, and it is not necessary to create an additional variable in addition to multiple tasks to be scheduled.

[0041] In step S102, the execution efficiency characteristics and submission time of each task to be scheduled in the task set are obtained.

[0042] The execution efficiency feature is used to reflect the efficiency of the computer device in executing the corresponding task, and can characterize the hardware requirements of the scheduled task on the computer device.

[0043] Optionally, the execution efficiency feature includes at least one of the following: resource occupancy data, reference runtime, the resource occupancy data includes at least one of the following: central processing unit (CPU) resource occupancy data, memory resource occupancy data, and graphics processing unit (GPU) resource occupancy data, and the reference runtime is a predicted value of the duration required to execute the corresponding task. The resource occupancy data can represent the hardware resources required to execute the corresponding task to be scheduled, and the reference runtime can represent the estimated duration of execution of the corresponding task to be scheduled, thereby reflecting the efficiency of the computer device in executing the corresponding task to be scheduled from the perspectives of resources and runtime, and using it as a reference for task scheduling, which helps to give priority to tasks that occupy less resources and have a shorter runtime, so as to improve resource utilization.

[0044] It should be understood that, based on the information such as the task type, model type, data type and data volume involved, the resource occupancy data of the task to be scheduled can often be roughly determined, while the reference running time cannot often be directly obtained. As an example, if the corresponding task to be scheduled is a model update task, the previous historical training time and historical update time of the model can be used as a reference, for example, the average of the historical training time and the historical update time can be used as the reference running time; or, based on the characteristics of the corresponding task to be scheduled (for example, whether the task type is a training / update type or an inference type, and the size of the model and the input data type, etc.), the historical execution time of tasks with the same or similar characteristics can be selected as the reference running time; or, for example, a model for predicting the reference running time can be trained in advance, and the characteristics of the task to be scheduled can be input into the model to obtain the reference running time output by the model. As long as a reasonable reference running time can be obtained, the present disclosure does not limit the specific acquisition method.

[0045] The submission time is the time when the task to be scheduled is submitted. The submission time can be used to reflect the submission order of different tasks to be scheduled, and can also be used to calculate the waiting time from submission to execution of the task to be scheduled.

[0046] In step S103, based on the execution efficiency characteristics and submission time of each task to be scheduled in the task set, the execution order of each task to be scheduled in the task set is determined using a sorting model.

[0047] By obtaining the execution efficiency characteristics and submission time of the tasks to be scheduled, and using an intelligent sorting model to determine the execution order of each task to be scheduled, it is possible to combine the hardware requirements of the tasks to be scheduled for the computer equipment and the submission time of the tasks to be scheduled, and reasonably balance the execution order of each task to be scheduled, which not only improves resource utilization, but also shortens the waiting time of tasks as much as possible, which helps to improve the overall efficiency of task scheduling. As an example, when using a sorting model, the execution efficiency characteristics and submission time of each task to be scheduled can be input into the sorting model together, or the execution efficiency characteristics can be input into the sorting model, and then combined with the submission time to adjust the execution order calculated by the model. Of course, other reasonable methods can also be used, and this disclosure does not limit this.

[0048] In step S104, each task to be scheduled in the task set is sent to the task execution system according to the execution order.

[0049] The task execution system includes at least one computer device, which is used to execute the scheduled tasks and can be a physical machine or a virtual machine. As an example, the task execution system is a container cluster, such as a Kubernetes (abbreviated as K8s, an open source container orchestration system for automating the deployment, expansion and management of applications) cluster. Each computer device can form one or more nodes (Node), and each node can run multiple container groups (Pod). The container group is the smallest deployable unit in Kubernetes. A container group contains one container or multiple tightly coupled and mutually cooperative containers, representing an independent application running instance in Kubernetes. In addition, Kubernetes also uses components such as controllers and schedulers to manage and coordinate the operation of these nodes and containers. The controller focuses on the status and operation of the application to ensure that the application is managed and maintained in accordance with the user's defined expectations; the scheduler focuses on the allocation and scheduling of resources. According to the available resources of the nodes in the cluster, the container group is allocated to the appropriate node to run to achieve optimal resource utilization. If a node where a container group is located fails or becomes unavailable, the scheduler will automatically schedule the container group to another available node to continue running.

[0050] Existing container clusters usually receive a large number of computing tasks (jobs) at one time, and then the controller and the scheduler coordinate with each other to create container groups for these tasks respectively, and assign them to appropriate nodes. The present disclosure can send the tasks to be scheduled (that is, the computing task job) to the task execution system in batches according to the execution order obtained in step S103. As an example, each task to be scheduled can be sorted according to the execution order to obtain a task queue, and then the tasks in the task queue are sent to the task execution system in order. The sending quantity parameter can also be configured to send a corresponding number of tasks to be scheduled to the task execution system each time, and the sending quantity parameter is, for example, 1, 2, 5, etc., which can realize the sequential batch sending of the tasks to be scheduled. It should be understood that when the sending quantity parameter is 1, it is to send the tasks to be scheduled one by one according to the execution order obtained in step S103, and the task execution system only needs to directly execute the single task to be scheduled received each time, that is, directly create a container group for the task to be scheduled, and assign it to the appropriate node. When the sending quantity parameter is greater than 1, the task execution system can coordinate the execution order of the multiple tasks to be scheduled received by itself, or use the execution order obtained in step S103, and can also dynamically adjust the actual execution order of a single batch of tasks based on the execution order obtained in step S103 according to the actual resource usage of the task execution system, and the present disclosure does not limit this. For the case where the sending quantity parameter is greater than 1, although it may be difficult to strictly control the execution order of multiple tasks to be scheduled sent in a single batch, since the number of tasks in a single batch is often small at this time, the overall efficiency of task scheduling can still be improved to a certain extent compared to the prior art. And if the execution order obtained in step S103 is required at this time, it is also necessary to send information representing the execution order of these tasks to be scheduled, and the present disclosure does not limit the specific form of this information.

[0051] Next, the step of determining the execution order, namely step S103, is further introduced.

[0052] In step S103, the two key factors that affect the execution order are the execution efficiency characteristics and the submission time. If the execution order is determined based only on the execution efficiency characteristics, efficiency will be given priority, and tasks with less resource usage and shorter running time will be executed first, which may cause some tasks with more resource usage and longer running time to wait for a long time. If the execution order is determined based only on the submission time, fairness will be given priority, and the tasks submitted first will be executed first. By reasonably taking both into account, the overall efficiency of task scheduling can be improved.

[0053] On this basis, optionally, the task scheduling method according to the exemplary embodiment of the present disclosure further includes: obtaining an efficiency fairness parameter, the efficiency fairness parameter is used to indicate the degree of influence of the execution efficiency feature or the submission time on the execution order; step S103 includes: based on the efficiency fairness parameter and the execution efficiency feature and submission time of each task to be scheduled in the task set, using the sorting model to determine the execution order. By introducing the efficiency fairness parameter, the emphasis on efficiency and fairness can be intervened by modifying the efficiency fairness parameter, so that the "balance" of efficiency and fairness is more quantitative, which helps to improve the control accuracy for the execution order. It should be understood that since the efficiency fairness parameter represents the degree of influence, and the positive influence is considered here, the efficiency fairness parameter must be greater than or equal to 0, and its value range can be, for example, [0,1], [0,100], etc. When the efficiency fairness parameter indicates the degree of influence of the execution efficiency feature on the execution order, the efficiency fairness parameter takes a value of 0, which means that only fairness is considered, that is, only the submission time is considered when determining the execution order, that is, the execution order is determined strictly according to the submission time; the efficiency fairness parameter takes a maximum value, which means that only efficiency is considered, that is, only the execution efficiency feature is considered when determining the execution order. When the efficiency fairness parameter indicates the degree of influence of the submission time on the execution order, the meaning of the value is completely opposite. It should also be understood that a task set has an efficiency fairness parameter, but not each task to be scheduled has an efficiency fairness parameter.

[0054] Regarding the use of efficiency and fairness parameters, in some embodiments, optionally, the operation of determining the execution order of each task to be scheduled in the task set using the sorting model based on the efficiency and fairness parameters and the execution efficiency characteristics and submission time of each task to be scheduled in the task set in step S103 includes: inputting the efficiency and fairness parameters and the execution efficiency characteristics and submission time of each task to be scheduled in the task set into the sorting model to determine the execution order. By inputting all of the information such as the efficiency and fairness parameters, the execution efficiency characteristics, and the submission time into the sorting model, the sorting can be completed end-to-end by the sorting model, which can simplify the data processing outside the model.

[0055] In other embodiments, optionally, the operation of determining the execution order of each task to be scheduled in the task set using the sorting model based on the efficiency fairness parameter and the execution efficiency characteristics and submission time of each task to be scheduled in the task set in step S103 includes: when the efficiency fairness parameter indicates that the degree of influence of the execution efficiency characteristics on the execution order is greater than 0, the execution efficiency characteristics of each task to be scheduled in the task set are input into the sorting model to obtain the candidate execution order of each task to be scheduled in the task set; based on the efficiency fairness parameter and the submission time of each task to be scheduled in the task set, the candidate execution order is adjusted using preset rules to obtain the execution order. By only using the execution efficiency characteristics as the input parameters of the sorting model, the efficiency fairness parameters and the submission time are used to adjust the candidate execution order output by the sorting model, which can make the physical meaning of the sorting model clear, that is, efficiency is given priority, and it is used to achieve sorting based on the execution efficiency characteristics. And at this time, the amount of data processed by the model is relatively small, which helps to reduce the model volume and reduce the amount of model calculation.

[0056] It should be understood that when the efficiency fairness parameter indicates that the degree of influence of the execution efficiency characteristics on the execution order is equal to 0, there is no need to input the execution efficiency characteristics into the sorting model, and the execution order can be obtained by directly sorting according to the submission time of each task to be scheduled. When the efficiency fairness parameter indicates that the degree of influence of the execution efficiency characteristics on the execution order reaches the maximum value (for example, 1, 100, etc.), that is, the degree of influence of the submission time on the execution order is equal to 0, the candidate execution order output by the model is used as the execution order, and there is no need to adjust it. For other cases, that is, when the efficiency fairness parameter is greater than 0 and less than the maximum value, it is necessary to use preset rules to adjust the candidate execution order based on the efficiency fairness parameter and the submission time, and under the same conditions, the greater the influence of the submission time on the execution order, the more adjustments to the candidate execution order, reflecting the influence of the submission time.

[0057] Regarding the execution of preset rules, it should be understood that the candidate execution order is essentially a ranking of the execution efficiency of multiple tasks to be scheduled, and the higher the execution efficiency of the task, the higher the ranking. Therefore, when the ranking is adjusted by actually executing the preset rules, the multiple tasks to be scheduled can be first arranged according to the candidate execution order, and then based on the efficiency fairness parameter and the submission time, the tasks to be scheduled that are later in the candidate execution order but have an earlier submission time are moved forward to cut in line. Alternatively, the multiple tasks to be scheduled can be first arranged in the order of submission time, and then based on the efficiency fairness parameter and the execution efficiency, the tasks with a late submission time but high execution efficiency are moved forward to cut in line. These are all implementation methods of the preset rules of the present disclosure, and fall within the protection scope of the present disclosure.

[0058] As an example, still taking the efficiency fairness parameter to represent the degree of influence of the execution efficiency feature on the execution order as an example, assuming that the value range of the efficiency fairness parameter is [0,1], and the current value of the efficiency fairness parameter is X, X is greater than 0 and less than 1. The execution efficiency features of multiple tasks to be scheduled can be first input into the sorting model to obtain the score of each task to be scheduled. The higher the score, the higher the execution efficiency of the corresponding task. Then sort the multiple tasks to be scheduled according to the submission time, and then check whether the scores of the tasks are gradually reduced from the head of the queue to the back. When a task with a higher score than the previous task appears, select Y tasks with higher scores from this task and the tasks after it (assuming that the number of these tasks is Z), and move these Y tasks forward to join the queue according to the score. For example, Y is the product of X and the total number of tasks to be scheduled. If Y is greater than Z, all these Z tasks are moved forward to join the queue. Y is also the product of X and Z.

[0059] In addition, as the task scheduling proceeds, the tasks to be scheduled may change, causing the task set to change. In this regard, in some embodiments, optionally, the task scheduling method according to the exemplary embodiment of the present disclosure also includes: removing the tasks to be scheduled that have been sent to the task execution system from the task set; in response to receiving new tasks to be scheduled and the execution efficiency characteristics and submission time of the new tasks to be scheduled, adding the new tasks to be scheduled to the task set to obtain an updated task set. By continuously updating the task set as the tasks to be scheduled are sent and executed and new tasks to be scheduled are added, the current situation of the tasks to be scheduled can always be reflected, ensuring the timeliness of the information.

[0060] In other embodiments, optionally, the task scheduling method according to the exemplary embodiment of the present disclosure also includes: continuously acquiring the computing resource data of the task execution system; judging whether there is a target task in the remaining tasks to be scheduled in the task set according to the computing resource data, wherein the target task is a task whose execution needs cannot be met by the current computing resources of the task execution system; if the judgment result is yes, the target task is removed from the task set until the computing resources of the task execution system meet the execution needs of the removed target task, and the removed target task is re-added to the task set to obtain an updated task set. By continuously acquiring the computing resource data of the task execution system and identifying the target task from the remaining tasks that have not yet been sent to the task execution system, it is possible to continuously monitor whether the task execution system is competent for the remaining tasks. On this basis, by removing the target task, that is, temporarily not sending the target task to the task execution system, it is possible to ensure that the tasks sent to the task execution system are tasks that the task execution system can handle, reducing the risk of waiting and wasting time due to the inability to execute the task, and helping to improve the efficiency of task execution. In subsequent monitoring, in addition to continuing to determine whether there are new target tasks among the remaining tasks to be scheduled in the task set, it is also necessary to determine whether the current computing resources of the task execution system can meet the execution needs of the target tasks previously removed from the task set. That is, based on the latest acquired computing resource data of the task execution system, determine whether the target tasks previously removed from the task set still belong to the target tasks. If it is determined that the task no longer belongs to the target task, the task is re-added to the task set so that the task can continue to be sent to the task execution system to ensure that the task can be executed.

[0061] It should be noted that removing the target task from the task set means that the target task will not participate in the task sending temporarily. As mentioned above, the task set represents the whole of all the tasks to be scheduled, and it is not necessary to create a variable. Therefore, in actual execution, as long as the target task cannot be sent to the task execution system in a short time, it will be fine. As an example, when the tasks to be scheduled are arranged into a task queue, if the target task is determined, the target task can be adjusted to the end of the queue. Accordingly, the task can be re-added to the task set, which can be executed as moving the task forward. How to determine the specific position to move forward is as long as it is reasonable. For example, the removed task can be moved forward to the original position, such as the task that is behind the task before the task is removed can be recorded as the marked task of the task. When the removed task is re-added to the task set, it is only necessary to find the marked task of the removed task from the task sequence, and then move the removed task to the front of its marked task. If its marked task is not found, it means that its marked task has been sent, and the task can be moved to the first place in the queue. For another example, the removed task can be directly moved forward to the first place in the queue. Of course, other reasonable methods may also be used, and the present disclosure does not limit this. As an example, after determining the target task, the order of the task queue may not be changed, and a non-sending flag may be added to the determined target task, so that the task is skipped when sending the task, and other tasks that are arranged behind it are sent first. Accordingly, by re-adding the task to the task set, the non-sending flag of the task can be removed, so that it can be sent normally.

[0062] It should be understood that since the above two types of embodiments regarding task set changes are not contradictory or mutually exclusive, the two types of embodiments can exist simultaneously to update the task set from different perspectives.

[0063] Regardless of the method used to update the task set, optionally, the task scheduling method according to the exemplary embodiment of the present disclosure also includes: for the updated task set, executing step S103 and step S104, that is, re-determining the execution order, and sending each task to be scheduled to the task execution system according to the newly determined execution order, which can ensure that the remaining tasks to be scheduled are still reasonably scheduled, and help to continuously ensure the overall efficiency of task scheduling. It should be understood that in special circumstances, if the original tasks to be scheduled in the task set have been sent to the task execution system, and a new task to be scheduled is received at this time, then since there is only one task to be scheduled at this time, the task is directly ranked first, and there is no need to perform calculations to determine the execution order. It should also be understood that in the case where the task set is not updated, there is no need to reorder.

[0064] After sending the task to be scheduled to the task execution system, optionally, the task scheduling method according to the exemplary embodiment of the present disclosure further includes: receiving the execution data of the task to be scheduled sent by the task execution system; and outputting the execution data of the task to be scheduled. By receiving and outputting the execution data of the task to be scheduled, information display can be realized so that the user can understand the execution status of the task. The execution data of the task to be scheduled may, for example, include at least one of the data such as the sending time, actual execution time, completion time, and execution result of the corresponding task to be scheduled, and may also include the execution order of each task to be scheduled. As an example, the tasks to be scheduled may be displayed one by one in the order of execution, and detailed data including the actual execution time, execution result, completion time, and the like may be displayed for each task to be scheduled. For example, a list may be used, each row represents a task to be scheduled, and each task to be scheduled is sorted in the order of execution. As the execution order is adjusted, the sorting of the list may also be updated and adjusted accordingly.

[0065] Further optionally, the task scheduling method according to the exemplary embodiment of the present disclosure also includes: receiving feedback information from the user on the execution data of the scheduled task; and updating the sorting model according to the feedback information. By receiving feedback from the user and updating the sorting model accordingly, it is possible to understand whether the user is satisfied with the execution of the task, and then continuously adjust and optimize the sorting model, which helps to continue to improve the overall efficiency of task scheduling. As an example, the feedback information can simply include a satisfactory or unsatisfactory conclusion, or it can specifically include information on whether the completion time of one or some tasks is satisfactory, so as to provide more model optimization information. Of course, the feedback information can also take other forms, and the present disclosure is not limited to this.

[0066] Next, combine Figure 2 and Figure 3 , introduces an exemplary embodiment of a task scheduling system involving a machine learning platform, which is used to implement K8s-based machine learning workload scheduling.

[0067] Reference Figure 2 The task scheduling system includes a user interface layer, a scheduling management layer, a K8s cluster management layer, and a K8s cluster execution environment. The user interface layer is used to interact with users, the scheduling management layer is used to execute the task scheduling method according to the exemplary embodiment of the present disclosure, and the K8s cluster management layer and the K8s cluster execution environment constitute a task execution system according to the exemplary embodiment of the present disclosure, wherein the K8s cluster management layer corresponds to the controller and the scheduler to realize the container management of K8s, and the K8s cluster execution environment realizes the specific execution of tasks.

[0068] Reference Figure 2 and Figure 3, the user first submits multiple tasks through the user interface layer and sets the relevant parameters of each task, including the CPU, memory, and GPU resources required for the task, as well as the efficiency and fairness parameters of these tasks. The user interface layer provides a user interface for users to submit tasks. The user interface also allows users to view the current task queue and set scheduling policies, such as setting efficiency and fairness parameters, and selecting a specific sorting model. The user interface layer provides users with a way to submit tasks and set relevant parameters, which can improve the user experience.

[0069] After receiving the task submitted by the user, the system enters the task attribute analysis module of the scheduling management layer. This module performs attribute analysis on each task, including but not limited to: CPU, memory, GPU resources required for the task, reference runtime, etc. The analysis results will be passed as input to the intelligent sorting algorithm module to provide data support for intelligent sorting. The intelligent sorting algorithm module is the core module of the present disclosure. It can intelligently sort tasks based on the task attribute analysis results, using the sorting model of the artificial intelligence algorithm, and comprehensively considering the resource occupancy of the task, the reference runtime, and the efficiency fairness parameter. For example, the efficiency fairness parameter indicates the degree of influence of the submission time on the execution order, and the value range is [0,100]. When the efficiency and fairness parameter is 0, tasks that occupy less resources and have a shorter execution time are given full priority, and the scores calculated by the model are used to sort the tasks; when it is 100, tasks are strictly executed in the order of submission time; when it is other intermediate values ​​between 0 and 100, the ranking model is used to calculate the score of each task, first sort by submission time, and then find Y tasks with high scores but low rankings, and move these Y tasks forward according to the scores, where Y = (100-X)%*W, W is the total number of tasks, so as to comprehensively consider efficiency and fairness for sorting. In addition, the dynamic adjustment module is used to monitor whether the tasks have changed (for example, whether new tasks submitted by users have been received), and to re-sort the intelligent sorting algorithm module when changes are determined.

[0070] After the scheduling management layer obtains the execution order, it sends the tasks one by one to the task allocation module of the K8s cluster management layer in sequence. The task allocation module can combine the cluster resource usage monitored by the resource monitoring module to send the tasks to specific nodes in the K8s cluster execution environment, ensuring that the assigned tasks are executed immediately if resources permit. The resource monitoring module is specifically used to continuously monitor the resource usage of each node to ensure the smooth progress of the task. The cluster resource usage monitored by it will also be fed back to the dynamic adjustment module in the scheduling management layer to adjust the execution order of tasks and / or wait for resource release, thereby improving resource utilization. Specifically, the dynamic adjustment module can adjust the corresponding tasks to the end of the queue when the cluster resources are insufficient to meet the execution needs of certain tasks, and when the cluster resources can meet the corresponding tasks, the intelligent sorting algorithm can recalculate the execution order. The scheduling management layer ensures the smooth execution of tasks.

[0071] When the K8s cluster executes a task, it will create a container group for the assigned task on a specific node and execute it according to the assignment of the task assignment module. After the execution is completed, the task result and status are fed back to the user interface layer through the scheduling management layer. The user receives feedback on the execution result and status of the task through the user interface layer, and can adjust and optimize subsequent tasks based on the feedback results, for example, by optimizing the sorting model.

[0072] The system prioritizes machine learning tasks through intelligent algorithms, which can improve the performance and adaptability of the task scheduling system, improve the efficiency of task execution, and provide a more efficient and fairer task scheduling solution for the machine learning platform. Its technical points mainly include task attribute analysis, intelligent sorting algorithm, and efficiency and fairness parameter adjustment. Task attribute analysis is a comprehensive analysis of the resource occupancy data (such as CPU, memory, GPU and other resource occupancy data) and reference running time of each task to obtain accurate execution efficiency characteristics. Intelligent sorting is to use a sorting model to prioritize tasks through artificial intelligence algorithms, giving priority to tasks that occupy less resources and have a short execution time, while taking into account the fairness of tasks that occupy more resources. The efficiency and fairness parameter is a parameter that adjusts efficiency and fairness. The value range is [0,100]. 0 means that efficiency is fully considered, and 100 means that fairness is fully considered. The weights of efficiency and fairness can be flexibly adjusted to ensure that different tasks can be reasonably scheduled.

[0073] By introducing an adaptive learning mechanism, as the intelligent sorting algorithm is continuously updated, the system can continuously adjust the scheduling sorting strategy according to the actual operating conditions, which can not only ensure that the intelligent sorting algorithm is always in the best state and improve the accuracy of scheduling, but also adjust the scheduling strategy according to real-time data, improve the adaptability of scheduling, and further improve resource utilization efficiency.

[0074] By supporting dynamic adjustment of task priority during task execution, it can adapt to various emergencies, improve scheduling flexibility, and quickly respond to urgent task needs and reduce task waiting time.

[0075] The task scheduling solution of the exemplary embodiment of the present disclosure has a technical value in that it can significantly improve the task execution efficiency of the machine learning platform through an intelligent sorting algorithm. Traditional scheduling methods are prone to waste of resources and task delays, while the present disclosure can dynamically adjust task priorities, optimize resource allocation, and reduce task waiting time. Its product value lies in that it can adapt to machine learning task scenarios of different scales and complexities, provide flexible scheduling strategies, and improve the overall performance and user experience of the platform. For algorithm engineers who need frequent tuning, this system can significantly improve work efficiency and experimental iteration speed. Its effect value lies in that through intelligent scheduling, it realizes the rapid allocation and execution of tasks, reduces the average waiting time, takes into account the fairness of tasks with large resource usage and long running time, and avoids long-term task accumulation.

[0076] Figure 4 is a block diagram illustrating a task scheduling apparatus according to an exemplary embodiment of the present disclosure.

[0077] Reference Figure 4 The task scheduling device 400 includes a first acquisition unit 401, a second acquisition unit 402, a sorting unit 403, and a sending unit 404.

[0078] The first acquisition unit 401 may acquire a task set, where the task set includes a plurality of tasks to be scheduled.

[0079] The second acquisition unit 402 may acquire the execution efficiency characteristics and submission time of each to-be-scheduled task in the task set, wherein the execution efficiency characteristics are used to reflect the efficiency of the computer device in executing the corresponding task.

[0080] Optionally, the execution efficiency feature includes at least one of the following: resource occupancy data, reference runtime, the resource occupancy data includes at least one of the following: central processing unit resource occupancy data, memory resource occupancy data, graphics processing unit resource occupancy data, and the reference runtime is a predicted value of the time required to execute the corresponding task.

[0081] The sorting unit 403 may determine the execution order of each to-be-scheduled task in the task set based on the execution efficiency characteristics and submission time of each to-be-scheduled task in the task set using a sorting model.

[0082] The sending unit 404 may send each to-be-scheduled task in the task set to a task execution system according to the execution order, wherein the task execution system includes at least one computer device.

[0083] Optionally, the task scheduling device 400 also includes a third acquisition unit (not shown in the figure), which can acquire efficiency fairness parameters, which are used to indicate the degree of influence of execution efficiency characteristics or submission time on the execution order; the sorting unit 403 can also use a sorting model to determine the execution order based on the efficiency fairness parameters and the execution efficiency characteristics and submission time of each task to be scheduled in the task set.

[0084] Optionally, the sorting unit 403 may also input the efficiency fairness parameter and the execution efficiency characteristics and submission time of each to-be-scheduled task in the task set into the sorting model to determine the execution order.

[0085] Optionally, the sorting unit 403 may also: when the efficiency fairness parameter indicates that the influence of the execution efficiency characteristics on the execution order is greater than 0, input the execution efficiency characteristics of each task to be scheduled in the task set into the sorting model to obtain the candidate execution order of each task to be scheduled in the task set; based on the efficiency fairness parameter and the submission time of each task to be scheduled in the task set, use preset rules to adjust the candidate execution order to obtain the execution order.

[0086] Optionally, the task scheduling device 400 further includes a removal unit (not shown in the figure) and an addition unit (not shown in the figure). The removal unit can remove the to-be-scheduled tasks that have been sent to the task execution system from the task set; the addition unit can add the new to-be-scheduled tasks to the task set in response to receiving the new to-be-scheduled tasks and the execution efficiency characteristics and submission time of the new to-be-scheduled tasks, thereby obtaining an updated task set.

[0087] Optionally, the task scheduling device 400 further includes a fourth acquisition unit (not shown in the figure), a judgment unit (not shown in the figure) and an update unit (not shown in the figure). The fourth acquisition unit can continuously acquire the computing resource data of the task execution system; the judgment unit can judge whether there is a target task among the remaining tasks to be scheduled in the task set based on the computing resource data, wherein the target task is a task whose execution needs cannot be met by the current computing resources of the task execution system; the update unit can remove the target task from the task set if the judgment result is yes, until the computing resources of the task execution system meet the execution needs of the removed target task, and then re-add the removed target task to the task set to obtain an updated task set.

[0088] Optionally, the sorting unit 403 and the sending unit 404 may also operate on an updated task set.

[0089] Optionally, the task scheduling device 400 further includes a receiving unit (not shown in the figure) and an output unit (not shown in the figure). The receiving unit can receive the to-be-scheduled task execution data sent by the task execution system; and the output unit can output the to-be-scheduled task execution data.

[0090] Optionally, the task scheduling device 400 further includes a feedback unit (not shown in the figure) and a model updating unit (not shown in the figure). The feedback unit can receive user feedback information on the execution data of the scheduled task; and the model updating unit can update the sorting model according to the feedback information.

[0091] The above has been referred to Figures 1 to 4 A task scheduling method and apparatus according to exemplary embodiments of the present disclosure are described.

[0092] Figure 4 Each unit in the task scheduling device shown may be configured as software, hardware, firmware, or any combination of the above items to perform a specific function. For example, each unit may correspond to a dedicated integrated circuit, or may correspond to a pure software code, or may correspond to a module that combines software and hardware. In addition, one or more functions implemented by each unit may also be uniformly executed by components in a physical entity device (e.g., a processor, a client, or a server, etc.).

[0093] In addition, refer to Figure 1 The described task scheduling method can be implemented by a program (or instruction) recorded on a computer-readable storage medium. For example, according to an exemplary embodiment of the present disclosure, a computer-readable storage medium storing instructions can be provided, wherein when the instructions are executed by at least one computing device, the at least one computing device is prompted to perform the task scheduling method according to the present disclosure.

[0094] The computer program in the computer-readable storage medium can be run in an environment deployed in a computer device such as a client, a host, an agent device, a server, etc. It should be noted that the computer program can also be used to perform additional steps in addition to the above steps or perform more specific processing when performing the above steps. The contents of these additional steps and further processing have been described in reference to Figure 1 It is mentioned in the description of the related method, so it will not be repeated here to avoid repetition.

[0095] It should be noted that the various units in the task scheduling device according to the exemplary embodiment of the present disclosure can completely rely on the operation of the computer program to realize the corresponding functions, that is, each unit corresponds to each step in the functional architecture of the computer program, so that the entire system is called through a special software package (for example, lib library) to realize the corresponding function.

[0096] on the other hand, Figure 4The various units shown may also be implemented by hardware, software, firmware, middleware, microcode, or any combination thereof. When implemented by software, firmware, middleware, or microcode, the program code or code segment for performing the corresponding operation may be stored in a computer-readable medium such as a storage medium, so that the processor may perform the corresponding operation by reading and running the corresponding program code or code segment.

[0097] For example, an exemplary embodiment of the present disclosure may also be implemented as a computing device, which includes a storage component and a processor, wherein a computer executable instruction set is stored in the storage component, and when the computer executable instruction set is executed by the processor, a task scheduling method according to an exemplary embodiment of the present disclosure is executed.

[0098] Specifically, the computing device can be deployed in a server or client, or can be deployed on a node device in a distributed network environment. In addition, the computing device can be a PC, a tablet device, a personal digital assistant, a smart phone, a web application, or other device capable of executing the above instruction set.

[0099] Here, the computing device is not necessarily a single computing device, but may also be any device or circuit collection that can execute the above instructions (or instruction sets) individually or jointly. The computing device may also be part of an integrated control system or system manager, or may be configured as a portable electronic device that is interconnected with a local or remote (e.g., via wireless transmission) interface.

[0100] In a computing device, a processor may include a central processing unit (CPU), a graphics processing unit (GPU), a programmable logic device, a dedicated processor system, a microcontroller, or a microprocessor. By way of example and not limitation, a processor may also include an analog processor, a digital processor, a microprocessor, a multi-core processor, a processor array, a network processor, etc.

[0101] According to the exemplary embodiment of the present disclosure, some operations described in the task scheduling method can be implemented by software, some operations can be implemented by hardware, and in addition, these operations can also be implemented by a combination of software and hardware.

[0102] The processor may execute instructions or codes stored in one of the storage components, which may also store data. Instructions and data may also be sent and received over a network via a network interface device, which may employ any known transmission protocol.

[0103] The storage component may be integrated with the processor, for example, RAM or flash memory is arranged within an integrated circuit microprocessor, etc. In addition, the storage component may include a separate device, such as an external disk drive, a storage array, or any other storage device that can be used by a database system. The storage component and the processor may be operatively coupled, or may communicate with each other, such as through an I / O port, a network connection, etc., so that the processor can read files stored in the storage component.

[0104] In addition, the computing device may also include a video display (such as a liquid crystal display) and a user interaction interface (such as a keyboard, a mouse, a touch input device, etc.) All components of the computing device may be connected to each other via a bus and / or a network.

[0105] The task scheduling method according to the exemplary embodiment of the present disclosure can be described as various interconnected or coupled functional blocks or functional diagrams. However, these functional blocks or functional diagrams can be equally integrated into a single logical device or operate according to non-exact boundaries.

[0106] Therefore, refer to Figure 1 The described task scheduling method may be implemented by a system including at least one computing device and at least one storage device storing instructions.

[0107] According to an exemplary embodiment of the present disclosure, at least one computing device is a computing device for executing a task scheduling method according to an exemplary embodiment of the present disclosure, and a computer executable instruction set is stored in the storage device. When the computer executable instruction set is executed by the at least one computing device, the execution reference Figure 1 The described task scheduling method.

[0108] According to an exemplary embodiment of the present disclosure, a computer program product may also be provided, including instructions, which, when executed by at least one computing device, prompt the at least one computing device to perform the task scheduling method according to the present disclosure.

[0109] The above describes various exemplary embodiments of the present disclosure, and it should be understood that the above description is only exemplary and not exhaustive, and the present disclosure is not limited to the disclosed exemplary embodiments. Without departing from the scope and spirit of the present disclosure, many modifications and changes are obvious to those of ordinary skill in the art. Therefore, the scope of protection of the present disclosure should be based on the scope of the claims.

Claims

1. A task scheduling method, characterized in that: The task scheduling method comprises: Obtain a task set, wherein the task set includes a plurality of tasks to be scheduled; Obtaining the execution efficiency characteristics and submission time of each to-be-scheduled task in the task set, wherein the execution efficiency characteristics are used to reflect the efficiency of the computer device in executing the corresponding task; Based on the execution efficiency characteristics and submission time of each task to be scheduled in the task set, using a sorting model, determining the execution order of each task to be scheduled in the task set; Each task to be scheduled in the task set is sent to a task execution system according to the execution order, wherein the task execution system includes at least one computer device.

2. The task scheduling method according to claim 1, characterized in that: The task scheduling method further includes: Acquire an efficiency fairness parameter, where the efficiency fairness parameter is used to indicate the degree of influence of the execution efficiency characteristic or the submission time on the execution order; The step of determining the execution order of each task to be scheduled in the task set by using a sorting model based on the execution efficiency characteristics and submission time of each task to be scheduled in the task set includes: The execution order is determined using the sorting model based on the efficiency fairness parameter and the execution efficiency characteristics and submission time of each to-be-scheduled task in the task set.

3. The task scheduling method according to claim 2, characterized in that: The determining the execution order of each task to be scheduled in the task set using the sorting model based on the efficiency fairness parameter and the execution efficiency characteristics and submission time of each task to be scheduled in the task set includes: The efficiency fairness parameter and the execution efficiency characteristics and submission time of each to-be-scheduled task in the task set are input into the sorting model to determine the execution order.

4. The task scheduling method according to claim 2, characterized in that: The determining the execution order of each task to be scheduled in the task set using the sorting model based on the efficiency fairness parameter and the execution efficiency characteristics and submission time of each task to be scheduled in the task set includes: When the efficiency fairness parameter indicates that the degree of influence of the execution efficiency feature on the execution order is greater than 0, the execution efficiency feature of each task to be scheduled in the task set is input into the sorting model to obtain a candidate execution order for each task to be scheduled in the task set; Based on the efficiency fairness parameter and the submission time of each to-be-scheduled task in the task set, the candidate execution order is adjusted using a preset rule to obtain the execution order.

5. The task scheduling method according to claim 1, characterized in that: The task scheduling method further includes: Removing the to-be-scheduled tasks that have been sent to the task execution system from the task set; In response to receiving a new task to be scheduled and the execution efficiency characteristics and submission time of the new task to be scheduled, the new task to be scheduled is added to the task set to obtain an updated task set.

6. The task scheduling method according to claim 1, characterized in that: The task scheduling method further includes: Continuously acquiring computing resource data of the task execution system; According to the computing resource data, determining whether there is a target task among the remaining tasks to be scheduled in the task set, wherein the target task is a task whose execution needs cannot be met by the current computing resources of the task execution system; If the judgment result is yes, the target task is removed from the task set until the computing resources of the task execution system meet the execution needs of the removed target task, and the removed target task is re-added to the task set to obtain an updated task set.

7. A task scheduling device, characterized in that: The task scheduling device comprises: A first acquisition unit is configured to acquire a task set, wherein the task set includes a plurality of tasks to be scheduled; A second acquisition unit is configured to acquire an execution efficiency feature and a submission time of each to-be-scheduled task in the task set, wherein the execution efficiency feature is used to reflect the efficiency of the computer device in executing the corresponding task; A sorting unit is configured to determine the execution order of each to-be-scheduled task in the task set using a sorting model based on the execution efficiency characteristics and submission time of each to-be-scheduled task in the task set; The sending unit is configured to send each to-be-scheduled task in the task set to a task execution system according to the execution order, wherein the task execution system includes at least one computer device.

8. A computer-readable storage medium storing instructions, characterized in that: When the instructions are executed by at least one computing device, the at least one computing device is prompted to execute the task scheduling method according to any one of claims 1 to 6.

9. A system comprising at least one computing device and at least one storage device storing instructions, characterized in that: When the instructions are executed by the at least one computing device, the at least one computing device is prompted to perform the task scheduling method according to any one of claims 1 to 6.

10. A computer program product comprising instructions, characterized in that When the instructions are executed by at least one computing device, the at least one computing device is prompted to perform the task scheduling method according to any one of claims 1 to 6.

Citation Information

Cited By

  • Control method, system and device and storage medium

    CN119861591A

  • Task scheduling method and device for fully fair scheduler and electronic equipment

    CN121455648A