Computing power task scheduling method, device and equipment and storage medium

Abstract

The invention discloses a computing power task scheduling method and device, equipment and a storage medium. The method comprises the following steps: determining a candidate job calculation graph through a branch service center according to a job dependency relationship of at least one to-be-worked sub-task in a to-be-scheduled computing power task; through a branch service center, based on a job time prediction model, according to the resource demand data and job scheduling data of the to-be-worked sub-task, determining predicted job time; correcting the candidate job calculation graph through the branch service center according to the job type, the predicted job time and the job dependency relationship of the to-be-worked sub-task to obtain a target job calculation graph; determining a target allocation resource through the branch service center according to the job type, the resource demand data and the current allocatable resource of the multi-level computing power scheduling system; and scheduling the to-be-operated sub-task through the branch service center based on the target allocation resource and the target operation calculation graph. According to the technical scheme, the reliability and timeliness of task response can be improved.

Description

Technical Field

[0001] This application relates to the field of computer technology and can be applied to the field of financial technology, especially the field of big data technology. Specifically, it relates to a computing power task scheduling method, apparatus, device and storage medium. Background Technology

[0002] In the process of digital and intelligent transformation of banks, in order to fully utilize the massive audio and video data of branches to monitor abnormal behavior of branch employees, banks have launched an intelligent audio and video monitoring and analysis system. Based on technologies such as artificial intelligence, machine vision, speech recognition, and natural language processing analysis, the system routinely and seamlessly monitors and analyzes abnormal and irregular behaviors of bank branch staff during business transactions and services, and feeds the analysis results back to the risk control and early warning center for response and processing, thereby achieving comprehensive risk review and rapid screening of potential risks at branches.

[0003] In the existing solution, the computing resources at the branch end exist in an isolated form, scattered and independent, lacking control and effective adjustment of global computing resources; in terms of computing tasks, it relies too much on the computing models of branch servers or data centers, and too many monitoring and analysis tasks and data processing requests result in high consumption of computing resources and low computing efficiency; in terms of computing power allocation, although it is convenient to manage and control the centralized computing power on the branch server side to execute tasks, it cannot execute time-sensitive response tasks in a timely manner. Summary of the Invention

[0004] This application provides a computing task scheduling method, apparatus, device, and storage medium to improve the reliability and timeliness of task response.

[0005] According to one aspect of this application, a computing power task scheduling method is provided, which is executed by a multi-level computing power scheduling system; the multi-level computing power scheduling system includes an equipment operation center, a branch processing center, and a branch service center; the branch processing center is communicatively connected to the branch service center and the equipment operation center respectively; the method includes:

[0006] The branch service center determines the computing power tasks to be scheduled based on the target device data received from the branch processing center.

[0007] The branch service center determines a candidate job computation graph for the computing power task to be scheduled based on the job dependencies of at least one subtask in the task to be scheduled; wherein, the candidate job computation graph is a directed acyclic graph used to characterize the job order of the subtasks to be scheduled.

[0008] The branch service center determines the predicted job time of the sub-task based on the job time prediction model and the resource requirement data and job scheduling data of the sub-task to be done.

[0009] The branch service center corrects the candidate job computation graph based on the job type of the sub-task to be scheduled, the predicted job time, and the job dependency relationship to obtain the target job computation graph of the computing power task to be scheduled.

[0010] The branch service center determines the target allocation resources for the subtask to be performed based on the job type, the resource requirement data, and the currently available resources of the multi-level computing power scheduling system; wherein, the branch service center includes multiple computing devices; and the currently available resources are the computing power resources of each of the multiple computing devices.

[0011] The branch service center schedules the pending subtasks in the pending computing power tasks based on the target allocated resources and the target job computation graph.

[0012] According to another aspect of this application, a computing power task scheduling device is provided, which is configured in a multi-level computing power scheduling system; the multi-level computing power scheduling system includes an equipment operation center, a branch processing center, and a branch service center; the branch processing center is communicatively connected to the branch service center and the equipment operation center respectively; the device includes:

[0013] The task determination module is used to determine the computing power tasks to be scheduled based on the target device data received from the branch processing center by the branch service center.

[0014] The computation graph determination module is used to determine the candidate job computation graph of the computing power task to be scheduled by the branch service center based on the job dependency relationship of at least one subtask to be scheduled in the computing power task to be scheduled; wherein, the candidate job computation graph is a directed acyclic graph used to characterize the job order of the subtasks to be scheduled.

[0015] The time prediction module is used to determine the predicted job time of the sub-task to be done based on the resource requirement data and job scheduling data of the sub-task to be done by the branch service center using the job time prediction model.

[0016] The computation graph correction module is used to correct the candidate job computation graph through the branch service center based on the job type of the sub-task to be scheduled, the predicted job time, and the job dependency relationship, so as to obtain the target job computation graph of the computing power task to be scheduled.

[0017] The resource determination module is used to determine the target allocation resources for the subtask to be performed by the branch service center based on the job type, the resource requirement data, and the currently available resources of the multi-level computing power scheduling system; wherein, the branch service center includes multiple computing devices; and the currently available resources are the computing power resources of each of the multiple computing devices.

[0018] The task scheduling module is used to schedule the pending subtasks in the computing power task to be scheduled by the branch service center based on the target allocated resources and the target job computation graph.

[0019] According to another aspect of this application, an electronic device is provided, the electronic device comprising:

[0020] One or more processors;

[0021] Memory, used to store one or more programs;

[0022] When the one or more programs are executed by the one or more processors, the one or more processors implement any of the computing power task scheduling methods provided in the embodiments of this application.

[0023] According to another aspect of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements any of the computing power task scheduling methods provided in the embodiments of this application.

[0024] According to another aspect of this application, a computer program product is provided, including a computer program that, when executed by a processor, implements any of the computing power task scheduling methods provided in the embodiments of this application.

[0025] This application involves the following steps: First, the branch service center determines the computing power tasks to be scheduled based on target device data received from the branch processing center. Second, the branch service center determines a candidate job computation graph for the computing power tasks to be scheduled based on the job dependencies of at least one subtask within the task. The candidate job computation graph is a directed acyclic graph representing the job order of the subtasks. Third, the branch service center determines the predicted job time for the subtasks based on a job time prediction model, considering the resource requirements and job scheduling data. Fourth, the branch service center modifies the candidate job computation graph based on the job type, predicted job time, and job dependencies of the subtasks to obtain the target job computation graph for the computing power tasks to be scheduled. Fifth, the branch service center determines the target allocation resources for the subtasks based on the job type, resource requirements, and the currently available resources of the multi-level computing power scheduling system. The branch service center includes multiple computing devices, and the currently available resources are the computing power resources of each device. Finally, the branch service center schedules the subtasks within the computing power tasks to be scheduled based on the target allocation resources and the target job computation graph. The above technical solution, by introducing a job computation graph and resource allocation to schedule computing tasks, can effectively improve the reliability and timeliness of task scheduling. Attached Figure Description

[0026] Figure 1a This is a flowchart of a computing power task scheduling method provided in Embodiment 1 of this application;

[0027] Figure 1b This is a schematic diagram of the candidate job calculation graph provided in Embodiment 1 of this application;

[0028] Figure 1c This is an architecture diagram of a multi-level computing power scheduling system provided in Embodiment 1 of this application;

[0029] Figure 2 This is a flowchart of a computing power task scheduling method according to Embodiment 2 of this application;

[0030] Figure 3 This is a schematic diagram of a computing power task scheduling device according to Embodiment 3 of this application;

[0031] Figure 4 This is a schematic diagram of the structure of an electronic device that implements the computing power task scheduling method of Embodiment 4 of this application. Detailed Implementation

[0032] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0033] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0034] Furthermore, it should be noted that the collection, storage, use, processing, transmission, provision, and disclosure of target device data and related data such as computing power tasks to be scheduled in the technical solution of this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0035] Example 1

[0036] Figure 1a This is a flowchart of a computing power task scheduling method according to Embodiment 1 of this application. This embodiment is applicable to the scheduling and resource allocation of computing power tasks in bank branches. It can be executed by a computing power task scheduling device, which can be implemented in hardware and / or software. This device can be configured in computer equipment, such as a multi-level computing power scheduling system. The multi-level computing power scheduling system includes an equipment operation center, a branch processing center, and a branch service center. The branch processing center is communicatively connected to both the branch service center and the equipment operation center. Figure 1a As shown, the method includes:

[0037] S110. The branch service center determines the computing power tasks to be scheduled based on the target device data received from the branch processing center.

[0038] In this embodiment, the multi-level computing power scheduling system refers to a system used for task scheduling across multiple levels (such as different processing centers, computing devices, etc.). A branch service center refers to a facility or system within a bank or financial institution, responsible for receiving data from the branch processing center and executing computational and scheduling tasks as needed; it corresponds to the data processing, storage, and analysis service centers of first- and second-level branches; the center relies on branch server resources to process branch-end computing power tasks and generate monitoring, analysis, and visualization results, sending back corresponding updated data to the branches and devices for parameter updates. The processing results are cached and notified to the early warning system. A branch processing center refers to the computing and data processing facilities within each bank branch or sub-branch, corresponding to the relevant edge device set, supporting local area transmission network, and data storage and caching modules of the bank's business outlets; it collects feature data processed by intelligent devices and batch data transmitted by general edge devices; the processing center's supporting connection consists of a local area network environment isolated by switches and edge computing server ports, while also connecting to some branch center server ports, dynamically allocated by the cloud computing scheduling center using scheduling strategies. Target device data refers to the batch data and some processing results received from the branches. The computing tasks to be scheduled refer to computing tasks that have not yet been allocated specific resources and executed after the branch service center has received the data; such computing tasks to be scheduled may include at least one of financial risk monitoring and visualization and intelligent analysis of customer behavior.

[0039] In one optional implementation, candidate device data is obtained by preprocessing data acquired from at least one terminal device through the device operation center and then transmitted to the branch processing center; target device data is obtained by configuring network parameters of the candidate device data through the branch processing center and then transmitted to the branch service center.

[0040] In this embodiment, the candidate device data is a set of data obtained from the terminal device and preliminarily processed by the device operation center; specifically, it refers to the valid data collected and processed by all edge devices of the current network point.

[0041] Specifically, at the equipment operation center level, data processing mainly revolves around the data input from various edge device source terminals. The goal is to judge and filter valid data based on data characteristics and preset algorithms. First, corresponding edge devices are deployed and corresponding equipment operation centers are built according to the actual situation of each branch. For larger branches, AI-enabled smart devices are centrally deployed, combined with some general edge devices for data collection. The smart devices execute tasks requiring device-side computing power, such as keyframe extraction and abnormal behavior detection, generating partial task results and processed data. The task results are promptly synchronized to the risk warning center, and the processed data is transmitted upwards to the branch processing center for further processing. For smaller branches, only general edge devices are deployed for data collection to reduce costs. The collected data is collected in batches and transmitted back to the branch processing center, where the processing center server calls the model for unified processing. This approach improves the quality of transmitted data while providing timely warnings through preliminary data processing and analysis. At the branch processing center level, all valid data collected and processed by the edge devices at the current branch are aggregated, and the computing power of the branch's edge computing server is used to process the data and execute tasks through the processing model. In this process, data from different equipment operation centers is forwarded to the current branch processing center through the core switch port of the local area network; at the same time, the edge computing server receives load balancing scheduling from the cloud computing scheduling center, and dynamically allocates network ports, bandwidth, computing power and other resources in real time according to business priorities; after data integration by the branch processing center, the packaged data is transmitted to the branch level for final processing.

[0042] It is understandable that the preprocessing of data by the equipment operation center is a computing power task of the equipment operation center; the configuration of network parameters of candidate equipment data by the network processing center is a computing power task of the network processing center; this application divides the task into three types according to the execution level of computing power tasks, and performs phased processing in the three centers of the transmission network respectively, which reduces the time loss caused by network transmission and centralized training, and improves the response timeliness of the overall task warning.

[0043] S120. The branch service center determines the candidate job computation graph of the computing power task to be scheduled based on the job dependency relationship of at least one subtask to be scheduled in the computing power task to be scheduled.

[0044] In this embodiment, a pending subtask refers to a single subtask among the pending computing power tasks that has not yet been completed or is awaiting scheduling. This pending subtask may include at least one of the following: real-time anti-fraud analysis, credit risk modeling, market risk stress testing, customer profiling enhancement, intelligent recommendation of financial products, customer behavior analysis, and account risk assessment. Job dependency refers to the relationship where certain tasks must wait for other tasks to complete before they can be executed. During scheduling, job dependencies determine the execution order of tasks, ensuring that tasks are performed in the correct sequence. The candidate job computation graph is a directed acyclic graph used to represent the job order of pending subtasks. This computation graph clearly shows the execution order between different subtasks, ensuring that task scheduling is performed while satisfying dependencies.

[0045] Specifically, the branch service center divides the task into task blocks according to the job dependency relationship of at least one subtask in the task to be scheduled, based on the principle of minimum partitioning, to obtain the candidate job computation graph of the task to be scheduled.

[0046] For example, see Figure 1b Operation B depends on the start of operation A, while operation C depends on the end of operation A. Based on the dependency and execution relationship, A and B are divided into the smallest job blocks and executed in the same executor session. Downstream operation C builds job blocks with other dependent operations. Assuming that operation D consumes a lot of I / O resources and executes slowly, while another set of dependent operations {EFG} completes execution, the resources occupied by {EFG} are released and the computation graph is effectively pruned to optimize the resource allocation structure.

[0047] S130. The branch service center determines the predicted operation time of the sub-task based on the operation time prediction model and the resource requirement data and operation scheduling data of the sub-task to be completed.

[0048] In this embodiment, the job time prediction model refers to a model that predicts the time required to complete a task or subtask based on historical data, job type, resource requirements, and other factors. Resource requirement data refers to the data on the computing resources (such as CPU, memory, storage, GPU, etc.) required by each task or subtask during execution. Job scheduling data refers to the data used in the task scheduling process to describe and manage information such as the execution order, resource allocation, priority, and time requirements of tasks (or subtasks). Predicted job time refers to the predicted time required to complete a task or subtask.

[0049] It should be noted that the job time prediction model can be trained on a multiple linear regression model based on the execution data of each job subtask of historical computing power tasks.

[0050] S140. The branch service center corrects the candidate job computation graph based on the job type, predicted job time and job dependency of the sub-task to be scheduled, and obtains the target job computation graph of the computing power task to be scheduled.

[0051] In this embodiment, job type refers to the nature or category of the task, such as computationally intensive tasks, data analysis tasks, and file processing tasks. Different types of tasks require different resource scheduling strategies and execution priorities. The target job computation graph is the final computation graph after adjustments based on the candidate job computation graph according to factors such as task type and resource requirements. It is the final task scheduling graph, which includes the order of jobs, dependencies, and resource allocation to ensure efficient task execution.

[0052] S150. Based on the job type, resource requirement data, and the currently available resources of the multi-level computing power scheduling system, the branch service center determines the target allocation resources for the sub-tasks to be performed.

[0053] In this embodiment, the branch service center includes multiple computing devices. The currently allocable resources refer to the computing resources that the system can use to allocate to tasks, including the computing power resources of all computing devices. Specifically, these resources refer to the individual computing power resources of each of the multiple computing devices. Target allocation resources refer to the specific hardware resources (such as GPU, CPU, memory, etc.) allocated to each subtask based on the task type, resource requirements, and available computing resources. These resources will be used to execute the various subtasks of the scheduled computing power task.

[0054] For example, based on the job type, the currently allocable resources of the multi-level computing power scheduling system are prioritized to allocate to higher-priority subtasks with jobs, and the remaining resources are then allocated to other subtasks waiting to be done, thereby determining the target allocation resources for each subtask waiting to be done.

[0055] S160. Based on the target resource allocation and target job calculation graph, the branch service center schedules the pending job subtasks in the computing power task to be scheduled.

[0056] For example, see Figure 1c Based on the target allocation resources and target job computation graph, the branch service center distributes the sub-tasks to be scheduled in the computing power tasks to the central GPU server, so that the sub-tasks to be scheduled can be executed by the central GPU server.

[0057] In one alternative implementation, see Figure 1c The multi-level computing power scheduling system also includes a cloud computing scheduling center; the cloud computing scheduling center is connected to the branch service center, the outlet processing center and the equipment operation center respectively.

[0058] In this embodiment, the cloud computing scheduling center refers to the control center responsible for dynamically allocating network bandwidth, servers, and other resources according to business, permissions, and priorities. It is an important control center for maintaining the overall network framework and resource load balance. At the same time, it performs unified monitoring of the allocated resources to ensure fault security and transmission stability.

[0059] For example, the cloud computing scheduling center can be responsible for global resource load balancing across devices, branches, and sub-branches; involving network and computing resource registration, parameter configuration, message passing, trend evaluation, etc.; and expanding or reclaiming corresponding resources according to the preset load balancing strategy and the actual usage of different devices, branches, and sub-branches.

[0060] Furthermore, after the cloud computing dispatch center schedules the computing power tasks to be scheduled at the branch service center, it monitors the execution time of the computing power tasks in real time to obtain the real-time execution time of the computing power tasks to be scheduled. If the real-time execution time meets the execution time condition, the cloud computing dispatch center determines the current resource gap of the computing power tasks to be scheduled based on the current available resources of the multi-level computing power scheduling system and the current resource requirements of the computing power tasks to be scheduled. The cloud computing dispatch center determines the resource adjustment method for the computing power tasks to be scheduled based on the current resource gap, the total resources of the multi-level computing power scheduling system, the current resource occupancy rate, and the resource warning threshold.

[0061] In this embodiment, real-time execution time refers to the task running time monitored by the system in real time during task execution. Execution time conditions are pre-set based on actual conditions or experience; for example, the execution time condition can be that the execution time is less than or equal to a preset time threshold. Current required resources refer to the resources required by the current task. Current resource gap refers to the insufficient availability of resources to meet the task's requirements. Total resources refer to the total computing power resources of the branch service center. Current resource occupancy rate refers to the proportion of computing power resources currently used by the task. Resource warning threshold refers to a preset resource percentage. Resource adjustment methods refer to determining how to optimize or reallocate computing resources based on real-time monitoring data (such as resource gaps, remaining available resources, etc.); adjustment methods may include strategies such as expanding resources, rescheduling tasks, and optimizing existing resources, aiming to improve the system's resource utilization and task execution efficiency.

[0062] Specifically, the cloud computing scheduling center determines the resource adjustment method for the computing tasks to be scheduled based on the current resource gap, the total resources of the multi-level computing power scheduling system, the current resource occupancy rate, and the resource warning threshold. This can be achieved by: the cloud computing scheduling center subtracting the current resource occupancy rate and the resource warning threshold of the multi-level computing power scheduling system to determine the remaining resource ratio of the multi-level computing power scheduling system; the cloud computing scheduling center determining the remaining available resources of the multi-level computing power scheduling system based on the remaining resource ratio and the total resources; and the cloud computing scheduling center determining the resource adjustment method for the computing tasks to be scheduled based on the remaining available resources and the current resource gap.

[0063] In this embodiment, the percentage of remaining resources refers to the percentage of remaining available resources without affecting the normal operation of the system. Remaining available resources refer to unused computing resources in the system; these resources can be used to meet the needs of current or future tasks.

[0064] For example, taking CPU computing power resources as an example, the evaluation of resource adjustment methods can be achieved through the following formula:

[0065]

[0066] in, This refers to the current task's shortfall in CPU computing resources. This refers to the CPU resource warning threshold. This refers to the utilization rate of CPU computing resources, i.e., the current resource utilization rate. It refers to the total amount of CPU computing power resources that can be allocated. This refers to the percentage of remaining resources. This refers to the remaining available resources.

[0067] Specifically, when the remaining available resources exceed the current resource gap, a resource expansion adjustment is requested. When multiple resource gaps exist, the resource adjustment process allocates and tests resources based on the highest and most scarce resource currently in use, typically allocating 30% of that resource gap to each task. After allocating the resource gaps for all tasks proportionally, execution tests are conducted to verify the effectiveness of the computing power allocation strategy for task execution. For tasks that achieve the expected results, the full allocation is performed based on the gap. For tasks that do not achieve the expected results, cluster resource expansion is implemented or some tasks are suspended, based on the overall computing power resource scarcity, to promptly release some occupied resources.

[0068] Furthermore, see Figure 1c After determining the resource adjustment method, the branch service center obtains the resource adjustment method from the cloud computing scheduling center through the security control module, and adjusts the computing resources accordingly.

[0069] Understandably, by conducting small-scale allocation tests, tasks can be effectively evaluated, and computing power gaps can be addressed, thereby improving the overall resource utilization efficiency of the scheduling center.

[0070] In another optional implementation, if the real-time execution time meets the execution time condition, the task status of the task is changed to timeout, the relevant resources of the task are reclaimed, and notifications and warnings are issued.

[0071] In another optional implementation, the cloud computing scheduling center can simultaneously monitor and schedule the computing resources of the branch service center and the equipment operation center in parallel, while making real-time adjustments to the computing resources of the branch service center.

[0072] For example, the scheduling strategies for the network portion of a network processing center can include priority scheduling strategies, network-level scheduling strategies, and busy / off-peak scheduling strategies. Priority strategies suspend, wait for, and schedule pending transmission tasks according to priority queues. In practice, first-come-first-served (FFS), shortest-job-first (SJF), and multi-level queue scheduling are combined to achieve efficient scheduling of different services and task types. Busy / off-peak scheduling prioritizes allocating bandwidth and server resources to basic services during peak periods, and allocates bandwidth for monitoring and analyzing data transmission during idle periods. Network-level scheduling performs load balancing at different network layers during transmission, dynamically scheduling different transmission tasks based on port and application request. For example, based on different transmission protocols, load balancing optimization is performed at the network transport layer according to the characteristics of transmitted data and packets, achieving transport layer routing combinations and end-to-end transmission control. At the application layer, based on different HTTP request protocols and header context, requests are sent to different server ports according to service type. In practice, the above network scheduling strategies can be combined and evaluated using a score-based approach. This score serves as the basis for network scheduling and resource allocation, including ports and bandwidth. Dynamic scheduling is achieved by combining these strategies with the actual network conditions, fully leveraging the advantages of different scheduling strategies in terms of data and architecture. The final evaluation score for the network scheduling strategies can be expressed as:

[0073]

[0074] in, This represents the final evaluation score of the network partial scheduling strategy. , and These represent the score weights corresponding to priority, busy / idle time, and network level scheduling strategies, respectively. This is the initial score. , and These represent the scores corresponding to the scheduling strategies at the priority, busy / idle time, and network level, respectively.

[0075] This application embodiment determines the computing power tasks to be scheduled by the branch service center based on target device data received from the branch processing center; determines the candidate job computation graph of the computing power tasks to be scheduled based on the job dependencies of at least one subtask in the computing power tasks to be scheduled; wherein, the candidate job computation graph is a directed acyclic graph used to represent the job order of the subtasks to be scheduled; determines the predicted job time of the subtasks to be scheduled based on the job time prediction model, according to the resource requirement data and job scheduling data of the subtasks to be scheduled; corrects the candidate job computation graph based on the job type, predicted job time and job dependencies of the subtasks to be scheduled to obtain the target job computation graph of the computing power tasks to be scheduled; determines the target allocation resources of the subtasks to be scheduled based on the job type, resource requirement data and the currently allocable resources of the multi-level computing power scheduling system; wherein, the branch service center includes multiple computing devices; the currently allocable resources are the computing power resources of each of the multiple computing devices; and schedules the subtasks to be scheduled based on the target allocation resources and the target job computation graph. The above technical solution, by introducing a job computation graph and resource allocation to schedule computing tasks, can effectively improve the reliability and timeliness of task scheduling.

[0076] Example 2

[0077] Figure 2 This is a flowchart of a computing power task scheduling method according to Embodiment 2 of this application. Based on the technical solutions of the above embodiments, this embodiment refines the process of "the branch service center correcting the candidate job computation graph according to the job type, predicted job time, and job dependency of the sub-task to be scheduled, to obtain the target job computation graph of the computing power task to be scheduled" into "the branch service center prioritizing the sub-tasks to be scheduled according to the predicted job time and a preset job time threshold, to obtain the priority type of the sub-tasks to be scheduled; wherein, the priority type includes priority processing type and delayed processing type; the branch service center correcting the candidate job computation graph according to the job type, priority type, and job dependency of the sub-tasks to be scheduled, to obtain the target job computation graph of the computing power task to be scheduled." It should be noted that for parts not detailed in this embodiment, please refer to the relevant descriptions in other embodiments. Figure 2 As shown, the method includes:

[0078] S210. The branch service center determines the computing power tasks to be scheduled based on the target device data received from the branch processing center.

[0079] S220. The branch service center determines the candidate job computation graph of the computing power task to be scheduled based on the job dependency relationship of at least one subtask to be scheduled in the computing power task to be scheduled.

[0080] S230. The branch service center determines the predicted operation time of the sub-task based on the operation time prediction model and the resource requirement data and operation scheduling data of the sub-task to be completed.

[0081] S240. The branch service center prioritizes the subtasks to be performed based on the predicted operation time and the preset operation time threshold, and obtains the priority type of the subtasks to be performed. The priority type includes priority processing type and delayed processing type.

[0082] In this embodiment, the job time threshold refers to a preset time limit, typically used to define the maximum allowed execution time of a job. Priority types indicate the urgency of task processing and are usually categorized based on task importance, timeliness, or dependencies. Priority processing type indicates a task with high priority that needs to be processed as quickly as possible in the system; these are typically urgent or important tasks. Delayed processing type indicates a task that can be postponed and is not urgent or requires immediate execution.

[0083] S250. The branch service center modifies the candidate job computation graph based on the job type, priority type and job dependency of the sub-task to be scheduled, and obtains the target job computation graph of the computing power task to be scheduled.

[0084] Optionally, the job type may include at least one of critical jobs and auxiliary jobs. Critical jobs are those that are highly important and urgent within the system; they typically require priority execution because they have a decisive impact on the completion of the overall task. Auxiliary jobs are those that contribute to the overall task, but whose execution order or timing is relatively flexible; they usually do not directly affect the system's main objectives, but are crucial for improving system efficiency, optimizing resource utilization, or providing supporting functions.

[0085] It should be noted that the candidate job computation graph may include at least one job scheduling branch.

[0086] Optionally, the branch service center moves the pending subtasks of critical type and priority type to the front of the branch for each job scheduling branch, and moves the jobs with upstream dependencies on the pending subtasks forward synchronously; and moves the pending subtasks of auxiliary type and priority type to the end of the branch, and moves the jobs with downstream dependencies on the pending subtasks backward synchronously, to obtain the target job computation graph of the computing power tasks to be scheduled.

[0087] In this embodiment, upstream dependency means that a job (A) depends on the completion of another job (B) before it can begin execution. In other words, job A must wait for the execution result of job B before it can execute; that is, job B is an upstream dependency of job A. Downstream dependency means that another job (B) can only begin execution after a job (A) is completed; that is, job A is a prerequisite task for job B, and job B depends on the execution result of job A; that is, job B is a downstream dependency of job A.

[0088] Furthermore, if a dependency conflict is detected during the forward or backward movement of the pending subtask, the branch service center determines the location of the dependency conflict and the dependent conflict subtask; based on the dependent conflict subtask, the branch service center inserts a virtual synchronization node or a replication subtask of the dependent conflict subtask at the location of the dependency conflict.

[0089] In this embodiment, dependency conflict refers to a situation during job scheduling where conflicts arise between job dependencies, preventing some jobs from executing in the expected order. A dependency conflict location refers to the specific position or node where a dependency conflict occurs during scheduling; in the job scheduling execution graph, this location typically refers to the time point at which a job is scheduled or a certain stage in the execution order where inconsistencies or conflicts in dependencies occur. A dependency-conflicting job subtask refers to a specific job subtask affected by a dependency conflict during job scheduling; this subtask cannot execute as planned due to the dependency conflict, potentially causing the scheduling system to adjust the execution order or take other conflict resolution measures. A virtual synchronization node is a virtual node used in job scheduling to resolve dependency conflicts; it is not a specific job task but a marker representing a synchronization point in job dependencies. In the case of dependency conflicts, the role of the virtual synchronization node is to separate conflicting job subtasks, ensuring they can execute in the correct order. Copying a job subtask refers to creating a copy of a job in job scheduling to resolve dependency conflicts. By copying conflicting job subtasks, the scheduling system can distribute the execution of a job to multiple locations to resolve dependency conflicts. Copying job subtasks can ensure that jobs can be completed on time, while avoiding scheduling delays caused by dependency conflicts.

[0090] S260. Based on the job type, resource requirement data, and the currently available resources of the multi-level computing power scheduling system, the branch service center determines the target allocation resources for the sub-tasks to be performed.

[0091] S270. Based on the target resource allocation and target job calculation graph, the branch service center schedules the pending job subtasks in the computing power task to be scheduled.

[0092] In this embodiment, the branch service center determines the computing power tasks to be scheduled based on target device data received from the branch processing center; the branch service center determines a candidate job computation graph for the computing power tasks to be scheduled based on the job dependencies of at least one subtask in the task; wherein, the candidate job computation graph is a directed acyclic graph used to represent the job order of the subtasks; the branch service center determines the predicted job time of the subtasks based on the job time prediction model and the resource requirement data and job scheduling data of the subtasks; and the branch service center prioritizes the subtasks based on the predicted job time and a preset job time threshold to obtain the subtasks to be scheduled. The process involves several steps: First, the priority type of the task is determined. This includes priority processing and delayed processing. Second, the branch service center modifies the candidate job computation graph based on the job type, priority type, and job dependencies of the sub-tasks to be scheduled, resulting in the target job computation graph for the scheduled computing power tasks. Third, the branch service center determines the target allocation resources for the sub-tasks based on the job type, resource requirement data, and the currently available resources of the multi-level computing power scheduling system. The branch service center comprises multiple computing devices, and the currently available resources are the computing power resources of each device. Finally, the branch service center schedules the sub-tasks within the scheduled computing power tasks based on the target allocation resources and the target job computation graph. This technical solution, by introducing job computation graphs and resource configuration for scheduling computing power tasks, effectively improves the reliability and timeliness of task scheduling.

[0093] Example 3

[0094] Figure 3 This is a schematic diagram of a computing power task scheduling device according to Embodiment 3 of this application. It is applicable to scheduling and allocating computing power tasks for bank branches. This computing power task scheduling device can be implemented in hardware and / or software and can be configured in computer equipment, such as a multi-level computing power scheduling system. The multi-level computing power scheduling system includes an equipment operation center, a branch processing center, and a branch service center. The branch processing center is communicatively connected to both the branch service center and the equipment operation center. Figure 3 As shown, the device includes:

[0095] The task determination module 310 is used to determine the computing power tasks to be scheduled based on the target device data received from the branch processing center through the branch service center;

[0096] The computation graph determination module 320 is used to determine the candidate job computation graph of the computing power task to be scheduled by the branch service center based on the job dependency relationship of at least one subtask to be scheduled in the computing power task to be scheduled; wherein, the candidate job computation graph is a directed acyclic graph used to represent the job order of the subtasks to be scheduled.

[0097] The time prediction module 330 is used to determine the predicted operation time of the sub-task to be done based on the resource requirement data and operation scheduling data of the sub-task to be done through the branch service center based on the operation time prediction model.

[0098] The computation graph correction module 340 is used to correct the candidate job computation graph by the branch service center according to the job type, predicted job time and job dependency of the sub-task to be scheduled, so as to obtain the target job computation graph of the computing power task to be scheduled.

[0099] The resource determination module 350 is used to determine the target allocation resources for the sub-tasks to be performed by the branch service center based on the job type, resource requirement data, and the currently available resources of the multi-level computing power scheduling system. The branch service center includes multiple computing devices, and the currently available resources are the computing power resources of each of the multiple computing devices.

[0100] The task scheduling module 360 ​​is used to schedule the pending subtasks in the computing power task to be scheduled by the branch service center based on the target allocation resources and target job calculation graph.

[0101] This application embodiment determines the computing power tasks to be scheduled by the branch service center based on target device data received from the branch processing center; determines the candidate job computation graph of the computing power tasks to be scheduled based on the job dependencies of at least one subtask in the computing power tasks to be scheduled; wherein, the candidate job computation graph is a directed acyclic graph used to represent the job order of the subtasks to be scheduled; determines the predicted job time of the subtasks to be scheduled based on the job time prediction model, according to the resource requirement data and job scheduling data of the subtasks to be scheduled; corrects the candidate job computation graph based on the job type, predicted job time and job dependencies of the subtasks to be scheduled to obtain the target job computation graph of the computing power tasks to be scheduled; determines the target allocation resources of the subtasks to be scheduled based on the job type, resource requirement data and the currently allocable resources of the multi-level computing power scheduling system; wherein, the branch service center includes multiple computing devices; the currently allocable resources are the computing power resources of each of the multiple computing devices; and schedules the subtasks to be scheduled based on the target allocation resources and the target job computation graph. The above technical solution, by introducing a job computation graph and resource allocation to schedule computing tasks, can effectively improve the reliability and timeliness of task scheduling.

[0102] Optionally, the device may also include:

[0103] The data preprocessing module is used to preprocess the data obtained from at least one terminal device through the equipment operation center to obtain candidate device data, and then transmit it to the network processing center.

[0104] The parameter configuration module is used to configure the network parameters of the candidate device data through the branch processing center to obtain the target device data and transmit it to the branch service center.

[0105] Optionally, the job time prediction model is obtained by training a multiple linear regression model based on the execution data of each job subtask of the historical computing power task.

[0106] Optionally, the computational graph correction module 340 includes:

[0107] The priority division unit is used by the branch service center to prioritize the subtasks to be done based on the predicted job time and the preset job time threshold, so as to obtain the priority type of the subtasks to be done; among which, the priority type includes priority processing type and delayed processing type.

[0108] The computation graph correction unit is used by the branch service center to correct the candidate job computation graph according to the job type, priority type and job dependency of the subtask to be scheduled, so as to obtain the target job computation graph of the computing power task to be scheduled.

[0109] Optionally, the job type includes critical jobs and auxiliary jobs; the candidate job computation graph includes at least one job scheduling branch; correspondingly, the computation graph correction unit is specifically used for:

[0110] For each job scheduling branch, the branch service center moves the pending subtasks that are critical and prioritized to the front end of the branch, and moves the jobs that have upstream dependencies on the pending subtasks forward synchronously based on job dependencies.

[0111] Additionally, the branch service center moves the subtasks of the pending tasks that are auxiliary tasks and have a priority type of delayed processing to the end of the branch, and based on the task dependencies, moves the tasks that have a downstream dependency relationship with the pending subtasks to the end of the branch, thus obtaining the target task computation graph of the computing power task to be scheduled.

[0112] Optionally, the computational graph correction unit is also specifically used for:

[0113] If a dependency conflict is detected during the forward or backward movement of a pending subtask, the branch service center will determine the location of the dependency conflict and the dependent conflict subtask.

[0114] Based on the dependency conflict job subtask, the branch service center inserts a virtual synchronization node or a replication job subtask of the dependency conflict job subtask at the dependency conflict location.

[0115] Optionally, the multi-level computing power scheduling system also includes a cloud computing scheduling center; the cloud computing scheduling center is communicatively connected to the branch service center, the outlet processing center, and the equipment operation center; the device also includes a resource scheduling module, which includes:

[0116] The time monitoring unit is used to monitor the execution time of the computing power tasks to be scheduled in real time after the cloud computing scheduling center schedules the computing power tasks to be scheduled in the branch service center, so as to obtain the real-time execution time of the computing power tasks to be scheduled.

[0117] The resource gap determination unit is used to determine the current resource gap of the computing power task to be scheduled by the cloud computing scheduling center, based on the current available resources of the multi-level computing power scheduling system and the current required resources of the computing power task to be scheduled, provided that the execution time condition is met in real time.

[0118] The adjustment method determination unit is used to determine the resource adjustment method for the computing power tasks to be scheduled by the cloud computing scheduling center based on the current resource gap, the total resources of the multi-level computing power scheduling system, the current resource occupancy rate, and the resource early warning threshold.

[0119] Optional, the adjustment method determines the unit, specifically used for:

[0120] The remaining resource ratio of the multi-level computing power scheduling system is determined by subtracting the current resource occupancy rate and resource early warning threshold of the multi-level computing power scheduling system from the cloud computing scheduling center.

[0121] The remaining available resources of the multi-level computing power scheduling system are determined by the cloud computing scheduling center based on the remaining resource ratio and total resource volume of the multi-level computing power scheduling system.

[0122] The cloud computing scheduling center determines the resource adjustment method for the computing power tasks to be scheduled based on the remaining available resources and the current resource gap.

[0123] Optionally, the computing power tasks to be scheduled include at least financial risk monitoring and visualization and intelligent analysis of customer behavior; the sub-tasks to be assigned include at least real-time anti-fraud analysis, credit risk modeling, market risk stress testing, customer profile enhancement, intelligent recommendation of financial products, customer behavior analysis and account risk assessment.

[0124] The computing power task scheduling device provided in this application can execute the computing power task scheduling method provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects for executing each computing power task scheduling method.

[0125] According to embodiments of this application, this application also provides an electronic device, a readable storage medium, and a computer program product.

[0126] Example 4

[0127] Figure 4 This is a schematic diagram of the structure of an electronic device 410 implementing the computing task scheduling method of the embodiments of this application. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or claimed herein.

[0128] like Figure 4 As shown, the electronic device 410 includes at least one processor 411 and a memory, such as a read-only memory 412 or a random access memory 413, communicatively connected to the at least one processor 411. The memory stores computer programs executable by the at least one processor. The processor 411 can perform various appropriate actions and processes based on the computer program stored in the read-only memory 412 or loaded from storage unit 418 into the random access memory 413. The random access memory 413 can also store various programs and data required for the operation of the electronic device 410. The processor 411, read-only memory 412, and random access memory 413 are interconnected via a bus 414. An input / output interface 415 is also connected to the bus 414.

[0129] Multiple components in electronic device 410 are connected to input / output interface 415, including: input unit 416, such as keyboard, mouse, etc.; output unit 417, such as various types of monitors, speakers, etc.; storage unit 418, such as disk, optical disk, etc.; and communication unit 419, such as network card, modem, wireless transceiver, etc. Communication unit 419 allows electronic device 410 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0130] Processor 411 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 411 include, but are not limited to, central processing units, graphics processing units, various special-purpose artificial intelligence computing chips, various processors running machine learning model algorithms, digital signal processors, and any suitable processor, controller, microcontroller, etc. Processor 411 performs the various methods and processes described above, such as computing task scheduling methods.

[0131] In some embodiments, the computing task scheduling method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 418. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 410 via read-only memory 412 and / or communication unit 419. When the computer program is loaded into random access memory 413 and executed by processor 411, one or more steps of the computing task scheduling method described above may be performed. Alternatively, in other embodiments, processor 411 may be configured as the computing task scheduling method by any other suitable means (e.g., by means of firmware).

[0132] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays, application-specific integrated circuits (ASICs), application-specific standard products (ASICs), systems-on-a-chip (SoCs), payload programmable logic devices, computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0133] Computer programs used to implement the methods of this application may be written in any combination of one or more programming languages. These computer programs may be provided to the processor of a general-purpose computer, a special-purpose computer, or other programmable computing task scheduler, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0134] In the context of this application, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, optical fibers, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0135] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a cathode ray tube or liquid crystal display monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0136] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0137] A computing system can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product within the cloud computing service system to address the shortcomings of traditional physical hosts and virtual private servers, such as high management difficulty and weak business scalability.

[0138] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this application can be achieved, and this is not limited herein.

[0139] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A computing power task scheduling method, characterized in that, The method is executed by a multi-level computing power scheduling system; the multi-level computing power scheduling system includes an equipment operation center, a branch processing center, and a branch service center; the branch processing center is communicatively connected to the branch service center and the equipment operation center, respectively; the method includes: The branch service center determines the computing power tasks to be scheduled based on the target device data received from the branch processing center. The branch service center determines a candidate job computation graph for the computing power task to be scheduled based on the job dependencies of at least one subtask in the task to be scheduled; wherein, the candidate job computation graph is a directed acyclic graph used to characterize the job order of the subtasks to be scheduled. The branch service center determines the predicted job time of the sub-task based on the job time prediction model and the resource requirement data and job scheduling data of the sub-task to be done. The branch service center corrects the candidate job computation graph based on the job type of the sub-task to be scheduled, the predicted job time, and the job dependency relationship to obtain the target job computation graph of the computing power task to be scheduled. The branch service center determines the target allocation resources for the subtask to be performed based on the job type, the resource requirement data, and the currently available resources of the multi-level computing power scheduling system; wherein, the branch service center includes multiple computing devices; and the currently available resources are the computing power resources of each of the multiple computing devices. The branch service center schedules the pending subtasks in the pending computing power tasks based on the target allocated resources and the target job computation graph.

2. The method according to claim 1, characterized in that, The method further includes: The device operation center preprocesses the data obtained from at least one terminal device to obtain candidate device data, which is then transmitted to the network processing center. After configuring the network parameters of the candidate device data through the branch processing center, the target device data is obtained and transmitted to the branch service center.

3. The method according to claim 1, characterized in that, The job time prediction model is obtained by training a multiple linear regression model based on the execution data of each job subtask of historical computing power tasks.

4. The method according to claim 1, characterized in that, The branch service center modifies the candidate job computation graph based on the job type of the sub-task to be scheduled, the predicted job time, and the job dependencies to obtain the target job computation graph for the computing power task to be scheduled, including: The branch service center prioritizes the subtasks to be done based on the predicted job time and a preset job time threshold, thereby obtaining the priority type of the subtasks to be done; wherein, the priority type includes priority processing type and delayed processing type. The branch service center modifies the candidate job computation graph based on the job type, priority type, and job dependency of the sub-task to be scheduled, thereby obtaining the target job computation graph of the computing power task to be scheduled.

5. The method according to claim 4, characterized in that, The job types include critical jobs and auxiliary jobs; the candidate job computation graph includes at least one job scheduling branch; correspondingly, the branch service center modifies the candidate job computation graph according to the job type, priority type, and job dependency of the subtask to be scheduled, to obtain the target job computation graph of the computing power task to be scheduled, including: For each job scheduling branch, the branch service center moves the pending subtasks with critical job type and priority type to the front end of the branch, and moves the jobs with upstream dependencies on the pending subtasks forward synchronously. Furthermore, the branch service center moves the pending subtasks with auxiliary job type and delayed processing priority type to the end of the branch, and based on the job dependency relationship, moves the jobs with downstream dependency relationship with the pending subtasks to the end of the branch, thereby obtaining the target job computation graph of the computing power task to be scheduled.

6. The method according to claim 5, characterized in that, The method further includes: If a dependency conflict is detected during the forward or backward movement of the pending subtask, the branch service center determines the location of the dependency conflict and the dependent conflict subtask. The branch service center inserts a virtual synchronization node or a replication subtask of the dependency conflict job at the dependency conflict location according to the dependency conflict job subtask.

7. The method according to claim 1, characterized in that, The multi-level computing power scheduling system further includes a cloud computing scheduling center; the cloud computing scheduling center is communicatively connected to the branch service center, the outlet processing center, and the equipment operation center, respectively; the method further includes: After the cloud computing scheduling center schedules the computing power task to be scheduled at the branch service center, it monitors the execution time of the computing power task to be scheduled in real time to obtain the real-time execution time of the computing power task to be scheduled. When the real-time execution time meets the execution time condition, the cloud computing scheduling center determines the current resource gap of the computing task to be scheduled based on the current available resources of the multi-level computing power scheduling system and the current required resources of the computing power task to be scheduled. The cloud computing scheduling center determines the resource adjustment method for the computing power tasks to be scheduled based on the current resource gap, the total resources of the multi-level computing power scheduling system, the current resource occupancy rate, and the resource early warning threshold.

8. The method according to claim 7, characterized in that, The cloud computing scheduling center determines the resource adjustment method for the computing power tasks to be scheduled based on the current resource gap, the total resources of the multi-level computing power scheduling system, the current resource occupancy rate, and the resource warning threshold. This includes: The remaining resource ratio of the multi-level computing power scheduling system is determined by subtracting the current resource occupancy rate and the resource early warning threshold of the multi-level computing power scheduling system from the current resource occupancy rate of the multi-level computing power scheduling system by the cloud computing scheduling center. The remaining available resources of the multi-level computing power scheduling system are determined by the cloud computing scheduling center based on the remaining resource ratio and the total resource volume of the multi-level computing power scheduling system. The cloud computing scheduling center determines the resource adjustment method for the computing power tasks to be scheduled based on the remaining available resources and the current resource gap.

9. The method according to claim 1, characterized in that, The computing power tasks to be scheduled include at least financial risk monitoring and visualization and intelligent analysis of customer behavior; the sub-tasks to be performed include at least real-time anti-fraud analysis, credit risk modeling, market risk stress testing, customer profile enhancement, intelligent recommendation of financial products, customer behavior analysis and account risk assessment.

10. A computing power task scheduling device, configured in a multi-level computing power scheduling system; the multi-level computing power scheduling system includes an equipment operation center, a branch processing center, and a branch service center; the branch processing center is communicatively connected to the branch service center and the equipment operation center respectively; the device includes: The task determination module is used to determine the computing power tasks to be scheduled based on the target device data received from the branch processing center by the branch service center. The computation graph determination module is used to determine the candidate job computation graph of the computing power task to be scheduled by the branch service center based on the job dependency relationship of at least one subtask to be scheduled in the computing power task to be scheduled; wherein, the candidate job computation graph is a directed acyclic graph used to characterize the job order of the subtasks to be scheduled. The time prediction module is used to determine the predicted job time of the sub-task to be done based on the resource requirement data and job scheduling data of the sub-task to be done by the branch service center using the job time prediction model. The computation graph correction module is used to correct the candidate job computation graph through the branch service center based on the job type of the sub-task to be scheduled, the predicted job time, and the job dependency relationship, so as to obtain the target job computation graph of the computing power task to be scheduled. The resource determination module is used to determine the target allocation resources for the subtask to be performed by the branch service center based on the job type, the resource requirement data, and the currently available resources of the multi-level computing power scheduling system; wherein, the branch service center includes multiple computing devices; and the currently available resources are the computing power resources of each of the multiple computing devices. The task scheduling module is used to schedule the pending subtasks in the computing power task to be scheduled by the branch service center based on the target allocated resources and the target job computation graph.

11. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the computing task scheduling method as described in any one of claims 1-9.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the computing power task scheduling method as described in any one of claims 1-9.

13. A computer program product comprising a computer program that, when executed by a processor, implements the computing power task scheduling method according to any one of claims 1-9.

Images