Task scheduling method, electronic equipment and storage medium

Abstract

The invention discloses a task scheduling method, electronic equipment and a storage medium. The method is applied to a configuration center side and comprises the following steps: acquiring a configuration instruction subjected to permission verification; acquiring multi-dimensional service scene data based on the configuration instruction, and inputting the multi-dimensional service scene data into a preset adaptive strategy model to generate a corresponding target adaptive strategy; the preset adaptive strategy model is obtained based on rule learning and transfer learning training; and performing visual simulation verification on the target adaptive strategy, and issuing the target adaptive strategy passing the verification to a server side. According to the scheme, the target self-adaptive strategy adaptive to the current service scene is generated through the multi-dimensional service scene data collected in real time and the preset self-adaptive strategy model, and the target self-adaptive strategy passing the visual simulation verification is issued to the server side for scheduling execution, so that the optimal task scheduling in the dynamic service scene can be realized, and the task scheduling efficiency is improved. And the autonomy and maintainability of the system are improved.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a task scheduling method, electronic device, and storage medium. Background Technology

[0002] In existing banking systems, branches within the same time zone share a single system, typically developed and maintained by the head office. However, due to differences in time zones, regulatory policies, and user groups across branches, the services offered by each branch also vary. This leads to inconsistent configuration of business preferences and system optimization across branches within the same deployment area, resulting in a situation where some aspects are neglected while others are addressed. Therefore, achieving optimal task scheduling for business operations has become a pressing issue. Summary of the Invention

[0003] This invention provides a task scheduling method, electronic device, and storage medium. By collecting multi-dimensional business scenario data in real time and using a preset adaptive strategy model, a target adaptive strategy adapted to the current business scenario is generated. The target adaptive strategy verified by visualization simulation is then sent to the server side for scheduling and execution. This enables optimal task scheduling in dynamic business scenarios and improves the system's autonomy and maintainability.

[0004] According to one aspect of the present invention, a task scheduling method is provided, applied to a configuration center side, the method comprising:

[0005] Obtain configuration commands that have been verified by authorization;

[0006] Multi-dimensional business scenario data is collected based on configuration instructions, and the multi-dimensional business scenario data is input into a preset adaptive strategy model to generate the corresponding target adaptive strategy; the preset adaptive strategy model is trained based on rule learning and transfer learning.

[0007] The target adaptive strategy is visualized and simulated for verification, and the verified target adaptive strategy is then sent to the server.

[0008] According to another aspect of the present invention, a task scheduling method is provided, applied to the server side, the method comprising:

[0009] Obtain the target adaptive strategy issued by the configuration center;

[0010] Generate process orchestration instructions based on target adaptive strategies and business requests;

[0011] The corresponding scheduling tasks are executed according to the process orchestration instructions, and the generated task execution results are fed back to the configuration center.

[0012] According to another aspect of the present invention, a task scheduling apparatus is provided, applied to a configuration center side, the apparatus comprising:

[0013] The configuration command acquisition module is used to acquire configuration commands that have been verified by permissions.

[0014] The strategy generation module is used to collect multi-dimensional business scenario data based on configuration instructions and input the multi-dimensional business scenario data into a preset adaptive strategy model to generate the corresponding target adaptive strategy; the preset adaptive strategy model is trained based on rule learning and transfer learning.

[0015] The strategy verification and distribution module is used to perform visual simulation verification of the target adaptive strategy and distribute the verified target adaptive strategy to the server side.

[0016] According to another aspect of the present invention, a task scheduling apparatus is provided, applied on a server side, the apparatus comprising:

[0017] The strategy acquisition module is used to acquire the target adaptive strategy issued by the configuration center.

[0018] The orchestration instruction generation module is used to generate process orchestration instructions based on the target adaptive strategy and business requests;

[0019] The scheduling and execution module is used to execute the corresponding scheduling tasks according to the process orchestration instructions and to feed back the generated task execution results to the configuration center.

[0020] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0021] At least one processor; and

[0022] A memory communicatively connected to the at least one processor; wherein,

[0023] The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the task scheduling method according to any embodiment of the present invention.

[0024] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the task scheduling method described in any embodiment of the present invention.

[0025] According to another aspect of the present invention, a computer program product is provided, the computer program product comprising a computer program that, when executed by a processor, implements the task scheduling method described in any embodiment of the present invention.

[0026] The task scheduling method provided in this invention is applied to the configuration center side. It acquires configuration instructions that have undergone permission verification; collects multi-dimensional business scenario data based on the configuration instructions; and inputs this multi-dimensional business scenario data into a preset adaptive strategy model to generate a corresponding target adaptive strategy. The preset adaptive strategy model is trained based on rule learning and transfer learning. The target adaptive strategy is then visually simulated and verified, and the verified target adaptive strategy is sent to the server side. This solution generates a target adaptive strategy adapted to the current business scenario by using real-time collected multi-dimensional business scenario data and a preset adaptive strategy model, and sends the visually simulated and verified target adaptive strategy to the server side for scheduling and execution. This enables optimal task scheduling in dynamic business scenarios, improving the system's autonomy and maintainability.

[0027] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. 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 structural diagram of a task scheduling system provided in Embodiment 1 of the present invention;

[0030] Figure 2 This is a flowchart of a task scheduling method provided in Embodiment 1 of the present invention;

[0031] Figure 3 This is a flowchart of a task scheduling method provided according to Embodiment 2 of the present invention;

[0032] Figure 4 This is a flowchart of a task scheduling method provided in Embodiment 3 of the present invention;

[0033] Figure 5 This is a flowchart of a task scheduling method provided in Embodiment 4 of the present invention;

[0034] Figure 6 This is a flowchart of a task scheduling method provided in Embodiment 5 of the present invention;

[0035] Figure 7 This is a structural diagram of the configuration center side provided according to Embodiment 5 of the present invention;

[0036] Figure 8 This is a flowchart of the adaptive strategy determination process provided in Embodiment 5 of the present invention;

[0037] Figure 9 This is a structural diagram of the server side provided according to Embodiment 5 of the present invention;

[0038] Figure 10 This is a schematic diagram of the structure of a task scheduling device according to Embodiment Six of the present invention;

[0039] Figure 11 This is a schematic diagram of the structure of a task scheduling device according to Embodiment 7 of the present invention;

[0040] Figure 12 This is a schematic diagram of the structure of an electronic device that implements the task scheduling method of this invention. Detailed Implementation

[0041] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0042] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention 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 the invention 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.

[0043] Example 1

[0044] Figure 1 This is a structural diagram of a task scheduling system provided in Embodiment 1 of the present invention. Figure 1As shown, the task scheduling system includes a configuration center 10 and a server 20. The configuration center 10 is mainly responsible for the intelligent generation and verification of adaptive policies, including processes such as permission verification and policy configuration management, policy generation and optimization, and policy verification and distribution. The server 20 is mainly responsible for the efficient execution and feedback of policies, including processes such as dynamic process orchestration, intelligent scheduling and execution, and task feedback optimization.

[0045] In this embodiment, by configuring the coordinated linkage between the central side 10 and the server side 20, differentiated scheduling can be performed according to the system operating environment, interaction environment, or branch business preferences, and adaptive strategies can be dynamically adjusted and optimized to improve business processing efficiency and business transaction quality, thereby enhancing user experience and transaction smoothness.

[0046] Based on the above task scheduling system Figure 2 This is a flowchart of a task scheduling method provided in Embodiment 1 of the present invention. This embodiment is applicable to situations where dynamic task scheduling is performed based on an adaptive strategy of the current business scenario. The method can be executed by a task scheduling device, which can be implemented in hardware and / or software and can be configured in an electronic device. Figure 2 As shown in the figure, the task scheduling method provided in this embodiment is applied to the configuration center side and specifically includes the following steps:

[0047] S110, Obtain the configuration command that has been verified by permissions.

[0048] In this context, configuration instructions refer to commands input by users (such as system developers) to the configuration center to define task scheduling rules, which drive the system's adaptive scheduling. These configuration instructions may include information such as the system's adaptive goals, adaptive reasons, and adaptive decision-making methods. The adaptive goals can refer to the state the system needs to achieve or the performance indicators to be optimized, serving as the optimization direction for subsequent strategy generation. Examples of adaptive goals include business response time < 500ms and cross-organization task success rate ≥ 99.9%. Adaptive reasons can refer to the conditions that trigger strategy adjustments, such as detecting resource load exceeding a preset threshold or a surge in abnormal logs. The adaptive decision-making methods can refer to the logic behind the generation of adaptive strategies, such as generating new strategies based on rule learning or migrating optimization strategies from similar historical scenarios.

[0049] Permission verification refers to the process by which the configuration center verifies the legitimacy of the initiator of the configuration command by authenticating the user's identity and matching role permissions. It may include verification methods such as digital signature verification and RBAC (Role-Based Access Control) model verification to ensure the security and compliance of configuration commands.

[0050] In this embodiment of the invention, users (such as system developers) can access the front-end interactive interface of the configuration center through a web browser, enter their account and password for identity authentication, and then the configuration center will perform permission authentication on the user's identity information through a preset RBAC model or other methods to confirm whether the user has the permission to initiate configuration commands. If the user permission verification fails, the front-end interactive interface will display a "no operation permission" prompt and terminate the process. If the user permission verification passes, the configuration command submitted by the user will be received to trigger the task scheduling adaptive process. The configuration command may include information such as the system adaptive target, the adaptive reason, and the adaptive decision method.

[0051] S120. Collect multi-dimensional business scenario data based on configuration instructions, and input the multi-dimensional business scenario data into the preset adaptive strategy model to generate the corresponding target adaptive strategy; the preset adaptive strategy model is trained based on rule learning and transfer learning.

[0052] Multidimensional business scenario data refers to a collection of multidimensional data reflecting the current operational status of a business scenario, such as external environment data, system data, and application requirement data. Target adaptive strategy refers to a dynamic task scheduling scheme generated by a pre-defined adaptive strategy model. This scheme may include specific policy rules such as resource allocation rules, task execution priorities, and exception handling rules, accurately adapting to the real-time needs of the current business scenario and guiding the server side to execute specific scheduling tasks.

[0053] A pre-trained adaptive strategy model refers to a model with dynamic scheduling capabilities that has been trained beforehand. Its core function is to generate an adaptive strategy—a scheduling strategy—based on input multi-dimensional business scenario data that adapts to the current business scenario. Essentially, it is a hybrid model that integrates rule learning and transfer learning. It can both follow historically validated effective rules and adapt to new scenarios by transferring historical knowledge. Rule learning is a machine learning method that extracts scheduling rules from historical business data, providing a basic framework for strategy generation and ensuring that the strategy conforms to known effective scheduling logic. Transfer learning is a learning mechanism that transfers existing historical business scenario knowledge (such as historical strategy parameters and scheduling experience) to new scenarios. By reusing information from historical data that is similar to the current scenario, it reduces the training cost in new scenarios, enabling the strategy to quickly adapt to unknown or dynamically changing business scenarios.

[0054] In this embodiment of the invention, the configuration center can collect multi-dimensional business scenario data, such as external environment data, system data, and application requirement data, based on configuration instructions. This ensures that the data input to the model can fully reflect the overall business scenario, providing a high-quality foundation for strategy generation. Then, the collected multi-dimensional business scenario data is preprocessed to ensure data quality. Finally, the preprocessed multi-dimensional business scenario data is input into a pre-trained preset adaptive strategy model. Through rule learning and transfer learning optimization, a target adaptive strategy adapted to the current business scenario is generated, such as resource allocation rules, task execution priorities, and exception handling rules.

[0055] It is important to understand that the preset adaptive strategy model in this embodiment possesses cross-scenario adaptability through transfer learning. When adding new business scenarios, it eliminates the need for model reconstruction; only fine-tuning of transfer parameters is required to generate effective strategies, thereby reducing system upgrade costs. For example, when a new partner institution is added, the preset adaptive strategy model can quickly migrate scheduling strategies from similar historical institutions, improving the adaptability and efficiency of strategy generation.

[0056] S130. Perform visual simulation verification of the target adaptive strategy, and send the verified target adaptive strategy to the server side.

[0057] Visual simulation verification can be understood as executing a target adaptive strategy in a virtual environment and displaying the strategy execution process and results through a graphical interface. It can dynamically display system performance indicators (such as transaction error rate, transaction response time, resource utilization, etc.) during the strategy execution process and supports manual interactive verification.

[0058] In this embodiment of the invention, the configuration center can be configured with a preset simulation scenario (virtual simulation environment) that is isomorphic to the current business scenario. This simulation environment can be used to simulate the current business scenario. Then, the generated target adaptive strategy is imported into the preset simulation environment, the core parameters in the strategy, such as resource allocation ratio, are parsed, and mapped into instructions executable by the simulation environment. Based on these instructions, the strategy simulation is started, the system performance indicators are monitored in real time, and the strategy execution process is displayed through a visual interface, such as dynamically displaying indicator change curves and heatmaps, so that system developers can intuitively observe the strategy execution status, bottlenecks, etc. through the visual interface. If the monitored system performance indicators do not meet the preset threshold, the model parameters of the preset adaptive strategy model are adjusted, the strategy is regenerated, and the simulation verification is performed repeatedly. If the monitored system performance indicators meet the preset threshold, the strategy simulation verification is determined to be passed, the verified target adaptive strategy is standardized and packaged, and sent to the server side for scheduling and execution.

[0059] The task scheduling method provided in this invention is applied to the configuration center side. It acquires configuration instructions that have undergone permission verification; collects multi-dimensional business scenario data based on the configuration instructions; and inputs this multi-dimensional business scenario data into a preset adaptive strategy model to generate a corresponding target adaptive strategy. The preset adaptive strategy model is trained based on rule learning and transfer learning. The target adaptive strategy is then visually simulated and verified, and the verified target adaptive strategy is sent to the server side. This solution generates a target adaptive strategy adapted to the current business scenario by using real-time collected multi-dimensional business scenario data and a preset adaptive strategy model, and sends the visually simulated and verified target adaptive strategy to the server side for scheduling and execution. This enables optimal task scheduling in dynamic business scenarios, improving the system's autonomy and maintainability.

[0060] Example 2

[0061] Figure 3 This is a flowchart of a task scheduling method provided in Embodiment 2 of the present invention. It is further optimized and extended based on the above embodiments and can be combined with various optional technical solutions in the above embodiments. For example... Figure 3 As shown in the figure, the task scheduling method provided in this embodiment 2 is applied to the configuration center side and specifically includes the following steps:

[0062] S210, Obtain the configuration command that has been verified by permissions.

[0063] In this embodiment of the invention, users (such as system developers) can access the front-end interactive interface of the configuration center through a web browser, enter their account and password for identity authentication, and then the configuration center will perform permission authentication on the user's identity information through a preset RBAC model or other methods to confirm whether the user has the permission to initiate configuration commands. If the user permission verification fails, the front-end interactive interface will display a "no operation permission" prompt and terminate the process. If the user permission verification passes, the configuration command submitted by the user will be received to trigger the task scheduling adaptive process. The configuration command may include information such as the system adaptive target, the adaptive reason, and the adaptive decision method.

[0064] S220: Collect multi-dimensional business scenario data for the current business scenario according to the configuration instructions; multi-dimensional business scenario data includes external environment data, system data, and application requirement data.

[0065] External environment data refers to external condition data that business operations depend on, such as network status, time zone clocks, and third-party interface status, used to assess the impact of external factors on business scheduling. System-specific data refers to the operational status data of the hardware and software systems that support the business, such as hardware resource load, software process status, and cluster health, used to assess the system's own capabilities in supporting task scheduling. Application requirement data refers to the operational requirements and rules of the business itself, such as business request characteristics, business rules, and target metrics, used to ensure that task scheduling conforms to business logic.

[0066] In this embodiment of the invention, the configuration center can determine relevant data collection parameters, such as collection range, accuracy, and priority, according to the adaptive target, adaptive reason, and adaptive decision-making method in the configuration instructions, thereby initiating the data collection process and collecting multi-dimensional business scenario data for the current business scenario; wherein, the multi-dimensional business scenario data may include external environment data, system own data, and application requirement data.

[0067] S230. Preprocess the multi-dimensional business scenario data to generate target data.

[0068] In this embodiment of the invention, the collected multi-dimensional business scenario data can be preprocessed by performing time stamp alignment, data cleaning, data standardization, and other preprocessing operations to obtain the corresponding target data.

[0069] S240. Input the target data sequentially into the anomaly detection module and transfer learning module of the preset adaptive strategy model to generate a target adaptive strategy that is adapted to the current business scenario. The anomaly detection module is used to identify abnormal patterns in the system operation process based on graph neural networks and preset clustering algorithms. The transfer learning module is used to generate a target adaptive strategy based on preset multi-task evolution algorithms and abnormal patterns.

[0070] The anomaly detection module refers to a functional module within a pre-defined adaptive strategy model used to identify abnormal patterns in system operation. This can be implemented using a Graph Neural Network (GNN) and a pre-defined clustering algorithm. Anomaly patterns can be understood as a set of features that deviate from the normal state of system operation. Examples include performance metrics such as short-term transaction error rates, transaction response times, CPU utilization, and API call failure rates exceeding pre-defined thresholds.

[0071] Graph Neural Networks (GNNs) are neural networks specifically designed for processing graph-structured data. They learn global relationships between nodes in a graph by aggregating features from neighboring nodes. In this embodiment, a GNN can be used to mine hidden relationships between log entries. The preset clustering algorithm can refer to a pre-configured density-based clustering algorithm, such as, but not limited to, the DBSCAN clustering algorithm. This algorithm clusters similar data into a single class by calculating the density distribution of data points, and points deviating from the cluster are identified as anomalies. This embodiment uses a preset clustering algorithm to identify abnormal patterns from log features.

[0072] The transfer learning module can refer to the functional module in the preset adaptive strategy model used to realize the adaptive transfer of scheduling strategies. It can generate new strategies based on historical adaptive strategies and current abnormal modes, and can realize knowledge transfer and optimization through preset multi-task evolution algorithms.

[0073] The pre-defined multi-task evolution algorithm is an optimization algorithm based on biological evolution theory. It can simultaneously optimize multiple related tasks (such as resource allocation, interface scheduling, etc.). Through selection, crossover, and mutation operations, it evolves new strategies adapted to the current scenario from historical strategies, taking into account multiple objectives (such as efficiency and stability).

[0074] In this embodiment of the invention, the configuration center can sequentially input the preprocessed target data into the anomaly detection module and the transfer learning module of the preset adaptive strategy model. The anomaly detection module can accurately identify the abnormal patterns in the current system operation process, and the transfer learning module can output a target adaptive strategy that dynamically adapts to the current business scenario.

[0075] Furthermore, based on the above embodiments of the invention, the anomaly detection module is specifically used for:

[0076] S11. Use the log entries generated during system operation as graph nodes, and perform structured parsing on each log entry to extract the corresponding template words and variable words;

[0077] S12. Combine and normalize the template words and variable words, and use the resulting feature vector as the initial feature of the corresponding graph node.

[0078] S13. Construct the log graph structure according to the connection relationships between the nodes in each graph;

[0079] S14. Use a graph neural network to process the initial features of nodes in the log graph structure and generate corresponding node update features;

[0080] S15. The node update features are reconstructed using a preset autoencoder to obtain the node reconstruction features;

[0081] S16. Use a preset clustering algorithm to perform density clustering on the node reconstruction features, and identify abnormal patterns in the system operation process based on the clustering results.

[0082] Template terms can refer to fixed text snippets in log entries that describe the event type, such as: API call failure, database connection timeout, etc. Variable terms can refer to dynamically changing parameter values ​​in log entries, such as: timestamps, error codes, IP addresses, etc.

[0083] A log graph structure can refer to a graph network constructed with log entries as nodes and variable relationships as edges, used to reflect the semantic relationships between logs. A pre-defined autoencoder can refer to a neural network composed of an encoder and a decoder, used to learn a compressed representation of the data and reconstruct the input, optimizing features through reconstruction loss to preserve core patterns.

[0084] In this embodiment of the invention, the specific process of using an anomaly detection module to identify abnormal modes during system operation includes:

[0085] ① Collect log entries generated during system operation (such as application logs, system kernel logs, interface call logs, etc.) and map each log entry to an independent node in the log graph structure. Perform structured parsing on each log entry, such as through regular expression matching and natural language processing, to separate fixed template words and variable words that change dynamically with the scenario, thereby achieving structured decomposition of log content.

[0086] ② For each log entry, combine its corresponding template words and variable words into a feature vector, then normalize the feature vector, and use the processed feature vector as the initial feature of the corresponding graph node.

[0087] ③ Extract the connection relationships between the nodes of each graph, that is, the association relationships between log entries. If the variable words of two logs have common parameters (such as both containing the same transaction ID, the same error code, etc.), then establish a directed or undirected edge between the two nodes, thereby connecting the isolated log nodes into a log graph structure containing association relationships, and capturing the potential logical associations between logs during system operation.

[0088] ④ The constructed log graph structure and initial node features are input into a graph neural network (as shown in the graph convolutional network GCN). Through the message passing mechanism of the graph neural network, each node aggregates the feature information of its neighboring nodes to achieve iterative updates of node features. The final generated node update features not only contain the information of a single log entry but also integrate the contextual information of related log entries, thus reflecting the system's operating status more comprehensively.

[0089] ⑤ The node update features output by the graph neural network are input into a predefined autoencoder consisting of an encoder and a decoder. The encoder compresses the node update features into low-dimensional latent vectors through dimensionality reduction mapping, and the decoder reconstructs vectors that are as close as possible to the original node update features based on the latent vectors. By minimizing the reconstruction loss (such as mean squared error), the autoencoder parameters are optimized so that the reconstructed node features (i.e., node reconstruction features) retain the core pattern of the log (normal or abnormal pattern), filter out noise interference, and enhance the discriminability of the features.

[0090] ⑥ The node reconstruction features output by the preset autoencoder are used as input, and preset clustering algorithms such as DBSCAN are used for clustering analysis. Normal logs will form high-density clusters due to their high feature similarity; while abnormal logs will become isolated points or small clusters in low-density areas because their features deviate from the normal pattern. Based on the clustering results, the log entries corresponding to isolated points or nodes that deviate from the core cluster are marked as abnormal, and their template words and variable word features are summarized to identify abnormal patterns in the system operation.

[0091] The anomaly detection module in this embodiment can identify hidden relationships across logs by using log graph structure and graph neural network, thereby improving the comprehensiveness of anomaly detection. By using log graph structure and density clustering, it can trace the associated logs of anomaly nodes, clarify the context in which the anomaly occurred, and facilitate the location of the cause of the anomaly.

[0092] Furthermore, based on the above embodiments of the invention, the transfer learning module is specifically used for:

[0093] S21. Search the preset historical business database for historical adaptive strategies that match the abnormal pattern;

[0094] S22. Call the preset multi-task evolution algorithm to optimize the historical adaptive strategy and generate a target adaptive strategy that is adapted to the current business scenario.

[0095] Among them, the preset historical business database can refer to a structured database that stores historical abnormal scenarios (historical abnormal patterns) and corresponding strategies, which is used to provide transferable historical knowledge and support the rapid generation of strategies.

[0096] In this embodiment of the invention, the specific process of generating a target adaptive strategy using a transfer learning module includes:

[0097] ① Key feature vectors can be extracted from the current anomaly mode output by the anomaly detection module, such as anomaly type (e.g., resource overload), scope of impact (e.g., involving institution A), error code, etc.

[0098] ② The cosine similarity algorithm is used to compare the key feature vector of the current abnormal pattern with the abnormal pattern feature vector of each record in the preset historical business database, and to select historical adaptive strategies with similarity greater than the preset threshold.

[0099] ③ Invoke a preset multi-task evolution algorithm to iteratively optimize the historical adaptive strategy and generate a target adaptive strategy adapted to the current business scenario. The multi-task evolution process may include:

[0100] 1) Population initialization: Decode the matched historical adaptive policies into decision vectors, such as CPU allocation rate = 0.7, cache threshold = 80%, etc.; Select the Top-K similar historical policy vectors as the initial population individuals, and inject random perturbation vectors (±5% mutation) to enhance diversity; Calculate the mean position of all historical policy vectors as the center point of the decision space for evolutionary search.

[0101] 2) Transfer strategy generation: Apply Gaussian perturbation to individuals with high fitness (such as the strategy with the lowest error rate); prioritize individuals close to the center of the decision space; generate new strategies through neighborhood search (perturbation in a small range around successful solutions) and back learning (generating symmetric solutions to expand the search space);

[0102] 3) Multi-task collaboration: Multiple related tasks are optimized in parallel, and each task independently maintains a subpopulation;

[0103] 4) Fitness evaluation: Calculate the fitness value for each new strategy based on the constraints of the current scenario and multi-objective indicators;

[0104] 5) Retain the top K% of policy individuals in terms of fitness, and repeat the crossover, mutation, and evaluation process until the iteration is terminated when any of the following conditions are met: the maximum preset number of rounds is reached (e.g., 100 rounds); the improvement of the optimal solution over N consecutive generations is less than a preset threshold; or the fitness is lower than a preset threshold.

[0105] 6) Policy output: Select the policy with the highest fitness (e.g., the lowest prediction error rate) as the target adaptive policy.

[0106] S250. Load the target adaptive strategy in the preset simulation environment to simulate the running state of the current business scenario.

[0107] The preset simulation environment can be understood as a virtual system built by the configuration center to simulate hardware resources, applications, user behaviors, etc. in real business scenarios, and to verify the execution effect of the strategy under near-real conditions.

[0108] In this embodiment of the invention, the configuration center can start a preset simulation environment, which includes virtual components consistent with the current business scenario, such as a virtual server cluster, a virtual business system, etc.; then the generated target adaptive strategy is imported into the preset simulation environment, triggering the target adaptive strategy to be executed in the simulation environment, simulating the running state of the real business scenario (such as transaction concurrency, resource load, etc.).

[0109] S260: Collect system performance indicators during the simulation process and display the system performance indicators through a visual interface for strategy simulation.

[0110] In this embodiment of the invention, a monitoring module can be configured in the preset simulation environment. This module collects system performance indicators (such as transaction error rate, transaction response time, resource utilization, etc.) during the strategy execution process at a preset frequency and visualizes them to dynamically demonstrate the execution process of the target adaptive strategy to system developers. For example, line charts can be used to display the changing trends of system indicators (such as the fluctuation curve of CPU utilization with request volume); bar charts can be used to compare the task success rate with the preset target; and the status of business process nodes can be visualized, with abnormal nodes highlighted and their root causes labeled.

[0111] S270. If the system performance indicators do not meet the preset performance indicator thresholds, adjust the model parameters of the preset adaptive strategy model and perform simulation verification on the newly generated target adaptive strategy again.

[0112] In this embodiment of the invention, the collected system performance indicators can be compared with the preset performance indicator thresholds. If the system performance indicators do not meet the preset performance indicator thresholds, it is determined that the current target adaptive strategy has not been verified. The relevant model parameters of the preset adaptive strategy model can be adjusted, and the target adaptive strategy can be regenerated based on the adjusted model. The newly generated target adaptive strategy can be simulated and verified again, i.e., return to execute S250.

[0113] S280. If the system performance indicators meet the preset performance indicator thresholds, then receive a manual confirmation instruction based on the simulation verification results of the current strategy.

[0114] In this embodiment of the invention, if the system performance indicators meet the preset performance indicator thresholds, the strategy visualization verification results are pushed to the system developers, and the developers are asked to provide corresponding manual confirmation instructions.

[0115] S290. If the manual confirmation instruction indicates that the strategy verification has failed, then adjust the model parameters of the preset adaptive strategy model and perform simulation verification again on the newly generated target adaptive strategy.

[0116] In this embodiment of the invention, if the manual confirmation instruction from the system developer indicates that the strategy verification has failed, the relevant model parameters of the preset adaptive strategy model are adjusted, and the target adaptive strategy is regenerated based on the adjusted model. The newly generated target adaptive strategy is then simulated and verified, i.e., the process returns to execution S250.

[0117] S2100. If the manual confirmation instruction indicates that the policy verification has passed, the current target adaptive policy is encapsulated into a policy package and sent to the server side.

[0118] In this embodiment of the invention, if the manual confirmation instruction from the system developer indicates that the policy verification has passed, the currently verified target adaptive policy is encapsulated into a policy package in a standardized format and sent to the server side for scheduling and execution. The policy package may include, but is not limited to, the following information: core policy content (resource allocation rules, task execution logic, exception handling process, etc.), metadata (policy version number, effective time, adaptation scenario tags, etc.), security verification information (SHA-256 hash value of policy content, digital signature of configuration center, etc.).

[0119] The task scheduling method provided in this invention is applied to the configuration center side. By collecting multi-dimensional business scenario data in real time and a preset adaptive strategy model, it generates a target adaptive strategy that adapts to the current business scenario. The target adaptive strategy that has passed the visualization simulation verification is then sent to the server side for scheduling and execution. This can achieve optimal task scheduling in dynamic business scenarios and improve the system's autonomy and maintainability.

[0120] Example 3

[0121] Figure 4 This is a flowchart of a task scheduling method provided in Embodiment 3 of the present invention. This embodiment is applicable to situations where dynamic task scheduling is performed based on an adaptive strategy of the current business scenario. The method can be executed by a task scheduling device, which can be implemented in hardware and / or software and can be configured in an electronic device. Figure 4 As shown in the figure, the task scheduling method provided in this embodiment three is applied to the server side and specifically includes the following steps:

[0122] S310, Obtain the target adaptive strategy issued by the configuration center.

[0123] In this embodiment of the invention, the server can receive the policy package encapsulated and issued by the configuration center and verify the policy package, such as verifying the integrity (hash value comparison) and authenticity (digital signature verification) of the policy package. After the verification is passed, the target adaptive policy is parsed from the policy package, and the business request will be executed based on the policy.

[0124] S320: Generate process orchestration instructions based on the target adaptive strategy and business requests.

[0125] Among them, the process orchestration instructions can be understood as a standardized task execution plan generated on the server side, which includes information such as a list of task nodes, execution sequence, resource quotas, and dependencies, and is used to guide the execution of business requests.

[0126] In this embodiment of the invention, after receiving a business request, the server can extract information such as business type, target organization identifier, and business execution dependencies. Based on the business type and target organization identifier, the server can retrieve and match the corresponding target process template in a preset process template library, and use a target adaptive strategy to adapt the parameters of the target process template, such as allocating resource quotas, adjusting interface parameters, and configuring exception handling rules. Finally, based on the adapted process template, an ordered task node topology sequence is constructed to generate a process orchestration instruction containing the task execution order.

[0127] S330: Execute the corresponding scheduling task according to the process orchestration instruction, and feed back the generated task execution result to the configuration center.

[0128] In this embodiment of the invention, the server can schedule each target organization to execute corresponding scheduling tasks according to the process orchestration instructions. After all the scheduling tasks are completed, the generated task execution results (such as execution status, task time, abnormal situations, etc.) are fed back to the configuration center so that the configuration center can adjust the parameters of the preset adaptive strategy model according to the task execution results, thereby improving the adaptability of the adaptive strategy to the business scenario.

[0129] The task scheduling method provided in this invention is applied to the server side. It obtains a target adaptive strategy issued by the configuration center; generates process orchestration instructions based on the target adaptive strategy and business requests; executes the corresponding scheduled tasks according to the process orchestration instructions; and feeds back the generated task execution results to the configuration center. This solution, by orchestrating and scheduling execution according to the target adaptive strategy issued by the configuration center in conjunction with business requests, can improve the system's adaptability to business changes and enhance the execution efficiency of task processes, achieving optimal task scheduling in dynamic business scenarios and improving the system's autonomy and maintainability.

[0130] Example 4

[0131] Figure 5 This is a flowchart of a task scheduling method provided in Embodiment 4 of the present invention. It is further optimized and extended based on the above embodiments and can be combined with various optional technical solutions in the above embodiments. For example... Figure 5As shown in the figure, the task scheduling method provided in this embodiment four is applied to the server side and specifically includes the following steps:

[0132] S410: Obtain the target adaptive strategy issued by the configuration center.

[0133] In this embodiment of the invention, the server can receive the policy package encapsulated and issued by the configuration center, and extract the target adaptive policy from the policy package after verifying it.

[0134] S420: Call the preset scheduler to parse the business request and determine the business type, target organization identifier, and business execution dependencies.

[0135] The pre-defined scheduler can be a core component on the server side used to parse business requests, match templates, and generate process orchestration instructions. It possesses a rule engine and process decomposition capabilities, serving as the central controller for task scheduling. The business type can be a classification identifier used to distinguish different businesses. The target institution identifier can be a code or ID used to uniquely identify the external institution involved in the business (such as a bank or payment platform), ensuring that process nodes are accurately associated with the corresponding institution interfaces. Business execution dependencies can be causal logical constraints between task nodes, used to ensure that scheduled tasks are executed in logical order.

[0136] In this embodiment of the invention, a preset scheduler on the server side can be invoked to perform structured parsing of the received business requests, thereby extracting the business type, target organization identifier, and business execution dependencies.

[0137] S430. Match the corresponding target process template in the preset process template library according to the business type and target organization identifier.

[0138] The preset process template library can refer to a database storing standardized business process frameworks. Each process template includes information such as template identifier, business type, and applicable organization identifier. The target process template can refer to a standardized process framework matched from the preset process template library that corresponds to the current business type and target organization.

[0139] In this embodiment of the invention, the server side is configured with a preset process template library. The preset scheduler can match the corresponding target process template in the preset process template library according to the business type and the target organization identifier, so as to ensure that the target process template is highly consistent with the business scenario.

[0140] S440. Adapt the parameters of the target process template using a target adaptive strategy.

[0141] In this embodiment of the invention, the preset scheduler can invoke a target adaptive strategy to dynamically adjust the parameters of the matched target process template, thereby adapting the target process template to the application scenario of the year. For example, resource quotas can be set for each process node in the process template according to the strategy allocation rules in the target adaptive strategy; interface parameters in the process template can be set using the organizational interface rules in the target adaptive strategy; and fault-tolerant nodes can be added to the process template using the exception handling rules in the target adaptive strategy.

[0142] S450, Decompose the adapted target process template into task nodes belonging to different target organizations.

[0143] In this embodiment of the invention, the target organization identifier can be used as the basis for splitting, and the processes of multiple organizations involved in the target process template can be split into several task nodes, with each scheduling task corresponding to a different scheduling task.

[0144] S460. Generate process orchestration instructions containing the execution order between each task node according to the business execution dependencies.

[0145] In this embodiment of the invention, the decomposed task nodes can be sorted topologically based on the parsed business execution dependencies, thereby generating a process orchestration instruction containing the execution order.

[0146] S470: Call the preset scheduler to send execution signals to the preset executor according to the process arrangement instructions.

[0147] Among them, the preset executor can refer to the component on the server side responsible for task scheduling of task nodes. It has the ability to schedule nodes, manage resources, handle exceptions, and summarize results, and is the core executor for task implementation.

[0148] In this embodiment of the invention, a preset scheduler can be invoked to parse the process orchestration instructions, extract core execution information (such as task node list, execution order, target organization identifier, resource quota, etc.), and generate standardized execution signals to drive the preset executor to execute the scheduled tasks according to the rules.

[0149] S480, in response to the execution signal, calls the preset executor to schedule and execute the scheduled tasks of each task node, and feeds back the task execution results to the configuration center.

[0150] In this embodiment of the invention, after receiving the execution signal, the preset scheduler can allocate corresponding resource quotas according to the resource requirements of each task node, and schedule each task node to execute the corresponding scheduling task according to the business execution dependency relationship. At the same time, it records the execution status and execution result of each task node. After all task nodes have completed execution, the execution results of each task node are summarized into task execution results and fed back to the configuration center. This allows the configuration center to adjust the parameters of the preset adaptive strategy model based on the task execution results, thereby improving the adaptability of the adaptive strategy to the business scenario.

[0151] The task scheduling method provided in this embodiment of the invention is applied to the server side. By following the target adaptive strategy issued by the configuration center and combining business requests for process orchestration and scheduling execution, it can improve the system's adaptability to business changes and improve the execution efficiency of task processes, achieve optimal task scheduling in dynamic business scenarios, and improve the system's autonomy and maintainability.

[0152] Example 5

[0153] Figure 6 This is a flowchart of a task scheduling method provided in Embodiment 5 of the present invention. Based on the above embodiments, this embodiment provides an implementation of the task scheduling method, which enables the perception, analysis, and processing of uncertain scenarios through repeated model training and rule learning and transfer learning using historical and real-time transaction behaviors and data in the runtime environment. This allows for rapid adaptation to unforeseen real-world changes during initial development, improving the system's autonomy. Figure 6 As shown, the task scheduling method provided in Embodiment 5 of the present invention, applied to the configuration center side, specifically includes the following steps:

[0154] S510. Verify the operator's permissions.

[0155] S520: Configure rules according to the configuration instructions that have been verified by permissions.

[0156] S530: Collect multi-dimensional business scenario data for the current business scenario.

[0157] The multi-dimensional business scenario data includes external environment data, system data, and application requirement data.

[0158] S540. Preprocess the multi-dimensional business scenario data and input the preprocessed target data into the preset adaptive strategy model to generate a target adaptive strategy that adapts to the current business scenario.

[0159] S550: Visualize and simulate the target adaptive strategy for verification.

[0160] S560. Determine whether the target adaptive strategy has been verified.

[0161] In this embodiment of the invention, if the target adaptive strategy fails to pass verification, S570 is executed; otherwise, S580 is executed.

[0162] S570. Adjust the model parameters of the preset adaptive strategy model and perform simulation verification on the newly generated target adaptive strategy again.

[0163] S580: Encapsulate the current target adaptive strategy into a strategy package and send the strategy package to the server side.

[0164] The technical solution of this invention enables the perception, analysis, and processing of uncertain scenarios. Based on the interaction experience between the system and the environment, an adaptive strategy is learned and can continue to evolve as the environment changes, deriving adaptive logic from operational data. This method, even when the knowledge of the system by architects and developers is incomplete (incompleteness can stem from many sources, such as designers lacking expert knowledge, difficulty in accurately modeling the system, and the unavailability of many key variables), utilizes machine learning technology to endow the software system with adaptive capabilities through a learning mechanism. In other words, the system automatically analyzes operational data to refine adaptive strategies, reducing the complexity of system design and development and increasing system maintainability.

[0165] Figure 7 This is a structural diagram of the configuration center side provided in Embodiment 5 of the present invention. Figure 7 As shown, the configuration center adopts a distributed architecture, including the following three modules:

[0166] ① Access Control Module: This module is based on a user-role-permission model, where users are associated with roles, and roles are associated with permissions. The database in the configuration center stores two tables: one table records the roles corresponding to each user, and the other table configures the corresponding permissions for different roles. When system developers want to modify the database, they retrieve user permissions through the aforementioned tables. If the user has the necessary permissions, the developer is allowed to perform the operation; otherwise, the developer is prompted that they do not have the required permissions.

[0167] ② Front-end Interaction Module: This module provides system developers with a configuration interface. It displays the steps that might be used according to the business logic. Developers can select specific steps for different branches and configure the adaptive goals, adaptive reasons, and adaptive decision-making methods. Simultaneously, the front-end interaction module allows for simulation model strategies, enabling operators to quickly understand the expected effects of the current task scheduling configuration.

[0168] ③ Adaptive Strategy Generation Module: This module generates adaptive strategies based on business rules, transaction data, and system behavior information. The process of determining the adaptive strategy is as follows: Figure 8 As shown, the specific process can be referred to in the above embodiments, and will not be repeated here.

[0169] Figure 9 This is a structural diagram of the server side provided in Embodiment 5 of the present invention. Figure 9 As shown, the server side includes a backend service module and a process orchestration module. The backend service module is responsible for calling the process orchestration of each bank according to business requests, executing the corresponding debugging task nodes; responding to business requests, classifying and suggesting based on monitored business behavior and system information, and outputting task execution results.

[0170] In this embodiment of the invention, the adaptive strategy includes the following two core functions:

[0171] 1. Anomaly Detection

[0172] In daily system operation, logs serve as a crucial data source for recording system status and events, carrying valuable information such as service records, fault causes, and system performance. Log-based adaptive diagnostic algorithms can automatically analyze different types of log data to assist in locating system faults and improve the efficiency and accuracy of fault diagnosis. This invention relies on system operation log data and combines machine learning and deep learning techniques to provide support for intelligent operation and maintenance work, including anomaly monitoring and application adaptation. It helps intelligent operation and maintenance systems intelligently adapt to and execute corresponding scripts based on the actual operating environment, automatically detecting and handling various anomalies that may occur during system operation. This ensures system availability and reliability in the event of a fault and improves the success rate of process execution. This is of great significance for reducing the number of system failures, shortening system fault recovery time, and improving system availability and stability.

[0173] This invention proposes an anomaly detection algorithm based on graph neural networks and density clustering. This method uses graph neural networks to learn and extract features from logs, and finally finds the features of abnormal logs based on clustering. For newly input unknown logs, embedding is performed, and the log feature vector with the closest distance is found without updating the network. The category of the known vector can be directly used as the basis for diagnosing anomalies in new logs. Considering that logs are updated over time, a batch of new vectors can also be added and the features can be re-clustered.

[0174] The detection algorithm consists of four parts: initializing the graph structure, model iteration, loss function, and density-based clustering. During training, the neural network learns to extract features from log nodes. During the evaluation phase, new node embeddings are obtained using the same embedding method as during initialization, without backpropagation.

[0175] ① Initialize the graph structure: First, perform structured parsing of the logs, classifying them into different categories according to log templates, and then establish a graph structure. A log category can consist of many log entries, where 'e' represents its template (event), corresponding to the log type; 'vax' represents the array of variable words in a log entry, the number of variable words depending on the log type. Different log categories construct different graphs, and graphs between different categories can be treated as independent problems, helping to reduce the demand on computing resources. Logs are selected as nodes, and subsequent log diagnosis is performed through node feature clustering. Log features consist of their template words and variable words. After structured parsing, the template and variable words are extracted. If the variable words are non-numeric, they are sequentially encoded, then combined into a vector, normalized, and used as the initial features of the nodes. Connections between nodes are determined based on whether variables co-occur.

[0176] ② Model Iteration: A convolutional neural network (CNN) structure is used to handle model iteration. During the information transmission phase, each node receives and processes information from its neighbors. Considering that new logs will continuously enter in real-world scenarios, it may be necessary to evict older logs. Algorithms such as LRU (Least Recently Used), FIFO (First In First Out), and LFU (Least Frequently Used) can be used to improve embedding and graph update speed. There are two methods for embedding new logs: one is to simply add them to the graph structure and then run the neural network in the evaluation state to obtain their features; the other is to backpropagate the neural network after adding them, update the parameters, and then update the features of all nodes.

[0177] ③ Loss Function: Considering that log anomaly detection is mostly an unlabeled classification task, meaning that the log data provided in actual operation is often unlabeled and cannot be used for supervised learning, the most common approach to handling unsupervised representation learning is to use a suitable decoder to reconstruct the adjacency matrix. An inner product decoder can be used. The inner product decoder is a commonly used decoder in graph neural networks, mainly used for link prediction or graph reconstruction tasks. Its core idea is to use the inner product of node embedding vectors to predict whether edges exist between nodes. The loss function in an autoencoder architecture is usually expressed as minimizing the similarity between the original adjacency matrix and the reconstructed adjacency matrix. In this way, the decoder can infer the structure of the entire graph from the node embeddings, i.e., predict whether edges exist between nodes, and use the difference between the prediction result and the original graph as the loss function.

[0178] ④ Density-based clustering: DBSCAN is a density-based clustering algorithm. Its clustering method can be described as the process of distinguishing the following types of points: (1) Core points: If there are a sufficient number of other points (min sample) within a specified radius (neighborhood radius, eps) of a point, then the point is considered a core point. These two parameters have a great influence on the clustering effect. The parameters with the largest silhouette coefficient can be found by analyzing data and exploring automatically. (2) Boundary points and noise points: Boundary points are points in the neighborhood of core points but do not meet the core point conditions themselves. Points that do not belong to core points or boundary points are considered noise points. (3) Cluster formation: The algorithm defines the region formed by the core point and its reachable core points (through direct or indirect connection) as a cluster. Considering that node feature extraction comes from the encoding of templates and variable words, and these words will establish connections between nodes based on their frequency of occurrence, the features of connected nodes will become similar during the graph iteration process. On the other hand, if we consider that most of the data in the log is stable, while anomalies are accompanied by fluctuations, the numerical part of the node features may be closer to the normal ones, while the normal and abnormal features are farther apart (this is also the training objective of autoencoders). However, at the same time, when observing log data over a larger time range, it can usually be found that it maintains density while generating a distance offset from the old data. In this case, distance-based clustering methods may lose their meaning. Therefore, the DBSCAN method is needed to cluster the log data features.

[0179] 2. Adaptive migration:

[0180] This invention primarily employs a multi-task evolutionary algorithm as the task optimizer, utilizing shared knowledge across different tasks to simultaneously optimize multiple tasks. Compared to single-task evolutionary algorithms, the advantages of multi-task evolutionary algorithms mainly include: (1) They can connect information from different domains, exhibiting better generalization ability. (2) By simultaneously processing multiple related problems, unnecessary redundant calculations can be avoided, reducing the computational cost of each problem and thus improving solution efficiency. (3) Through collaboration among multiple tasks, sharing useful knowledge from multiple tasks can effectively alleviate local optima.

[0181] To address the blindness of migration in the early stages of evolution, this invention incorporates prior knowledge of the optimization problem into the design of the migration strategy. Individuals closer to the center of the search space are more likely to approach the global optimum. Therefore, in the design of the migration strategy for multi-task optimization algorithms, migrating several individuals from the resource task that are close to the center of the decision space of the target task to guide the initial evolution of the target task should quickly guide the target task toward the optimal position. Introducing the center of the decision space in the migration strategy design accelerates the convergence of the algorithm. Furthermore, successfully migrated solutions from the previous generation have a high probability of improving the solution in the next generation of the target task. Consider using successfully migrated solutions from the previous generation and solutions similar to them in the resource task as the migration solutions for the next generation. The migration content processing mechanism mainly consists of two parts: neighborhood search and back learning. Typically, searching around an excellent solution has a high probability of finding another, even better solution. To improve the efficiency of knowledge transfer, a local search operation is performed around successfully migrated solutions with a certain probability. The search probability should not be too high, otherwise it will disrupt the original evolutionary direction of the successfully migrated solutions. Local search can be achieved by introducing Gaussian perturbations.

[0182] The task scheduling method provided in this embodiment of the invention has at least the following beneficial effects:

[0183] ① It can learn adaptive strategies that meet the requirements of system autonomy under dynamic changes in the operating environment, interactive environment, or the system's own state; and use efficient algorithms to solve problems such as search, optimization, and uncertainty analysis implicit in the adaptive behavior of software systems.

[0184] ② Adopt machine learning to construct adaptive strategies for software systems, transform the adaptive behavior of software systems into problems that machine learning excels at, such as classification and clustering, and use techniques such as unsupervised learning and reinforcement learning to obtain adaptive strategies;

[0185] ③ A unified configuration module is used to manage the adaptive model in a unified manner, providing simulation capabilities. It is easy to operate and improves readability, allowing non-developers to quickly understand the business operation process and expected results.

[0186] ④ A process orchestration module was designed on the server side, which can enable the reuse of steps.

[0187] Example 6

[0188] Figure 10 This is a schematic diagram of a task scheduling device provided in Embodiment Six of the present invention. Figure 10 As shown, the device is applied to the configuration center side and includes:

[0189] Configuration instruction acquisition module 61 is used to acquire configuration instructions that have been verified by permissions;

[0190] The strategy generation module 62 is used to collect multi-dimensional business scenario data based on configuration instructions and input the multi-dimensional business scenario data into a preset adaptive strategy model to generate a corresponding target adaptive strategy; the preset adaptive strategy model is trained based on rule learning and transfer learning.

[0191] The strategy verification and distribution module 63 is used to perform visual simulation verification of the target adaptive strategy and distribute the verified target adaptive strategy to the server side.

[0192] Furthermore, based on the above embodiments of the invention, the strategy generation module 62 includes:

[0193] The data acquisition unit is used to collect multi-dimensional business scenario data of the current business scenario according to the configuration instructions; the multi-dimensional business scenario data includes external environment data, system own data, and application requirement data.

[0194] The preprocessing unit is used to preprocess multi-dimensional business scenario data to generate target data;

[0195] The strategy generation unit is used to sequentially input target data into the anomaly detection module and transfer learning module of the preset adaptive strategy model to generate a target adaptive strategy that is adapted to the current business scenario. The anomaly detection module is used to identify abnormal patterns in the system operation process based on graph neural networks and preset clustering algorithms, and the transfer learning module is used to generate target adaptive strategies based on preset multi-task evolution algorithms and abnormal patterns.

[0196] Furthermore, based on the above embodiments of the invention, the anomaly detection module is used for:

[0197] Log entries generated during system operation are used as graph nodes, and each log entry is structured and parsed to extract the corresponding template words and variable words;

[0198] The template words and variable words are combined and normalized, and the resulting feature vector is used as the initial feature of the corresponding graph node.

[0199] Construct the log graph structure according to the connection relationships between the nodes in the graph;

[0200] A graph neural network is used to process the initial features of nodes in the log graph structure to generate corresponding node update features;

[0201] The node update features are reconstructed using a preset autoencoder to obtain the node reconstructed features;

[0202] A preset clustering algorithm is used to perform density clustering on the reconstructed features of nodes, and abnormal patterns in the system operation process are identified based on the clustering results.

[0203] Furthermore, based on the above embodiments of the invention, the transfer learning module is used for:

[0204] Search the preset historical business database for historical adaptive strategies that match the abnormal pattern;

[0205] The preset multi-task evolution algorithm is invoked to evolve and optimize the historical adaptive strategy, generating a target adaptive strategy that is adapted to the current business scenario.

[0206] Furthermore, based on the above embodiments of the invention, the strategy verification and distribution module 63 includes:

[0207] The strategy loading unit is used to load the target adaptive strategy in a preset simulation environment to simulate the running state of the current business scenario;

[0208] The simulation demonstration unit is used to collect system performance indicators during the simulation process and display the system performance indicators through a visual interface to simulate the strategy.

[0209] The first execution unit is used to adjust the model parameters of the preset adaptive strategy model and perform simulation verification on the newly generated target adaptive strategy again if the system performance index does not meet the preset performance index threshold.

[0210] The second execution unit is used to receive a manual confirmation instruction based on the simulation verification result of the current strategy if the system performance index meets the preset performance index threshold.

[0211] The third execution unit is used to adjust the model parameters of the preset adaptive strategy model and perform simulation verification on the newly generated target adaptive strategy again if the manual confirmation instruction indicates that the strategy verification has failed.

[0212] The fourth execution unit is used to encapsulate the current target adaptive policy into a policy package and send the policy package to the server side if the manual confirmation instruction is that the policy verification is successful.

[0213] The task scheduling device provided in the embodiments of the present invention can execute the task scheduling method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0214] Example 7

[0215] Figure 11 This is a schematic diagram of a task scheduling device provided in Embodiment 7 of the present invention. Figure 11 As shown, this device is applied to the server side and includes:

[0216] The strategy acquisition module 71 is used to acquire the target adaptive strategy issued by the configuration center.

[0217] The orchestration instruction generation module 72 is used to generate process orchestration instructions based on the target adaptive strategy and business requests;

[0218] The scheduling execution module 73 is used to execute the corresponding scheduling tasks according to the process orchestration instructions and to feed back the generated task execution results to the configuration center.

[0219] Furthermore, based on the above embodiments of the invention, the instruction generation module 72 includes:

[0220] The business request parsing unit is used to call the preset scheduler to parse business requests and determine the business type, target organization identifier, and business execution dependencies;

[0221] The template matching unit is used to match the corresponding target process template in the preset process template library according to the business type and target organization identifier;

[0222] The parameter adaptation unit is used to adapt the parameters of the target process template using a target adaptive strategy.

[0223] The task node determination unit is used to decompose the adapted target process template into task nodes belonging to different target organizations.

[0224] The orchestration instruction generation unit is used to generate process orchestration instructions containing the execution order between each task node according to the business execution dependencies.

[0225] Furthermore, based on the above embodiments of the invention, the scheduling execution module 73 includes:

[0226] The execution signal sending unit is used to call the preset scheduler to send execution signals to the preset executor according to the process arrangement instructions;

[0227] The scheduling and execution unit is used to respond to the execution signal, call the preset executor to schedule and execute the scheduled tasks of each task node, and feed back the task execution results to the configuration center.

[0228] The task scheduling device provided in the embodiments of the present invention can execute the task scheduling method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0229] Example 8

[0230] Figure 12A schematic diagram of an electronic device 80 that can be used to implement embodiments of the present invention is shown. 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 (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0231] like Figure 12 As shown, the electronic device 80 includes at least one processor 81 and a memory, such as a read-only memory (ROM) 82 and a random access memory (RAM) 83, communicatively connected to the at least one processor 81. The memory stores computer programs executable by the at least one processor. The processor 81 can perform various appropriate actions and processes based on the computer program stored in the ROM 82 or loaded from storage unit 88 into the RAM 83. The RAM 83 can also store various programs and data required for the operation of the electronic device 80. The processor 81, ROM 82, and RAM 83 are interconnected via a bus 84. An input / output (I / O) interface 85 is also connected to the bus 84.

[0232] Multiple components in electronic device 80 are connected to I / O interface 85, including: input unit 86, such as keyboard, mouse, etc.; output unit 87, such as various types of monitors, speakers, etc.; storage unit 88, such as disk, optical disk, etc.; and communication unit 89, such as network card, modem, wireless transceiver, etc. Communication unit 89 allows electronic device 80 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0233] Processor 81 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 81 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 81 performs the various methods and processes described above, such as task scheduling methods.

[0234] In some embodiments, the task scheduling method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 88. In some embodiments, part or all of the computer program may be loaded and / or mounted on electronic device 80 via ROM 82 and / or communication unit 89. When the computer program is loaded into RAM 83 and executed by processor 81, one or more steps of the task scheduling method described above may be performed. Alternatively, in other embodiments, processor 81 may be configured to perform the task scheduling method by any other suitable means (e.g., by means of firmware).

[0235] 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 (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), 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.

[0236] In some embodiments, the task scheduling method may be implemented as a computer program, which is implicitly included in a computer program product. When executed by a processor, the computer program implements the task scheduling method of the present invention. The computer program product can be understood as a software product that primarily implements its solution through a computer program. The computer program used to implement the method of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer program causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a standalone software package, or entirely on a remote machine or server.

[0237] In the context of this invention, 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 may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may 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 (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0238] 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 CRT (cathode ray tube) or LCD (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).

[0239] 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.

[0240] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through a communication network. 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 hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0241] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0242] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. 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 invention should be included within the scope of protection of this invention.

Claims

1. A task scheduling method, characterized in that, Applied to the configuration center side, the method includes: Obtain configuration commands that have been verified by authorization; Based on the configuration instructions, multi-dimensional business scenario data is collected, and the multi-dimensional business scenario data is input into a preset adaptive strategy model to generate a corresponding target adaptive strategy; the preset adaptive strategy model is trained based on rule learning and transfer learning. The target adaptive strategy is visualized and simulated for verification, and the verified target adaptive strategy is then sent to the server.

2. The method according to claim 1, characterized in that, The step of collecting multi-dimensional business scenario data based on the configuration instructions and inputting the multi-dimensional business scenario data into a preset adaptive strategy model to generate a corresponding target adaptive strategy includes: Collect the multi-dimensional business scenario data of the current business scenario according to the configuration instructions; the multi-dimensional business scenario data includes external environment data, system own data, and application requirement data; The multi-dimensional business scenario data is preprocessed to generate target data; The target data is sequentially input into the anomaly detection module and transfer learning module of the preset adaptive strategy model to generate the target adaptive strategy adapted to the current business scenario; the anomaly detection module is used to identify abnormal patterns in the system operation process based on graph neural networks and preset clustering algorithms, and the transfer learning module is used to generate the target adaptive strategy based on preset multi-task evolution algorithm and the abnormal patterns.

3. The method according to claim 2, characterized in that, The anomaly detection module is used for: Log entries generated during system operation are used as graph nodes, and each log entry is structured and parsed to extract corresponding template words and variable words; The template words and variable words are combined and normalized, and the resulting feature vector is used as the initial node feature of the corresponding graph node. Construct a log graph structure according to the connection relationships between the graph nodes; The graph neural network is used to process the initial features of the nodes in the log graph structure to generate corresponding node update features; The node update features are reconstructed using a preset autoencoder to obtain the node reconstructed features; The preset clustering algorithm is used to perform density clustering on the reconstructed features of the nodes, and the abnormal patterns in the system operation process are identified based on the clustering results.

4. The method according to claim 2, characterized in that, The transfer learning module is used for: Search the preset historical business database for historical adaptive strategies that match the abnormal pattern; The preset multi-task evolution algorithm is invoked to evolve and optimize the historical adaptive strategy, generating the target adaptive strategy that is adapted to the current business scenario.

5. The method according to claim 1, characterized in that, The step of visually simulating and verifying the target adaptive strategy, and then sending the verified target adaptive strategy to the server side, includes: The target adaptive strategy is loaded into a preset simulation environment to simulate the running state of the current business scenario; Collect system performance indicators during the simulation process, and display the system performance indicators through a visual interface for strategy simulation. If the system performance index does not meet the preset performance index threshold, the model parameters of the preset adaptive strategy model are adjusted, and the newly generated target adaptive strategy is simulated and verified again. If the system performance indicators meet the preset performance indicator thresholds, then a manual confirmation instruction based on the simulation verification results of the current strategy is received. If the manual confirmation instruction indicates that the strategy verification has failed, then the model parameters of the preset adaptive strategy model are adjusted, and the newly generated target adaptive strategy is simulated and verified again. If the manual confirmation instruction indicates that the policy verification is successful, then the current target adaptive policy is encapsulated into a policy package and the policy package is sent to the server side.

6. A task scheduling method, characterized in that, Applied to the server side, the method includes: Obtain the target adaptive strategy issued by the configuration center; Generate process orchestration instructions based on the target adaptive strategy and business requests; The corresponding scheduling task is executed according to the process orchestration instructions, and the generated task execution result is fed back to the configuration center.

7. The method according to claim 6, characterized in that, The step of generating process orchestration instructions based on the target adaptive strategy and business requests includes: The preset scheduler is invoked to parse the business request and determine the business type, target organization identifier, and business execution dependencies; Match the corresponding target process template in the preset process template library according to the business type and the target organization identifier; The target process template is adapted to the parameters using the target adaptive strategy. The adapted target process template is broken down into task nodes belonging to different target organizations; Based on the business execution dependencies, generate the process orchestration instructions that include the execution order between each of the task nodes.

8. The method according to claim 6, characterized in that, The step of executing the corresponding scheduling task according to the process orchestration instructions and feeding back the generated task execution result to the configuration center includes: The preset scheduler is invoked to send execution signals to the preset executor according to the process orchestration instructions; In response to the execution signal, the preset executor is invoked to schedule and execute the scheduled tasks of each task node, and the obtained task execution results are fed back to the configuration center.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the task scheduling method of any one of claims 1-5 or 6-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the task scheduling method of any one of claims 1-5 or 6-8.

Images