A task dynamic scheduling method, system, platform and storage medium based on load sensing and efficiency monitoring

Abstract

The application discloses a kind of based on load perception and efficiency monitoring task dynamic scheduling method, system, platform and storage medium;The present application scheme is by constructing the multidimensional load perception mechanism of multitask processing node and task level processing efficiency time sequence attenuation detection model, realizes the real-time accurate quantification of global processing resource state and individual task processing efficiency;On this basis, based on the adaptive migration decision mechanism of multi-condition cooperative triggering, combined with the multi-constraint global combination optimization solution of node adaptation degree and processing continuity protection, precise identification and optimal migration scheduling of task between high load and low efficiency node are realized;Further, through the double-layer verification mechanism of rule verification and multi-prediction model integrated voting of migration scheme, the credibility and reliability of migration execution are ensured, finally relying on migration effect tracking and model parameter closed-loop feedback adaptive update, the continuous evolution of system decision strategy and the dynamic optimal balance of global processing efficiency are realized.

Description

Technical Field

[0001] This invention belongs to the field of task scheduling technology, specifically relating to a dynamic task scheduling method, system, platform, and storage medium based on load awareness and efficiency monitoring. Background Technology

[0002] In distributed task processing scenarios, task initiators typically delegate tasks to multiple task processing nodes simultaneously. Current multi-node collaborative processing models lack a dynamic adjustment mechanism after initial task allocation. When a node experiences processing capacity bottlenecks or a decline in efficiency, its tasks cannot be promptly transferred to better nodes, leading to processing delays. Furthermore, the lack of unified monitoring of real-time load status across nodes makes it impossible to perceive the uneven distribution of global processing capacity, easily resulting in resource waste where some nodes are overloaded while others are idle. Additionally, the triggering conditions for task transfer and the selection of target nodes lack quantitative decision-making basis, with the timing and direction of transfer relying primarily on the subjective experience of administrators. Moreover, the lack of mechanisms to protect the continuity of task processing during transfer means that frequent or inappropriate transfers can lead to negative effects such as interrupted processing progress and broken external interaction relationships.

[0003] In existing technologies, some task scheduling methods have attempted to address the aforementioned problems. For example, some solutions monitor server load status and combine it with deep reinforcement learning for task scheduling, but they only consider server load and do not fully perceive the real-time mapping relationship between load and processing capacity. Other solutions achieve intelligent task scheduling through dynamic priority algorithms, but lack time-series monitoring of task processing efficiency degradation and closed-loop feedback optimization mechanisms. In the field of distributed systems, although some research has proposed load-aware dynamic rescheduling strategies, these are mainly aimed at deep learning hybrid deployment scenarios, and their triggering conditions and target selection mechanisms are not applicable to general task processing scenarios.

[0004] Therefore, in order to address the aforementioned technical deficiencies, there is an urgent need to design and develop a task dynamic scheduling method, system, platform, and storage medium based on load awareness and efficiency monitoring. Summary of the Invention

[0005] To overcome the shortcomings and difficulties of the existing technology, the present invention aims to provide a task dynamic scheduling method, system, platform and storage medium based on load awareness and efficiency monitoring, which can accurately predict the effect of unexecuted strategies; monitor the load status of multiple nodes in real time, quantitatively evaluate the trend of processing efficiency changes, intelligently trigger task flow and ensure flow quality.

[0006] The first objective of this invention is to provide a task dynamic scheduling method based on load perception and efficiency monitoring; the second objective of this invention is to provide a task dynamic scheduling system based on load perception and efficiency monitoring; the third objective of this invention is to provide a task dynamic scheduling platform based on load perception and efficiency monitoring; and the fourth objective of this invention is to provide a computer-readable storage medium.

[0007] The first objective of this invention is achieved by the method comprising: First data corresponding to multiple task processing nodes is created, generated, and obtained. Second data corresponding to each task processing node is constructed based on the first data. The first data is load status data, and the second data is a load pressure characterization value. Establish and acquire third data corresponding to each task processing node and the tasks in progress, and identify whether there is efficiency decay in each task in progress based on the third data; wherein, the third data is processing efficiency time series data; In response to the second data and the efficiency decay identification results of the tasks in progress at each task processing node meeting the preset trigger conditions, a corresponding set of tasks to be migrated is created and generated. Based on the receiving adaptability of each candidate receiving node to each task in the set of tasks to be migrated, and the migration priority of each task, the migration scheme is optimized under preset constraints, and corresponding candidate migration schemes are generated; wherein, the target migration scheme is determined from the candidate migration schemes and executed.

[0008] Furthermore, the step of establishing, generating, and acquiring third data corresponding to each task processing node and the tasks in progress, and identifying whether there is efficiency degradation in each task in progress based on the third data, also includes: Based on at least two of the following: data on the achievement of phased sub-goals of tasks in progress, data on the decline of interaction activity over time, and data on the completion of milestone nodes, a processing efficiency index for tasks in progress is generated by merging them. Within a preset time window, the changing trend of the processing efficiency index is fitted, and a corresponding efficiency decay signal strength is calculated and generated based on the slope of the changing trend and the deviation of the processing efficiency index from the historical average; wherein, when the efficiency decay signal strength exceeds a first threshold, the task in progress is marked as having efficiency decay; the first threshold is a decay trigger threshold.

[0009] Furthermore, the preset triggering condition includes at least one of the following: The dispersion of the second data of each task processing node exceeds the second threshold; wherein, the second data is a load pressure characterization value; and the second threshold is a load imbalance threshold. In any task processing node, the proportion of tasks with declining efficiency exceeds a third threshold; wherein, the third threshold is a decline aggregation threshold. The load pressure change rate of any task processing node exceeds the fourth threshold and its load pressure characterization value exceeds the fifth threshold; wherein, the fourth threshold is the load growth rate threshold; and the fifth threshold is the high load threshold.

[0010] Furthermore, the load pressure characterization value is generated in the following manner: Generate and acquire fourth data corresponding to the task processing node and in multiple load dimensions; wherein, the fourth data is the original load data; Based on the degree of deviation between the original load data of each load dimension and their respective corresponding normalized reference values, a fifth data point corresponding to each load dimension is generated; wherein, the fifth data point is a normalized load value. Based on the preset weight values ​​of each load dimension, the normalized load values ​​are processed by weighted aggregation operation to generate corresponding load pressure characterization values.

[0011] Furthermore, the receiving adaptability is determined in the following way: The receiving suitability of candidate receiving nodes for the target task is determined based on at least two of the following factors: The statistical values ​​of the processing effectiveness of the candidate receiving node for the category to which the target task belongs in the historical period, the processing convenience score of the candidate receiving node in the geographical area associated with the target task, the current load pressure characterization value of the candidate receiving node, the quantitative score of the candidate receiving node's ability to process the target task at its current stage, and the estimated value of the processing continuity loss of migrating the target task from the source processing node to the candidate receiving node.

[0012] Furthermore, the estimated continuity loss for processing is determined based on at least two of the following factors: The quantitative values ​​of the processing time accumulated by the target task at the source processing node, the similarity between the source processing node and the candidate receiving node in the processing flow architecture, and the degree of relational dependence of the external interaction objects associated with the target task.

[0013] Furthermore, the step of optimizing the migration scheme under preset constraints based on the receiving adaptability of each candidate receiving node to each task in the set of tasks to be migrated, and the migration priority of each task, and constructing and generating corresponding candidate migration schemes, further includes: An optimization model is constructed with the goal of maximizing the overall benefit of the migration scheme; wherein the overall benefit is calculated as the sum of the products of the migration priority of each task to be migrated and its corresponding receiving adaptability. Under preset constraints, the optimization model is solved to generate corresponding candidate migration schemes; wherein, the preset constraints include: each task to be migrated can migrate to at most one candidate receiving node, the task to be migrated does not migrate back to its source processing node, and the load pressure characterization value of each candidate receiving node after accepting the migration task does not exceed the fifth threshold.

[0014] Furthermore, the step of optimizing the migration scheme under preset constraints based on the receiving adaptability of each candidate receiving node to each task in the set of tasks to be migrated, and the migration priority of each task, and constructing and generating corresponding candidate migration schemes, further includes: Based on the preset set of compliance rules, each candidate migration scheme is verified and processed item by item, and candidate migration schemes that fail the verification are filtered out. The remaining candidate migration schemes are input into multiple trained prediction models; wherein the multiple prediction models include at least an efficiency gain prediction model based on a tree model, a periodic change prediction model based on a sequence model, and a risk assessment model based on a probabilistic graphical model. The prediction results output by each prediction model are generated and obtained, and a comprehensive evaluation value of each candidate migration scheme is calculated based on each prediction result; wherein, the candidate migration scheme whose comprehensive evaluation value meets the preset conditions is selected as the target migration scheme.

[0015] Furthermore, the method also includes: After executing the target migration scheme, monitor the change in processing efficiency of the migrated tasks within a preset observation window; The value of the first threshold is adjusted based on the ratio of the number of tasks whose processing efficiency changes exceed a preset gain threshold to the total number of migrated tasks; wherein, the first threshold is an attenuation trigger threshold. And / or, calculate the statistical correlation metric between the load data of each load dimension and the change in processing efficiency, adjust the preset weight value corresponding to each load dimension according to the statistical correlation metric, and perform normalization processing on the adjusted preset weight value.

[0016] The second objective of this invention is achieved by providing a system for implementing the dynamic task scheduling method based on load awareness and efficiency monitoring, wherein the system comprises: The first creation and generation unit is used to create, generate, and acquire first data corresponding to multiple task processing nodes, and construct and generate second data corresponding to each task processing node based on the first data; wherein, the first data is load status data; and the second data is a load pressure characterization value. A generation unit is established to generate and acquire third data corresponding to each task processing node and the tasks in progress, and to identify whether there is efficiency decay in each task in progress based on the third data; wherein, the third data is processing efficiency time-series data. The second creation and generation unit is used to create and generate a corresponding set of tasks to be migrated in response to the second data of each task processing node and the efficiency decay identification result of the tasks in progress meeting the preset trigger conditions. A construction generation unit is used to optimize the migration scheme under preset constraints based on the reception adaptability of each candidate receiving node to each task in the set of tasks to be migrated, and the migration priority of each task, and to construct and generate corresponding candidate migration schemes; wherein, a target migration scheme is determined from the candidate migration schemes and executed.

[0017] Furthermore, the generation unit further includes: The first generation module is used to generate a processing efficiency index for the task based on at least two of the following: data on the achievement of phased sub-goals of the task, data on the decline of interaction activity over time, and data on the completion of milestone nodes. The second generation module is used to fit the changing trend of the processing efficiency index within a preset time window, and calculate and generate a corresponding efficiency decay signal intensity based on the slope of the changing trend and the deviation of the processing efficiency index from the historical average; wherein, when the efficiency decay signal intensity exceeds a first threshold, the task in progress is marked as having efficiency decay; the first threshold is a decay trigger threshold.

[0018] And / or, the second creation / generation unit further includes: The first determination module is used to determine whether the dispersion of the second data of each task processing node exceeds a second threshold; wherein, the second data is a load pressure characterization value; and the second threshold is a load imbalance threshold. The second determination module is used when the proportion of tasks with efficiency degradation in any task processing node exceeds a third threshold; wherein the third threshold is a degradation aggregation threshold. The third determination module is used when the load pressure change rate of any task processing node exceeds a fourth threshold and its load pressure characterization value exceeds a fifth threshold; wherein, the fourth threshold is the load growth rate threshold; and the fifth threshold is the high load threshold.

[0019] The load pressure characterization value is generated through the following module: The third generation module is used to generate and obtain fourth data corresponding to the task processing node and in multiple load dimensions; wherein, the fourth data is the original load data; The fourth generation module is used to construct and generate fifth data corresponding to each load dimension based on the degree of deviation between the original load data of each load dimension and their respective corresponding normalized reference values; wherein, the fifth data is a normalized load value; The fifth generation module is used to perform weighted aggregation operations on each normalized load value according to the preset weight values ​​of each load dimension, and generate the corresponding load pressure characterization value.

[0020] The receiving adaptability is determined by the following functional modules: The first processing module is used to determine the receiving adaptability of candidate receiving nodes to the target task based on at least two of the following factors: The second processing module is used for the statistical value of the processing effectiveness of the candidate receiving node for the category to which the target task belongs in the historical period, the processing convenience score of the candidate receiving node in the geographical area associated with the target task, the current load pressure characterization value of the candidate receiving node, the quantitative score of the candidate receiving node's ability to process the target task at its current stage, and the estimated value of the processing continuity loss when migrating the target task from the source processing node to the candidate receiving node.

[0021] The estimated continuity loss for processing is determined based on at least two of the following factors: The quantitative values ​​of the processing time accumulated by the target task at the source processing node, the similarity between the source processing node and the candidate receiving node in the processing flow architecture, and the degree of relational dependence of the external interaction objects associated with the target task.

[0022] Furthermore, and / or, the construction generation unit further includes: The first construction module is used to construct an optimization model with the goal of maximizing the overall benefit of the migration scheme; wherein the overall benefit is calculated and generated based on the sum of the products of the migration priority of each task to be migrated and the corresponding receiving adaptability. The second construction module is used to solve the optimization model under preset constraints and establish corresponding candidate migration schemes; wherein the preset constraints include: each task to be migrated can migrate to at most one candidate receiving node, the task to be migrated does not migrate back to its source processing node, and the load pressure characterization value of each candidate receiving node after accepting the migration task does not exceed the fifth threshold.

[0023] The third processing module is used to verify and process each candidate migration scheme according to the preset compliance rule set, and filter out the candidate migration schemes that fail the verification. The fourth processing module is used to input the remaining candidate migration schemes into multiple trained prediction models respectively; wherein, the multiple prediction models include at least an efficiency gain prediction model based on a tree model, a periodic change prediction model based on a sequence model, and a risk assessment model based on a probabilistic graphical model. The sixth generation module is used to generate and obtain the prediction results output by each prediction model, and calculate and generate a comprehensive evaluation value for each candidate migration scheme based on the prediction results; wherein, the candidate migration scheme whose comprehensive evaluation value meets the preset conditions is selected as the target migration scheme.

[0024] Furthermore, the system also includes: The monitoring unit is used to monitor the change in processing efficiency of the migrated tasks within a preset observation window after the target migration scheme is executed. The first adjustment module is used to adjust the value of the first threshold according to the ratio of the number of tasks whose processing efficiency changes exceed a preset gain threshold to the total number of migrated tasks; wherein, the first threshold is an attenuation trigger threshold. And / or, the second adjustment module is used to calculate the statistical correlation metric between the load data of each load dimension and the change in processing efficiency, adjust the preset weight value corresponding to each load dimension according to the statistical correlation metric, and perform normalization processing on the adjusted preset weight value.

[0025] The third objective of this invention is achieved as follows: it includes a processor, a memory, and a task dynamic scheduling platform control program based on load awareness and efficiency monitoring; wherein the task dynamic scheduling platform control program based on load awareness and efficiency monitoring is executed on the processor, the task dynamic scheduling platform control program based on load awareness and efficiency monitoring is stored in the memory, and the task dynamic scheduling platform control program based on load awareness and efficiency monitoring implements the task dynamic scheduling method based on load awareness and efficiency monitoring.

[0026] The fourth objective of this invention is achieved as follows: the computer-readable storage medium stores a task dynamic scheduling platform control program based on load perception and efficiency monitoring, and the task dynamic scheduling platform control program based on load perception and efficiency monitoring implements the task dynamic scheduling method based on load perception and efficiency monitoring.

[0027] This invention achieves real-time and accurate quantification of global processing resource status and individual task processing efficiency by constructing a multi-dimensional load perception mechanism for multi-task processing nodes and a task-level processing efficiency time-series decay detection model. Based on this, an adaptive migration decision-making mechanism triggered by multiple conditions, combined with a multi-constraint global combination optimization solution that balances node adaptability and processing continuity protection, enables accurate identification and optimal migration scheduling of tasks between high-load and low-efficiency nodes. Furthermore, a two-layer verification mechanism involving rule verification of the migration scheme and integrated voting of multiple prediction models ensures the credibility and reliability of migration execution. Finally, relying on migration effect tracking and closed-loop feedback adaptive updates of model parameters, the continuous evolution of system decision-making strategies and the dynamic optimal balance of global processing efficiency are achieved. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a schematic diagram of the process steps of a task dynamic scheduling method based on load awareness and efficiency monitoring according to the present invention. Figure 2 This is a schematic diagram of a task dynamic scheduling system architecture based on load awareness and efficiency monitoring according to the present invention; Figure 3 This is a schematic diagram of a task dynamic scheduling platform architecture based on load awareness and efficiency monitoring according to the present invention; Figure 4 This is a schematic diagram of a computer-readable storage medium architecture in one embodiment of the present invention. Detailed Implementation

[0030] To facilitate a clearer understanding of the objectives, technical solutions, and advantages of this invention, the invention will be further described below in conjunction with the accompanying drawings and specific embodiments. Those skilled in the art can easily understand other advantages and effects of this invention from the content disclosed in this specification.

[0031] This invention can also be implemented or applied through other different specific examples, and various details in this specification can also be modified and changed based on different viewpoints and applications without departing from the spirit of this invention.

[0032] It should be noted that if the embodiments of the present invention involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0033] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Secondly, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0034] Preferably, the task dynamic scheduling method based on load awareness and efficiency monitoring of the present invention is applied in one or more terminals or servers. The terminal is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0035] The terminal can be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal can interact with the customer via a keyboard, mouse, remote control, touchpad, or voice control device.

[0036] This invention provides a method, system, platform, and storage medium for dynamic task scheduling based on load awareness and efficiency monitoring. For example... Figure 1 The diagram shown is a flowchart of a task dynamic scheduling method based on load awareness and efficiency monitoring provided in an embodiment of the present invention.

[0037] In this embodiment, the task dynamic scheduling method based on load perception and efficiency monitoring can be applied to terminals or fixed terminals with display functions. The terminals are not limited to personal computers, smartphones, tablets, desktop computers or all-in-one computers with cameras, etc.

[0038] The task dynamic scheduling method based on load awareness and efficiency monitoring can also be applied to a hardware environment consisting of a terminal and a server connected to the terminal via a network. The network includes, but is not limited to, a wide area network (WAN), a metropolitan area network (MAN), or a local area network (LAN). The task dynamic scheduling method based on load awareness and efficiency monitoring in this embodiment can be executed by the server, by the terminal, or by both the server and the terminal.

[0039] For example, for terminals requiring dynamic task scheduling based on load awareness and efficiency monitoring, the dynamic task scheduling function based on load awareness and efficiency monitoring provided by the method of this invention can be directly integrated into the terminal, or a client for implementing the method of this invention can be installed. Alternatively, the method provided by this invention can also run on servers or other devices in the form of a Software Development Kit (SDK), providing an interface for dynamic task scheduling based on load awareness and efficiency monitoring in the form of an SDK. Terminals or other devices can then implement dynamic task scheduling based on load awareness and efficiency monitoring through the provided interface. The invention will be further described below with reference to the accompanying drawings.

[0040] like Figure 1 As shown, this invention provides a dynamic task scheduling method based on load awareness and efficiency monitoring, the method comprising the following steps: S01. Create, generate, and obtain first data corresponding to multiple task processing nodes, and construct and generate second data corresponding to each task processing node based on the first data; wherein, the first data is load status data; and the second data is a load pressure characterization value. S02. Establish, generate, and acquire third data corresponding to each task processing node and the tasks in progress, and identify whether there is efficiency decay in each task in progress based on the third data; wherein, the third data is processing efficiency time-series data. S03. In response to the second data of each task processing node and the efficiency decay identification result of the in-process task satisfying the preset triggering condition, a corresponding set of tasks to be migrated is created. S04. Based on the receiving adaptability of each candidate receiving node to each task in the set of tasks to be migrated, and the migration priority of each task, optimize the migration scheme under preset constraints, and construct and generate corresponding candidate migration schemes; wherein, determine the target migration scheme from the candidate migration schemes and execute it.

[0041] The process of establishing, generating, and acquiring third data corresponding to each task processing node and the tasks in progress, and identifying whether there is efficiency degradation in each task in progress based on the third data, further includes: S021. Based on at least two of the following: data on the achievement of phased sub-goals of the task in progress, data on the decline of interactive activity over time, and data on the completion of milestone nodes, generate a processing efficiency index for the task in progress. S022. Within a preset time window, fit the changing trend of the processing efficiency index, and calculate and generate a corresponding efficiency decay signal strength based on the slope of the changing trend and the deviation of the processing efficiency index from the historical average; wherein, when the efficiency decay signal strength exceeds a first threshold, the task in progress is marked as having efficiency decay; the first threshold is a decay trigger threshold.

[0042] The preset triggering conditions include at least one of the following: The dispersion of the second data of each task processing node exceeds the second threshold; wherein, the second data is a load pressure characterization value; and the second threshold is a load imbalance threshold. In any task processing node, the proportion of tasks with declining efficiency exceeds a third threshold; wherein, the third threshold is a decline aggregation threshold. The load pressure change rate of any task processing node exceeds the fourth threshold and its load pressure characterization value exceeds the fifth threshold; wherein, the fourth threshold is the load growth rate threshold; and the fifth threshold is the high load threshold.

[0043] The load pressure characterization value is generated in the following manner: S031. Generate and obtain fourth data corresponding to the task processing node and in multiple load dimensions; wherein, the fourth data is the original load data; S032. Based on the degree of deviation between the original load data of each load dimension and their respective corresponding normalized reference values, construct and generate the fifth data corresponding to each load dimension; wherein, the fifth data is the normalized load value; S033. Based on the preset weight values ​​of each load dimension, perform weighted aggregation operations to process each normalized load value and generate the corresponding load pressure characterization value.

[0044] The receiving adaptability is determined in the following way: The receiving suitability of candidate receiving nodes for the target task is determined based on at least two of the following factors: The statistical values ​​of the processing effectiveness of the candidate receiving node for the category to which the target task belongs in the historical period, the processing convenience score of the candidate receiving node in the geographical area associated with the target task, the current load pressure characterization value of the candidate receiving node, the quantitative score of the candidate receiving node's ability to process the target task at its current stage, and the estimated value of the processing continuity loss of migrating the target task from the source processing node to the candidate receiving node.

[0045] The estimated continuity loss for processing is determined based on at least two of the following factors: The quantitative values ​​of the processing time accumulated by the target task at the source processing node, the similarity between the source processing node and the candidate receiving node in the processing flow architecture, and the degree of relational dependence of the external interaction objects associated with the target task.

[0046] The step of optimizing the migration scheme under preset constraints based on the receiving adaptability of each candidate receiving node to each task in the set of tasks to be migrated, and the migration priority of each task, and constructing and generating corresponding candidate migration schemes, further includes: S041. Construct an optimization model with the goal of maximizing the overall benefit of the migration scheme; wherein the overall benefit is calculated and generated based on the sum of the products of the migration priority of each task to be migrated and the corresponding receiving adaptability. S042. Under preset constraints, the optimization model is solved to generate corresponding candidate migration schemes; wherein, the preset constraints include: each task to be migrated can migrate to at most one candidate receiving node, the task to be migrated does not migrate back to its source processing node, and the load pressure characterization value of each candidate receiving node after accepting the migration task does not exceed the fifth threshold.

[0047] The step of optimizing the migration scheme under preset constraints based on the receiving adaptability of each candidate receiving node to each task in the set of tasks to be migrated, and the migration priority of each task, and constructing and generating corresponding candidate migration schemes, further includes: S043. Based on the preset compliance rule set, verify and process each candidate migration scheme item by item, and filter out candidate migration schemes that fail the verification. S044. Input the remaining candidate migration schemes into multiple trained prediction models respectively; wherein, the multiple prediction models include at least an efficiency gain prediction model based on a tree model, a periodic change prediction model based on a sequence model, and a risk assessment model based on a probabilistic graphical model. S045. Generate and obtain the prediction results output by each prediction model, and calculate and generate a comprehensive evaluation value for each candidate migration scheme based on each prediction result; wherein, the candidate migration scheme whose comprehensive evaluation value meets the preset conditions is selected as the target migration scheme.

[0048] The method further includes the following steps: S051. After executing the target migration scheme, monitor the change in processing efficiency of the migrated tasks within a preset observation window; S052. Adjust the value of the first threshold according to the ratio of the number of tasks whose processing efficiency changes exceed a preset gain threshold to the total number of migrated tasks; wherein, the first threshold is an attenuation trigger threshold. And / or, S053, calculate the statistical correlation metric between the load data of each load dimension and the change in processing efficiency, adjust the preset weight value corresponding to each load dimension according to the statistical correlation metric, and perform normalization processing on the adjusted preset weight value.

[0049] Specifically, in this embodiment of the invention, a dynamic task scheduling method based on load awareness and efficiency monitoring is proposed, comprising the following steps: Step S1: Construct a real-time load status monitoring model for multi-task processing nodes, periodically collect multi-dimensional load indicators of each node and calculate the comprehensive load pressure index; its output serves as the input for steps S3 and S4.

[0050] Step S2: Construct a task-level processing efficiency decay detection model to perform time-series tracking of the processing efficiency of each task and identify efficiency decay signals; its output serves as the input for step S3.

[0051] Step S3: Based on the load pressure index output in step S1 and the efficiency decay signal output in step S2, the migration trigger decision engine determines whether to start task migration and determines the set of tasks to be migrated; its output serves as the input for step S4.

[0052] Step S4: Based on the load pressure index output in step S1 and the set of tasks to be migrated output in step S3, execute the migration target node matching and migration scheme optimization algorithm to generate candidate migration schemes; the output is used as the input for step S5.

[0053] Step S5: Perform automated rationality verification on the candidate migration schemes output in step S4, confirm the target migration scheme and perform task handover; its output serves as the input for step S6.

[0054] Step S6: Continuously track and evaluate the migration effect after step S5, and send the evaluation results back to update the load monitoring model parameters of step S1 and the efficiency decay detection threshold of step S2.

[0055] Furthermore, in step S1, real-time load status monitoring of multi-task processing nodes is carried out, a real-time load status monitoring model for multi-task processing nodes is constructed, and multi-dimensional load indicators of each node are periodically collected and a comprehensive load pressure index is calculated.

[0056] For the Each task processing node operates at a preset data collection cycle. The following load metrics are collected periodically: Task capacity load metrics (including but not limited to the total number of tasks currently in hand, total weighted task complexity, new task reception rate, and slope of task backlog growth trend); Human resource load metrics (including but not limited to front-line processing staff hour utilization rate, average number of tasks in hand per person, hourly load ratio, and capacity gap rate caused by personnel changes); Process bottleneck load metrics (including but not limited to queue length of process steps, waiting time for external dependencies, and average dwell time of key approval steps); Resource turnover load metrics (including but not limited to the proportion of prepaid expenses to budget, amount of unpaid expenses, and processing and settlement cycle delay).

[0057] For the Each task processing node at time Comprehensive load pressure index The calculation is as follows: (1) In the formula, This represents the total number of load dimensions, and its value range is positive integers. For the first Preset importance weights for each load dimension, with a value range of [value range missing]. ,and ; For the first The node at the th Normalized load values ​​in each dimension, with a range of [value range missing]. The calculation method is as follows: (2) In the formula, No. Each node at time... The Original load value; , For the first Preset normalized lower and upper bounds for dimensional load; It is a preset normalization smoothing constant to prevent the denominator from being zero, and the range is positive real numbers; Further calculations of the load pressure change rate are used to predict load trends: (3) In the formula, This is the preset trend observation window duration, with a range of positive real numbers. The output of step S1 is the comprehensive load pressure index of each node. and its rate of change This serves as the input for steps S3 and S4.

[0058] In step S2, task-level processing efficiency degradation detection, a task-level processing efficiency degradation detection model is constructed to perform time-series tracking of the processing efficiency of each task and identify efficiency degradation signals.

[0059] For the The node handling the first For each case, construct a time-series sequence of its processing efficiency. Defined in the... Processing efficiency index per observation period : (4) In the formula, For the first The task in the first The amount of phased achievements within each cycle, with a range of non-negative real numbers; For the first The case in the 1st The expected phased target quantity for each cycle is determined by the phased target in the task processing plan, with a value range of positive real numbers, and the data source is the task planning management module; For the first The time point (cycle number) at which the task most recently achieved a phased result is represented by a non-negative integer; The preset activity decay sensitivity coefficient controls the rate at which the efficiency score decays due to time without results; the value range is positive real numbers. For the first The task in the first The processing progress of each cycle is calculated based on the completion percentage of processing milestones (including but not limited to completion of Phase 1, Phase 2, Phase 3, Phase 4, Phase 5, and Phase 6), with a value range of [value range missing]. The data source is the task progress management system; The preset efficiency sub-dimension weights have a value range of [value range missing]. ,and ; This is a preset smoothing constant with a range of positive real numbers. For the task Efficiency timing The decay trend was detected using sliding window linear regression. (5) In the formula, The preset attenuation detection sliding window length (number of periods) has a value range of positive integers; This represents the average of the time sequence numbers within the window. : The mean of efficiency indicators within the window; Define the efficiency attenuation signal strength: (6) In the formula, For the task The historical average efficiency value since the assignment, with a range of positive real numbers, is calculated from the historical data in step S2. This is the preset attenuation amplification index, with values ​​ranging from real numbers greater than 0; when Exceeding the preset attenuation signal trigger threshold At that time, the task was marked as an "efficiency decay task".

[0060] The output of step S2 is the set of efficiency decay signals for all tasks at each node. This serves as the input for step S3.

[0061] Step S3, migration trigger decision, is based on the load pressure index output in step S1 and the efficiency decay signal output in step S2. The migration trigger decision engine determines whether to initiate migration based on the following combined conditions: Condition type A, triggered by load imbalance: (7) in, This represents the global average load pressure index. This is the preset load imbalance trigger threshold, with a value range of positive real numbers; Condition type B, efficiency decay clustering trigger: (8) In the formula, For the first The number of currently active tasks on each node. The preset threshold for triggering the decay task percentage has a value range of [value range missing]. ; Condition type C, triggered by deteriorating load trend: (9) In the formula, The preset load increase trigger threshold, This is the preset high load water level threshold.

[0062] When any one of conditions A, B, or C is met, the migration triggers the decision engine to start the migration process.

[0063] Once migration is triggered, tasks to be migrated are selected from high-load nodes or nodes with declining efficiency; tasks are defined. Migration priority score: (10) In the formula, For the task The cumulative number of achievements has been achieved, with a range of non-negative real numbers, and the data source is the achievement recording system. :Task The baseline target total, with a range of positive real numbers, is sourced from the task ledger system; For the task The remaining effective processing time, with a value range of non-negative real numbers, and the data source is the task status table; For the task The total processing time window, with a value range of positive real numbers, and the data source is the task status table; The preset migration priority dimension weights have a value range of [value range missing]. ,and The data source is the system policy configuration table.

[0064] Select the top migration priority scores The task constitutes the set of tasks to be migrated ,in It is dynamically determined by the following formula: (11) In the formula, This is a preset migration ratio adjustment coefficient with a value range of [value range missing]. ; The preset target load water level has a value range of [value range missing]. ; The output of step S3 is the set of tasks to be migrated. This serves as the input for step S4.

[0065] Step S4: Target node matching and scheme optimization for migration, including the set of tasks to be migrated. Each task in and each candidate receiving node ( ), calculate the receiving adaptation score : (12) In the formula, For the target node For the task The statistical value of the prior treatment effectiveness of the type, with a value range of [value range missing]. The data source is the node performance statistics table; For the target node In the mission The processing convenience score for the area is within a certain range. The data source is the node region coverage data table; The current load pressure index of the target node is calculated in step S1; For the target node For the task The specific processing capability score for the current processing stage (including but not limited to stage one, stage two, stage three, stage four, and stage five), with a value range of [value range missing]. The data source is the node capability label table; The calculation method for the processing continuity loss when migrating a task from the source node to the target node is described below. The preset weights for the receiver adaptation dimension have a value range of [value range missing]. ,and .

[0066] Assessment of Continuous Loss Management: (13) In the formula, For the task The data has been processed in the source node, and the value range is a non-negative real number. The data source is the task management system. The similarity of the processing flows between the source node and the target node is calculated from the normalized complement of the edit distance of the standardized processing flow templates of the two nodes, with a value range of [value range missing]. ; For the task The dependency on external interaction object relationships is calculated by normalizing factors such as the number of established interaction channels and the depth of historical interaction records, with a value range of [value range missing]. The data source is the interactive relationship management system; The preset weights for the continuous loss sub-dimensions have a range of values. ,and ; To optimize the global migration scheme, construct a global migration optimization problem: (14) Constraints: (Each task to be migrated can be migrated to a maximum of one target node); (Do not migrate back to the source node); (After migration, the load on the target node should not exceed the high load level;) .

[0067] In the formula, The migration decision matrix; To receive tasks The estimated increment of the target node load pressure index is obtained by estimating the ratio of task complexity to node processing capacity.

[0068] The solution is obtained using a combinatorial optimization algorithm, generating the first... One candidate migration scheme The output of step S4 is a set of candidate migration schemes. This serves as the input for step S5.

[0069] Step S5 involves automated rationality verification and migration execution, applying the candidate migration scheme set output in step S4. Each scheme in Perform the following two automated verifications: Automatic compliance verification based on a rules engine encodes business compliance requirements into a rule set. (Including but not limited to: the same task within the preset minimum holding period) Migration is not allowed within the same task, and the cumulative number of migrations for the same task shall not exceed the preset maximum number of migrations. The migration of tasks involving intermediate stages of the process must meet the statutory time limit requirements for changes, and the qualification level of the target node after migration must not be lower than the minimum qualification level required by the task. .

[0070] Regarding the plan Each migration instruction in the process is verified one by one: (15) when At that time, the plan It was deemed non-compliant and removed from the candidate set.

[0071] Decision confirmation based on ensemble model voting involves training multiple transfer effect prediction models (including but not limited to efficiency improvement prediction models based on gradient boosting trees, processing cycle reduction prediction models based on recurrent neural networks, and transfer risk probability prediction models based on Bayesian networks) for each compliant solution. Predict the post-transfer performance metrics separately: (16) The overall score of the scheme is calculated by combining the prediction results of each model: (17) In the formula, Output weights for the preset model.

[0072] When at least two out of multiple models predict a positive effect (i.e., the predicted efficiency improvement is greater than the preset minimum improvement threshold) The processing cycle is shortened, the predicted value is greater than 0, and the migration risk probability is lower than the preset risk tolerance threshold. At that time, the plan was confirmed through integrated voting.

[0073] Choose those that pass both verifications and The best solution Execute as the target migration plan.

[0074] For the target migration scheme For each migration instruction, a task handover data package is generated (including but not limited to the full task archive data package, current processing progress snapshot, established external interaction records, in-transit process status, and processing suggestions for the target node), and the task affiliation is updated.

[0075] The output of step S5 is the record of the executed migration results, which serves as the input for step S6.

[0076] Step S6: Tracking the transfer effect and updating model parameters. For each transferred task, during the preset observation period after the transfer... We will continuously track the following indicators: (18) In the formula, The migration execution time; For efficiency observation period length; This represents the change in efficiency after migration; a positive value indicates that the migration is effective, and a negative value indicates that the migration is ineffective.

[0077] For adaptive updating of model parameters, the efficiency decay detection threshold in step S2 is updated based on the migration effect evaluation results: (19) In the formula, The number of tasks that improve efficiency after migration ( The number of tasks), the data source is the tracking results of step S6; This represents the total number of migration tasks; The target migration success rate is preset, with a value range of [value range missing]. ; The step size is adaptively adjusted to a preset threshold, and the value range is positive real numbers; When the actual migration success rate is lower than the target value Increase (raise the migration trigger threshold to reduce unnecessary migrations); when the actual migration success rate is higher than the target value, Lower the threshold (increase the chance of migration).

[0078] Simultaneously update the weights of each load dimension in step S1. Gradient adjustment based on migration effect feedback is adopted: (20) In the formula, This is a statistical correlation measure, measuring the first... The correlation between dimensional load metrics and migration performance The preset weight adjustment step size; after the update, the weight adjustment step size is adjusted accordingly. Normalization process to ensure .

[0079] To achieve the above objectives, the present invention also provides a dynamic task scheduling system based on load awareness and efficiency monitoring, such as... Figure 2 As shown, the system is applied to the task dynamic scheduling method based on load awareness and efficiency monitoring, and the system includes: The first creation and generation unit is used to create, generate, and acquire first data corresponding to multiple task processing nodes, and construct and generate second data corresponding to each task processing node based on the first data; wherein, the first data is load status data; and the second data is a load pressure characterization value. A generation unit is established to generate and acquire third data corresponding to each task processing node and the tasks in progress, and to identify whether there is efficiency decay in each task in progress based on the third data; wherein, the third data is processing efficiency time-series data. The second creation and generation unit is used to create and generate a corresponding set of tasks to be migrated in response to the second data of each task processing node and the efficiency decay identification result of the tasks in progress meeting the preset trigger conditions. A construction generation unit is used to optimize the migration scheme under preset constraints based on the reception adaptability of each candidate receiving node to each task in the set of tasks to be migrated, and the migration priority of each task, and to construct and generate corresponding candidate migration schemes; wherein, a target migration scheme is determined from the candidate migration schemes and executed.

[0080] The establishment and generation unit further includes: The first generation module is used to generate a processing efficiency index for the task based on at least two of the following: data on the achievement of phased sub-goals of the task, data on the decay of interaction activity over time, and data on the completion of milestone nodes. The second generation module is used to fit the changing trend of the processing efficiency index within a preset time window, and calculate and generate a corresponding efficiency decay signal intensity based on the slope of the changing trend and the deviation of the processing efficiency index from the historical average; wherein, when the efficiency decay signal intensity exceeds a first threshold, the task in progress is marked as having efficiency decay; the first threshold is a decay trigger threshold.

[0081] And / or, the second creation / generation unit further includes: The first determination module is used to determine whether the dispersion of the second data of each task processing node exceeds a second threshold; wherein, the second data is a load pressure characterization value; and the second threshold is a load imbalance threshold. The second determination module is used when the proportion of tasks with efficiency degradation in any task processing node exceeds a third threshold; wherein the third threshold is a degradation aggregation threshold. The third determination module is used when the load pressure change rate of any task processing node exceeds a fourth threshold and its load pressure characterization value exceeds a fifth threshold; wherein, the fourth threshold is the load growth rate threshold; and the fifth threshold is the high load threshold.

[0082] The load pressure characterization value is generated through the following module: The third generation module is used to generate and obtain fourth data corresponding to the task processing node and in multiple load dimensions; wherein, the fourth data is the original load data; The fourth generation module is used to construct and generate fifth data corresponding to each load dimension based on the degree of deviation between the original load data of each load dimension and their respective corresponding normalized reference values; wherein, the fifth data is a normalized load value; The fifth generation module is used to perform weighted aggregation operations on each normalized load value according to the preset weight values ​​of each load dimension, and generate the corresponding load pressure characterization value.

[0083] The receiving adaptability is determined by the following functional modules: The first processing module is used to determine the receiving adaptability of candidate receiving nodes to the target task based on at least two of the following factors: The second processing module is used for the statistical value of the processing effectiveness of the candidate receiving node for the category to which the target task belongs in the historical period, the processing convenience score of the candidate receiving node in the geographical area associated with the target task, the current load pressure characterization value of the candidate receiving node, the quantitative score of the candidate receiving node's ability to process the target task at its current stage, and the estimated value of the processing continuity loss when migrating the target task from the source processing node to the candidate receiving node.

[0084] The estimated continuity loss for processing is determined based on at least two of the following factors: The quantitative values ​​of the processing time accumulated by the target task at the source processing node, the similarity between the source processing node and the candidate receiving node in the processing flow architecture, and the degree of relational dependence of the external interaction objects associated with the target task.

[0085] And / or, the construction generation unit further includes: The first construction module is used to construct an optimization model with the goal of maximizing the overall benefit of the migration scheme; wherein the overall benefit is calculated and generated based on the sum of the products of the migration priority of each task to be migrated and the corresponding receiving adaptability. The second construction module is used to solve the optimization model under preset constraints and establish corresponding candidate migration schemes; wherein the preset constraints include: each task to be migrated can migrate to at most one candidate receiving node, the task to be migrated does not migrate back to its source processing node, and the load pressure characterization value of each candidate receiving node after accepting the migration task does not exceed the fifth threshold.

[0086] The third processing module is used to verify and process each candidate migration scheme according to the preset compliance rule set, and filter out the candidate migration schemes that fail the verification. The fourth processing module is used to input the remaining candidate migration schemes into multiple trained prediction models respectively; wherein, the multiple prediction models include at least an efficiency gain prediction model based on a tree model, a periodic change prediction model based on a sequence model, and a risk assessment model based on a probabilistic graphical model. The sixth generation module is used to generate and obtain the prediction results output by each prediction model, and calculate and generate a comprehensive evaluation value for each candidate migration scheme based on the prediction results; wherein, the candidate migration scheme whose comprehensive evaluation value meets the preset conditions is selected as the target migration scheme.

[0087] And / or, the system further includes: The monitoring unit is used to monitor the change in processing efficiency of the migrated tasks within a preset observation window after the target migration scheme is executed. The first adjustment module is used to adjust the value of the first threshold according to the ratio of the number of tasks whose processing efficiency changes exceed a preset gain threshold to the total number of migrated tasks; wherein, the first threshold is an attenuation trigger threshold. And / or, the second adjustment module is used to calculate the statistical correlation metric between the load data of each load dimension and the change in processing efficiency, adjust the preset weight value corresponding to each load dimension according to the statistical correlation metric, and perform normalization processing on the adjusted preset weight value.

[0088] In the system solution embodiment of the present invention, the specific details of the method steps involved in the dynamic task scheduling based on load perception and efficiency monitoring have been described above. That is to say, the functional modules in the system are used to implement the steps or sub-steps in the above method embodiment, which will not be repeated here.

[0089] To achieve the above objectives, the present invention also provides a task dynamic scheduling platform based on load awareness and efficiency monitoring, such as... Figure 3As shown, the system includes a processor, a memory, and a task dynamic scheduling platform control program based on load awareness and efficiency monitoring. The processor executes the task dynamic scheduling platform control program based on load awareness and efficiency monitoring, which is stored in the memory. This program implements the task dynamic scheduling method steps based on load awareness and efficiency monitoring. Specific details of the steps have been described above and will not be repeated here.

[0090] In this embodiment of the invention, the task dynamic scheduling platform based on load awareness and efficiency monitoring has a built-in processor that can be composed of integrated circuits. For example, it can be composed of a single packaged integrated circuit, or multiple integrated circuits packaged with the same or different functions. This includes combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor connects to various components using various interfaces and lines, and executes programs or units stored in memory, as well as calling data stored in memory, to perform various functions of task dynamic scheduling based on load awareness and efficiency monitoring and to process data. The memory is used to store program code and various data. It is installed in a task dynamic scheduling platform based on load awareness and efficiency monitoring, and enables high-speed and automatic access to programs or data during operation. The memory includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0091] To achieve the above objectives, the present invention also provides a computer-readable storage medium, such as... Figure 4 As shown, the computer-readable storage medium stores a task dynamic scheduling platform control program based on load perception and efficiency monitoring. The task dynamic scheduling platform control program based on load perception and efficiency monitoring implements the task dynamic scheduling method steps based on load perception and efficiency monitoring. The specific details of the steps have been described above and will not be repeated here.

[0092] In the description of embodiments of the present invention, it should be noted that any process or method description in the flowcharts or otherwise described herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the present invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order according to the functions involved, as should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0093] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processing module, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, a “computer-readable medium” can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, the computer-readable medium can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0094] This invention achieves real-time and accurate quantification of global processing resource status and individual task processing efficiency by constructing a multi-dimensional load perception mechanism for multi-task processing nodes and a task-level processing efficiency time-series decay detection model. Based on this, an adaptive migration decision-making mechanism triggered by multiple conditions, combined with a multi-constraint global combination optimization solution that balances node adaptability and processing continuity protection, enables accurate identification and optimal migration scheduling of tasks between high-load and low-efficiency nodes. Furthermore, a two-layer verification mechanism involving rule verification of the migration scheme and integrated voting of multiple prediction models ensures the credibility and reliability of migration execution. Finally, relying on migration effect tracking and closed-loop feedback adaptive updates of model parameters, the continuous evolution of system decision-making strategies and the dynamic optimal balance of global processing efficiency are achieved.

[0095] In other words, this invention achieves real-time monitoring and quantitative evaluation of the load status of task processing nodes by constructing a multi-dimensional load index system and calculating the comprehensive load pressure index and its rate of change, overcoming the deficiency of lacking unified monitoring of load status in existing technologies. By constructing a three-dimensional processing efficiency index that integrates the achievement of phased results, activity decay, and progress advancement, and using sliding window linear regression to detect efficiency decay trends, it can accurately identify inefficient tasks and provide a quantitative basis for workflow decisions. Furthermore, it designs three independent triggering conditions: load imbalance triggering, efficiency decay aggregation triggering, and load trend deterioration triggering, forming an "OR" logic decision, which improves the sensitivity and robustness of workflow triggering. It also models the matching of tasks to target nodes as a constrained combinatorial optimization problem, comprehensively considering reception adaptability and processing continuity protection, and solves the optimal workflow scheme from a global perspective, avoiding system performance degradation caused by local decisions. In addition, by introducing processing continuity loss evaluation (comprehensively considering the processed time, process similarity, and relationship maintenance dependency) into reception adaptability, it avoids task progress interruptions and external interaction relationship breaks caused by frequent or improper workflows.

[0096] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A dynamic task scheduling method based on load awareness and efficiency monitoring, characterized in that, The method includes: First data corresponding to multiple task processing nodes is created, generated, and obtained. Second data corresponding to each task processing node is constructed based on the first data. The first data is load status data, and the second data is a load pressure characterization value. Establish and acquire third data corresponding to each task processing node and the tasks in progress, and identify whether there is efficiency decay in each task in progress based on the third data; wherein, the third data is processing efficiency time series data; In response to the second data and the efficiency decay identification results of the tasks in progress at each task processing node meeting the preset trigger conditions, a corresponding set of tasks to be migrated is created and generated. Based on the receiving adaptability of each candidate receiving node to each task in the set of tasks to be migrated, and the migration priority of each task, the migration scheme is optimized under preset constraints, and corresponding candidate migration schemes are generated; wherein, the target migration scheme is determined from the candidate migration schemes and executed.

2. The task dynamic scheduling method based on load awareness and efficiency monitoring according to claim 1, characterized in that, The process of establishing, generating, and acquiring third data corresponding to each task processing node and the tasks in progress, and identifying whether there is efficiency degradation in each task in progress based on the third data, further includes: Based on at least two of the following: data on the achievement of phased sub-goals of tasks in progress, data on the decline of interactive activity over time, and data on the completion of milestone nodes, a processing efficiency index for tasks in progress is generated by merging them. Within a preset time window, the changing trend of the processing efficiency index is fitted, and a corresponding efficiency decay signal strength is calculated and generated based on the slope of the changing trend and the deviation of the processing efficiency index from the historical average; wherein, when the efficiency decay signal strength exceeds a first threshold, the task in progress is marked as having efficiency decay; the first threshold is a decay trigger threshold.

3. The task dynamic scheduling method based on load awareness and efficiency monitoring according to claim 1, characterized in that, The preset triggering conditions include at least one of the following: The dispersion of the second data of each task processing node exceeds the second threshold; wherein, the second data is a load pressure characterization value; and the second threshold is a load imbalance threshold. In any task processing node, the proportion of tasks with declining efficiency exceeds a third threshold; wherein, the third threshold is a decline aggregation threshold. The load pressure change rate of any task processing node exceeds the fourth threshold and its load pressure characterization value exceeds the fifth threshold; wherein, the fourth threshold is the load growth rate threshold; and the fifth threshold is the high load threshold.

4. A task dynamic scheduling method based on load awareness and efficiency monitoring according to claim 1 or 3, characterized in that, The load pressure characterization value is generated in the following manner: Generate and acquire fourth data corresponding to the task processing node and in multiple load dimensions; wherein, the fourth data is the original load data; Based on the degree of deviation between the original load data of each load dimension and their respective corresponding normalized reference values, a fifth data point corresponding to each load dimension is generated; wherein, the fifth data point is a normalized load value. Based on the preset weight values ​​of each load dimension, the normalized load values ​​are processed by weighted aggregation operation to generate corresponding load pressure characterization values.

5. The task dynamic scheduling method based on load awareness and efficiency monitoring according to claim 1, characterized in that, The step of optimizing the migration scheme under preset constraints based on the receiving adaptability of each candidate receiving node to each task in the set of tasks to be migrated, and the migration priority of each task, and constructing and generating corresponding candidate migration schemes, further includes: An optimization model is constructed with the goal of maximizing the overall benefit of the migration scheme; wherein the overall benefit is calculated as the sum of the products of the migration priority of each task to be migrated and its corresponding receiving adaptability. Under preset constraints, the optimization model is solved to generate corresponding candidate migration schemes; wherein, the preset constraints include: each task to be migrated can migrate to at most one candidate receiving node, the task to be migrated does not migrate back to its source processing node, and the load pressure characterization value of each candidate receiving node after accepting the migration task does not exceed the fifth threshold.

6. A task dynamic scheduling method based on load awareness and efficiency monitoring according to claim 1 or 5, characterized in that, The step of optimizing the migration scheme under preset constraints based on the receiving adaptability of each candidate receiving node to each task in the set of tasks to be migrated, and the migration priority of each task, and constructing and generating corresponding candidate migration schemes, further includes: Based on the preset set of compliance rules, each candidate migration scheme is verified and processed item by item, and candidate migration schemes that fail the verification are filtered out. The remaining candidate migration schemes are input into multiple trained prediction models; wherein the multiple prediction models include at least an efficiency gain prediction model based on a tree model, a periodic change prediction model based on a sequence model, and a risk assessment model based on a probabilistic graphical model. The prediction results output by each prediction model are generated and obtained, and a comprehensive evaluation value of each candidate migration scheme is calculated based on each prediction result; wherein, the candidate migration scheme whose comprehensive evaluation value meets the preset conditions is selected as the target migration scheme.

7. The task dynamic scheduling method based on load awareness and efficiency monitoring according to claim 1, characterized in that, The method further includes: After executing the target migration scheme, monitor the change in processing efficiency of the migrated tasks within a preset observation window; The value of the first threshold is adjusted based on the ratio of the number of tasks whose processing efficiency changes exceed a preset gain threshold to the total number of migrated tasks; wherein, the first threshold is an attenuation trigger threshold.

8. A task dynamic scheduling system based on load awareness and efficiency monitoring, characterized in that, The system is applied to the task dynamic scheduling method based on load awareness and efficiency monitoring as described in any one of claims 1 to 7, and the system includes: The first creation and generation unit is used to create, generate, and acquire first data corresponding to multiple task processing nodes, and construct and generate second data corresponding to each task processing node based on the first data; wherein, the first data is load status data; and the second data is a load pressure characterization value. A generation unit is established to generate and acquire third data corresponding to each task processing node and the tasks in progress, and to identify whether there is efficiency decay in each task in progress based on the third data; wherein, the third data is processing efficiency time-series data. The second creation and generation unit is used to create and generate a corresponding set of tasks to be migrated in response to the second data of each task processing node and the efficiency decay identification result of the tasks in progress meeting the preset trigger conditions. A construction generation unit is used to optimize the migration scheme under preset constraints based on the reception adaptability of each candidate receiving node to each task in the set of tasks to be migrated, and the migration priority of each task, and to construct and generate corresponding candidate migration schemes; wherein, a target migration scheme is determined from the candidate migration schemes and executed.

9. A task dynamic scheduling platform based on load awareness and efficiency monitoring, characterized in that, The system includes a processor, a memory, and a task dynamic scheduling platform control program based on load awareness and efficiency monitoring. The task dynamic scheduling platform control program based on load awareness and efficiency monitoring is executed on the processor and stored in the memory. The task dynamic scheduling platform control program based on load awareness and efficiency monitoring implements the task dynamic scheduling method based on load awareness and efficiency monitoring as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a task dynamic scheduling platform control program based on load perception and efficiency monitoring. The task dynamic scheduling platform control program based on load perception and efficiency monitoring implements the task dynamic scheduling method based on load perception and efficiency monitoring as described in any one of claims 1 to 7.

Images