Data migration task scheduling method and device, equipment, storage medium and product

Abstract

The invention discloses a data migration task scheduling method and device, equipment, a storage medium and a product, relates to the technical field of networks, and is used for solving the problem that dynamic change of network topology is difficult to adapt and the problem of resource competition in the process of executing multiple data migration tasks at the same time. The method is applied to a central controller deployed with a network topology prediction model and a multi-task decision model, and comprises the following steps: acquiring historical statistical characteristics and topology change event information of a target network topology; processing the historical statistical characteristics and the topology change event information by using a network topology prediction model to obtain global development characteristics of the target network topology; obtaining task information of each data migration task and a pre-selected path corresponding to each data migration task; processing each piece of task information, each pre-selected path corresponding to each piece of task information and the global development characteristics by utilizing a multi-task decision model to obtain a target migration decision of each data migration task; and executing a corresponding target migration decision on each data migration task.

Description

Technical Field

[0001] The present invention relates to the field of network technology, and in particular to a method for scheduling data migration tasks, and also to a scheduling device, electronic equipment, computer-readable storage medium, and computer program product for data migration tasks. Background Art

[0002] In modern data center operations and large-scale data management, efficiently and securely migrating massive amounts of data from one storage source to a new destination is a common and critical task. For example, scenarios such as storage system upgrades, data center relocations, disaster recovery data synchronization, or distributing large datasets to specific computing clusters all involve large-scale data migration.

[0003] Large-scale data migration is a core operation in modern data centers, but existing solutions often struggle to cope with the dynamic changes in network topology and the complex resource competition generated when multiple migration tasks are executed simultaneously. On the one hand, network equipment failures, link upgrades, and routing policy adjustments can all cause real-time changes in the data center network topology. However, traditional migration solutions only select paths based on static topology, and performance can drop sharply or even be interrupted after topology changes. On the other hand, when multiple data migration tasks are executed concurrently, they compete for the computing resources and network bandwidth of the worker machines. The lack of an effective coordination mechanism can lead to mutual interference between tasks, unfair resource allocation, and starvation of some tasks, thereby reducing overall migration efficiency and affecting business stability.

[0004] Therefore, how to proactively perceive and adapt to dynamic changes in network topology, intelligently coordinate resource competition among multiple tasks, and thereby effectively ensure overall migration efficiency and business stability is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0005] An object of the present invention is to provide a scheduling method for data migration tasks, which can actively perceive and adapt to dynamic changes in network topology, while intelligently coordinating resource competition among multiple tasks, thereby effectively ensuring overall migration efficiency and business stability; another object of the present invention is to provide a scheduling device, electronic device, computer-readable storage medium and computer program product for data migration tasks, all of which have the above-mentioned beneficial effects.

[0006] In a first aspect, the present invention provides a method for scheduling data migration tasks, which is applied to a central controller, wherein the central controller is deployed with a network topology prediction model and a multi-task decision model, and the method includes: Obtain historical statistical characteristics of the target network topology and topology change event information; Processing the historical statistical features and the topology change event information using the network topology prediction model to obtain global development features of the target network topology; Obtaining task information of each data migration task and a preselected path corresponding to each data migration task; Using the multi-task decision model, processing each of the task information, each of the pre-selected paths corresponding to the task information, and the global development characteristics to obtain a target migration decision for each of the data migration tasks; The corresponding target migration decision is executed for each data migration task.

[0007] The historical statistical characteristics of the target network topology are obtained, including: Acquire historical working data of each network node and each network link in the target network topology; generating a historical status feature of each of the network nodes according to the historical working data of each of the network nodes; generating a historical status feature of each of the network links according to the historical working data of each of the network links; The historical statistical features of the target network topology are generated according to the historical status features of each of the network nodes and the historical status features of each of the network links.

[0008] The topology change event information of the target network topology is obtained, including: monitoring, by a software-defined network controller, topology change events in the target network topology; Collecting event information of each topology change event as topology change event information of the target network topology; The event information includes one or more combinations of the event type of the topology change event, event-related network nodes, event-related network links, and event occurrence timestamp.

[0009] The network topology prediction model is used to process the historical statistical features and topology change event information to obtain the global development features of the target network topology, including: Obtaining a current network topology map of the target network topology; In the network topology prediction model, the historical statistical features are modified using the current network topology map and the topology change event information to obtain the global development features of the target network topology.

[0010] The task information of each data migration task is obtained, including: For each data migration task, one or more combinations of source node information, target node information, migration data volume, task priority, and expected service level of the data migration task are obtained as task information of the data migration task.

[0011] The step of obtaining the preselected path corresponding to each of the data migration tasks includes: Determining a predicted network topology graph of the target network topology according to the global development characteristics; Calculating the link cost of each network link in the predicted network topology diagram according to the global development characteristics; According to the source node and the target node of each data migration task, the link cost is processed using the K shortest path algorithm to obtain a preselected path corresponding to each data migration task.

[0012] Wherein, the multi-task decision model includes a global feature processing network and a path feature processing network; Accordingly, the multi-task decision model is used to process each task information, each pre-selected path corresponding to each task information, and the global development characteristics to obtain a target migration decision for each data migration task, including: Using the global feature processing network to process each of the task information and the global development features to obtain global context features; Processing each of the preselected paths corresponding to each of the task information using the path feature processing network to obtain preselected path features; A target migration decision for each of the data migration tasks is determined according to the global context feature and the path feature.

[0013] The multi-task decision model further includes an evaluation network; and the data migration task scheduling method further includes: Using the evaluation network to evaluate the target migration decision of each data migration task to obtain each evaluation result; The multi-task decision model is optimized and updated with reference to the evaluation results and the preset reward function.

[0014] The data migration task scheduling method further includes: Obtaining worker cluster status information and dependencies and conflicts between the data migration tasks; Accordingly, the global feature processing network is used to process each of the task information and the global development features to obtain global context features, including: The global feature processing network is used to process the task information, the global development features, the worker cluster status information, and the dependency and conflict relationships between the data migration tasks to obtain the global context features.

[0015] Executing the corresponding target migration decision for each data migration task includes: Determine, according to the target migration decision, a target work machine for executing the data migration task, a target migration path for the data migration task, and migration required resources; The migration requirement resources are allocated to the data migration task, and the data migration task is executed by the target working machine according to the target migration path.

[0016] The migration requirement resources include the required bandwidth share of the data migration task; and the data migration task scheduling method further includes: Determining a shared bandwidth aggregation point in a target migration path of the data migration task; The required bandwidth share is adjusted according to the bandwidth usage and the upper limit threshold of the shared bandwidth aggregation point.

[0017] In a second aspect, the present invention further discloses a data migration task scheduling device, which is applied to a central controller, wherein the central controller is deployed with a network topology prediction model and a multi-task decision model, and the device comprises: The first acquisition module is used to obtain historical statistical characteristics of the target network topology and topology change event information; A first processing module is configured to process the historical statistical features and the topology change event information using the network topology prediction model to obtain global development features of the target network topology; A second acquisition module is used to acquire task information of each data migration task and a preselected path corresponding to each data migration task; a second processing module, configured to process each of the task information, each of the pre-selected paths corresponding to each of the task information, and the global development characteristics using the multi-task decision model to obtain a target migration decision for each of the data migration tasks; An execution module is used to execute the corresponding target migration decision for each data migration task.

[0018] In a third aspect, the present invention further discloses an electronic device, comprising: memory for storing computer programs; A processor is configured to implement the steps of any one of the above-mentioned methods for scheduling data migration tasks when executing the computer program.

[0019] In a fourth aspect, the present invention further discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the scheduling method for any one of the data migration tasks described above.

[0020] In a fifth aspect, the present invention further discloses a computer program product, comprising a computer program / instruction, which, when executed by a processor, implements the steps of any one of the above-mentioned methods for scheduling data migration tasks.

[0021] The present invention provides a scheduling method for data migration tasks, which is applied to a central controller, wherein the central controller is deployed with a network topology prediction model and a multi-task decision model, and the method comprises: obtaining historical statistical characteristics and topology change event information of a target network topology; processing the historical statistical characteristics and the topology change event information using the network topology prediction model to obtain global development characteristics of the target network topology; obtaining task information of each data migration task and a preselected path corresponding to each data migration task; processing each task information, each preselected path corresponding to each task information, and the global development characteristics using the multi-task decision model to obtain a target migration decision for each data migration task; and executing the corresponding target migration decision for each data migration task.

[0022] By applying the technical solution provided by the present invention, a network topology prediction model and a multi-task decision model are deployed in the central controller, wherein the network topology prediction model can be used to process the historical statistical characteristics and topology change event information of the target network topology to obtain its global development characteristics, thereby realizing active perception of dynamic changes in the network topology; the multi-task decision model can be used to process the task information, pre-selected paths and global development characteristics of each data migration task to obtain the target migration decision of each data migration task, thereby realizing intelligent coordination of resource competition among multiple data migration tasks. It can be seen that this technical solution can actively perceive and adapt to dynamic changes in the network topology by deploying the network topology prediction model and the multi-task decision model in the central controller, while intelligently coordinating resource competition among multiple tasks, thereby effectively ensuring overall migration efficiency and business stability.

[0023] The data migration task scheduling device, electronic device, computer-readable storage medium, and computer program product provided by the present invention also have the above-mentioned technical effects, and the present invention will not be described in detail here. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the prior art and the embodiments of the present invention, the following is a brief introduction to the drawings required for describing the prior art and the embodiments of the present invention. Of course, the drawings related to the embodiments of the present invention described below only describe some of the embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without inventive work, and the obtained other drawings also fall within the scope of protection of the present invention.

[0025] Figure 1 A flowchart of a method for scheduling data migration tasks provided by an embodiment of the present invention; Figure 2 A flowchart of another method for scheduling data migration tasks provided by an embodiment of the present invention; Figure 3 A schematic diagram of the structure of a data migration task scheduling device provided by an embodiment of the present invention; Figure 4 The present invention provides a schematic structural diagram of an electronic device. DETAILED DESCRIPTION

[0026] The core of the present invention is to provide a scheduling method for data migration tasks, which can actively perceive and adapt to dynamic changes in network topology, and intelligently coordinate resource competition among multiple tasks, thereby effectively ensuring overall migration efficiency and business stability; another core of the present invention is to provide a scheduling device, electronic device, computer-readable storage medium and computer program product for data migration tasks, all of which have the above-mentioned beneficial effects.

[0027] To more clearly and completely describe the technical solutions in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be introduced below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0028] An embodiment of the present invention provides a method for scheduling a data migration task.

[0029] Please refer to Figure 1 , Figure 1 A flow chart of a method for scheduling data migration tasks provided in an embodiment of the present invention is provided. The method for scheduling data migration tasks can be applied to a central controller in which a network topology prediction model and a multi-task decision model are deployed. The implementation process may include the following S101 to S105.

[0030] S101: Acquire historical statistical characteristics and topology change event information of a target network topology.

[0031] This step aims to capture historical statistical features and topology change event information within the target network topology. The target network topology, representing the network topology where each data migration task is currently located, consists of network nodes and network links. Network nodes include various worker nodes (source and target nodes corresponding to the data migration task), worker machines (devices used to execute the data migration task), and switches (devices used to forward the data to be migrated in the data migration task). Network links provide wired / wireless connections between network nodes.

[0032] First, historical statistical features are used to characterize the historical operating status of each network node and link within the target network topology. For example, they can indicate which network nodes or links are under long-term high load, which network nodes or network links experience significant long-term performance fluctuations, and which network nodes or network links frequently experience failures. It is understood that historical statistical features are essentially a static historical summary of the target network topology.

[0033] Specifically, historical statistical features can be extracted based on the historical working data of the target network topology. Therefore, in one possible implementation, obtaining the historical statistical features of the target network topology can include: obtaining the historical working data of each network node and the historical working data of each network link in the target network topology; generating the historical status features of each network node based on the historical working data of each network node; generating the historical status features of each network link based on the historical working data of each network link; generating the historical statistical features of the target network topology based on the historical status features of each network node and the historical status features of each network link. Furthermore, the historical working data can include but is not limited to working status, load size, bandwidth share occupancy, etc. By sorting the historical working data in chronological order, the historical status features of each network node and each network link in the target network topology can be extracted, thereby obtaining the historical statistical features of the target network topology.

[0034] Secondly, topology change events refer to changes occurring at network nodes and links within the target network topology, such as link up / down (up / down), node removal / addition, and other topology change events. It's understood that topology change events are essentially dynamic, real-time events within the target network topology.

[0035] Specifically, topology change event information can be obtained by monitoring the target network topology for change events. Therefore, in one possible implementation, obtaining the topology change event information of the target network topology can include: monitoring the topology change events in the target network topology through a software-defined network controller (SDN controller); counting the event information of each topology change event as the topology change event information of the target network topology; wherein the event information includes a combination of one or more of the event type of the topology change event, the event-related network node, the event-related network link, and the event occurrence timestamp. Similarly, in actual application scenarios, the topology change events monitored by the SDN controller can be sorted in chronological order to obtain a topology change event stream for subsequent processing.

[0036] S102: Using a network topology prediction model to process historical statistical features and topology change event information, to obtain global development features of the target network topology.

[0037] This step aims to predict the global development characteristics of the target network topology based on a network topology prediction model. The network topology prediction model is pre-built and can be directly called upon. In one possible implementation, the network topology prediction model can be built based on a TE-GCN (temporally enhanced graph convolutional network), a graph neural network specifically designed to handle dynamic changes in network topology. This TE-GCN not only learns a representation of the current network topology but also captures patterns in the evolution of the topology over time and predicts potential changes in link quality, providing forward-looking insights into network status for the multi-task decision-making model described later.

[0038] Furthermore, the global development characteristics are used to characterize the future working status of each network node and each network link in the target network topology, corresponding to the above-mentioned historical statistical characteristics. On this basis, the network topology prediction model is used to process the historical statistical characteristics and topology change event information to obtain the global development characteristics of the target network topology, which can include: obtaining the current network topology map of the target network topology; in the network topology prediction model, the current network topology map and topology change event information are used to correct the historical statistical characteristics to obtain the global development characteristics of the target network topology. It can be seen that "historical statistical characteristics" are static historical summaries and are the cornerstone of constructing the target network topology portrait; while "topology change event information" is a dynamic real-time event, which is used to make real-time / regular corrections and predictions to this cornerstone portrait.

[0039] S103: Acquire task information of each data migration task and a pre-selected path corresponding to each data migration task.

[0040] This step aims to obtain the task information and pre-selected paths for each data migration task. Among them, a data migration task is a task that requires migrating certain data from a certain network node to another network node (both network nodes are network nodes within the target network topology) based on certain business needs. It can be understood that within the target network topology, the number of data migration tasks occurring simultaneously is not unique, and complex resource competition will inevitably arise during the simultaneous execution of multiple data migration tasks. The task information and pre-selected paths for each data migration task are obtained in this step. Its purpose is to reasonably allocate resources for each data migration task, so as to effectively reduce the resource competition problem within the target network topology, and thereby effectively ensure the stability of the network system.

[0041] Obtaining task information for each data migration task may include, for each data migration task, obtaining, as the task information for the data migration task, one or more combinations of the following: source node information, target node information, migration data volume, task priority, and expected service level. The source node information and target node information may be node IP addresses, the migration data volume may be the amount of data to be migrated in the data migration task, the task priority may be the priority level of the data migration task, and the expected service level may include, for example, the bandwidth share and maximum tolerable delay expected to be achieved during the execution of the data migration task.

[0042] Among them, obtaining the pre-selected path corresponding to each data migration task may include: determining a predicted network topology diagram of the target network topology based on global development characteristics; calculating the link cost of each network link in the predicted network topology diagram based on global development characteristics; and processing each link cost using the K shortest path algorithm based on the source node and target node of each data migration task to obtain the pre-selected path corresponding to each data migration task. It is understandable that blindly screening all possible paths in the target network topology will inevitably lead to a decrease in work efficiency due to the huge workload, and this problem is particularly obvious when the target network topology is more complex. Therefore, obtaining the pre-selected path corresponding to each data migration task can preliminarily filter out some paths whose performance is obviously poor relative to the data migration task, that is, filter out a small number of high-quality candidate paths, which facilitates the subsequent selection of the target path that is most suitable for the corresponding data migration task and ensures overall migration efficiency.

[0043] In one possible implementation, calculating the link cost of each network link in the predicted network topology diagram based on the global development characteristics may include: determining a predicted available bandwidth, a predicted link delay, and a predicted link risk score for each network link in the predicted network topology diagram based on the global development characteristics; and calculating the predicted available bandwidth, the predicted link delay, and the predicted link risk score according to a link cost calculation formula to obtain the link cost of each network link. The link cost calculation formula is: ; in, represents the link cost, represents the predicted available bandwidth, represents the predicted link delay, represents the predicted link risk score, 、 、 These are adjustable weight values used to balance the importance of different factors.

[0044] S104: Using a multi-task decision model, process each task information, each pre-selected path corresponding to each task information, and global development characteristics to obtain a target migration decision for each data migration task.

[0045] This step aims to determine the target migration decision based on a multi-task decision model. The multi-task decision model is pre-built and can be directly called upon. In one possible implementation, the multi-task decision model can be constructed based on an MTR-DRL network (multi-task competitive deep reinforcement learning network). Its state space can explicitly model the resource requirements and competition between multiple tasks. Its action space allows for joint decisions on resource allocation (such as bandwidth quota and priority) for multiple concurrent tasks. Its reward function can comprehensively evaluate the overall migration efficiency and resource utilization fairness of multiple tasks.

[0046] Furthermore, the target migration decision for each data migration task, that is, the execution method for each data migration task, includes but is not limited to the target migration path of the data migration task, various types of resources that can be allocated (such as network resources, computing resources, storage resources, etc.), migration time / priority, executable work machines, etc.

[0047] Among them, the multi-task decision model may include a global feature processing network and a path feature processing network; accordingly, the multi-task decision model is used to process each task information, each pre-selected path corresponding to each task information, and the global development characteristics to obtain the target migration decision of each data migration task, which may include: using the global feature processing network to process each task information and the global development characteristics to obtain global context characteristics; using the path feature processing network to process each pre-selected path corresponding to each task information to obtain pre-selected path characteristics; determining the target migration decision of each data migration task based on the global context characteristics and path characteristics.

[0048] Specifically, the multi-task decision model may include a global feature processing network and a path feature processing network, wherein the global feature processing network is used to process each task information and global development features to obtain global context features, that is, to obtain the overall situation of the current environment; the path feature processing network is used to process each pre-selected path corresponding to each task information to obtain pre-selected path features, that is, to obtain detailed feature information of each pre-path. Finally, the two features output by the two networks are fused, and the fused information can be processed using a cross-task attention mechanism to obtain the target migration decision for each data migration task. It can be understood that all data migration tasks can share the output results of the path feature processing network. In addition, to ensure work efficiency, the global feature processing network and the path feature processing network can run synchronously.

[0049] Furthermore, the scheduling method of the data migration task may also include: obtaining the dependency and conflict relationships between the working machine cluster status information and each data migration task; accordingly, using the global feature processing network to process each task information and global development features to obtain global context features, which may include: using the global feature processing network to process each task information, global development features, working machine cluster status information, and the dependency and conflict relationships between each data migration task to obtain global context features.

[0050] To ensure the accuracy of each target migration decision and improve the overall performance of the network system, the worker cluster status information and the dependencies and conflicts between each data migration task can be incorporated into the global feature processing network for processing. This ensures the comprehensiveness of the acquired global environment information. The worker cluster status information refers to the working status of the worker clusters used to execute each data migration task, such as the current resource utilization of each worker and the resources allocated to each data migration task.

[0051] In addition, the multi-task decision model may also include an evaluation network; accordingly, the scheduling method of the data migration task may also include: using the evaluation network to evaluate the target migration decision of each data migration task to obtain each evaluation result; and optimizing and updating the multi-task decision model with reference to each evaluation result and a preset reward function.

[0052] Similarly, in order to further ensure the accuracy of each target migration decision and further improve the overall performance of the network system, an evaluation network can be added to the multi-task decision model to achieve effective evaluation of the output results of the multi-task decision model, and the evaluation results can be combined with the preset reward function to perform reverse optimization and update of the multi-task decision model. Among them, the preset reward function is a decisive indicator for evaluating the quality of the model strategy, directly reflecting the core goal of the task and constraining the behavior of the system. In one possible implementation, the preset reward function is as follows: Rt = R eff -P over -P networkQuality -P SLA - C time ; Among them, R t is the reward target, R eff Rewards efficient data transmission, P over Indicates the resource limit of the penalty worker, P networkQuality Indicates the penalty for network link congestion, P SLA Indicates that the penalty task service level agreement violates the C time represents the fixed cost that encourages the model to complete the task as quickly as possible. By applying a small fixed negative reward at each decision time step, the model is incentivized to complete all transfer tasks as quickly as possible while meeting other conditions.

[0053] S105: Execute corresponding target migration decisions for each data migration task.

[0054] This step aims to achieve the orderly execution of each target migration decision, thereby achieving the effective execution of each data migration task within the target network topology. In one possible implementation, the output of the multi-task decision model in S104 can be specifically an independent multi-task joint action, which indicates the execution order of each data migration task.

[0055] Among them, executing the corresponding target migration decision for each data migration task can include: determining the target work machine for executing the data migration task, the target migration path and migration requirement resources of the data migration task according to the target migration decision; allocating migration requirement resources to the data migration task, and executing the data migration task through the target work machine according to the target migration path.

[0056] Furthermore, the migration demand resources may include the required bandwidth share of the data migration task; accordingly, the scheduling method of the data migration task may also include: determining a shared bandwidth aggregation point in the target migration path of the data migration task; and adjusting the required bandwidth share according to the bandwidth usage and upper limit threshold of the shared bandwidth aggregation point. The required bandwidth share refers to the minimum bandwidth share required by the corresponding data migration task. Taking any target data migration task as an example, the shared bandwidth aggregation point in the target migration path of the target data migration task refers to the location in the target migration path where the target data migration task may share bandwidth with other data migration tasks, which is essentially the outbound port of the switch. Therefore, according to the bandwidth usage of the shared bandwidth aggregation point and its pre-set upper limit threshold, the actual bandwidth share of the target migration task can be dynamically adjusted between the required bandwidth share and the upper limit threshold. For example, when other data migration tasks do not occupy their bandwidth share, the actual bandwidth share of the target data migration task can be increased based on the required bandwidth share to achieve efficient utilization of bandwidth resources.

[0057] It can be seen that the scheduling method for data migration tasks provided by the embodiment of the present invention deploys a network topology prediction model and a multi-task decision model in the central controller, wherein the network topology prediction model can be used to process the historical statistical characteristics and topology change event information of the target network topology to obtain its global development characteristics, thereby realizing active perception of dynamic changes in the network topology; the multi-task decision model can be used to process the task information, pre-selected paths and global development characteristics of each data migration task to obtain the target migration decision of each data migration task, thereby realizing intelligent coordination of resource competition among multiple data migration tasks. It can be seen that this technical solution can actively perceive and adapt to dynamic changes in the network topology by deploying the network topology prediction model and the multi-task decision model in the central controller, while intelligently coordinating resource competition among multiple tasks, thereby effectively ensuring the overall migration efficiency and business stability.

[0058] An embodiment of the present invention provides another method for scheduling data migration tasks.

[0059] Please refer to Figure 2 , Figure 2 This is a flow chart of another method for scheduling data migration tasks provided by an embodiment of the present invention, and its implementation process is as follows.

[0060] 1. Data processing based on the network topology prediction model (TE-GCN model) aims to capture network topology changes in real time to generate deep features that can reflect the topology evolution trend.

[0061] 1. Get topology change event information.

[0062] (1) Obtain real-time topology change events (such as network link UP / DOWN, bandwidth adjustment, network node removal / addition, etc.) by subscribing to the SDN controller.

[0063] (2) The above topology change events form a topology change event stream, which records the event type, involved network nodes / network links, timestamps and related attributes, and obtains topology change event information.

[0064] 2. Obtain historical statistical features.

[0065] At each historical time step, a feature vector is generated for each network node and each network link in the target network topology. This feature vector contains the real-time indicators (such as load) of the corresponding time node. Based on these feature vectors, historical statistical features of N historical time steps can be obtained, such as load mean, variance, failure frequency, etc.

[0066] 3. Data processing.

[0067] (1) Topology Evolution Encoder: This is the core of the network topology prediction model. It uses a time series model based on an attention mechanism to implement data processing. It aims to learn how to adjust the embedding representation of network nodes and network links based on historical change events so that they not only reflect the current state but also implicitly predict future short-term changes. The input of the topology evolution encoder is the above-mentioned topology change event information, historical statistical features, and the physical network topology map of the network topology at the current moment.

[0068] (2) Graph Convolution / Attention Layer: The output of the topology evolution encoder is the input to the graph convolution / attention layer. Based on the network node and network link features adjusted by the topology evolution encoder, multi-layer graph convolution operations are applied to generate the final network node embedding vectors. These embedding vectors contain an understanding of the dynamics of the network topology. The output of the graph convolution / attention layer is the global evolution feature of the network topology. This global evolution feature is a high-dimensional mathematical vector that condenses the state of the network node itself and the context of the environment in the network topology.

[0069] In other words, through data processing of the network topology prediction model, adjustments to historical statistical features are made based on the current physical network topology and topology change event information, resulting in a global development feature that better reflects the future development trend of the network topology. The adjustments include: ① Increase the risk feature value: Assume that a dimension in the feature vector of a network link is specifically used to represent "instability" or "failure risk." Then, when the topology evolution encoder processes the event that the network link frequently goes up or down, it will learn to increase the value of this dimension for the network link. Lowering quality ratings: The model may also learn to lower the values representing dimensions such as “link quality,” “link reliability,” or “expected bandwidth.” ③ Impact on attention weight: In the downstream graph attention network, the model will learn to assign a very low attention weight to link features with high "instability" values. This means that when calculating network path scores or selecting network paths, the comprehensive evaluation of the path containing this network link will naturally decrease, resulting in a high probability of avoiding this network link when selecting the target migration link for the subsequent data migration task.

[0070] 2. Data processing based on the multi-task decision-making model (MTR-DRL model).

[0071] 1. Obtain global status: This includes information about all data migration tasks within the network topology, information about the worker cluster, and the output of the network topology prediction model.

[0072] (1) Relevant information of all data migration tasks in the network topology: source network nodes, target network nodes, migration data volume, migration completion ratio, currently allocated resources (such as bandwidth share, CPU share, etc.), real-time throughput, migration task priority, expected service level (such as expected bandwidth, maximum tolerable delay, etc.) and its satisfaction, dependency and conflict relationships between different data migration tasks, etc.

[0073] (2) Relevant information about the worker cluster in the network topology: the current total resource utilization of each worker (resource categories such as CPU, memory, disk I / O, total network interface bandwidth, etc.), the resources allocated to each data migration task, the gap between the resources and the preset upper threshold, etc.

[0074] 2. Data processing based on the Proximal Policy Optimization (PPO) algorithm, a reinforcement learning algorithm.

[0075] It's understandable that in the PPO algorithm, the actor network outputs actions, while the critic network evaluates the quality of those actions, guiding the optimization of the actor network. Updates to the critic network more accurately assess the performance of the actor network. The more accurate the critic network's evaluation, the better the actor network's training results. The actor-critic network structure is a classic AC framework for reinforcement learning.

[0076] (1) Actor network, including a global feature processing network module and a path feature processing network module. The global state data is processed based on the global feature processing network to obtain global context features, which represent the overall situation of the current environment; the pre-selected path is processed based on the path feature processing network to obtain pre-selected path features. The two network modules can run in parallel, outputting global context features and pre-selected path features for fusion processing, and using the cross-task attention mechanism to process the fusion information and obtain the target migration decision of each data migration task (such as the target migration path of the data migration task, the various resources that can be allocated (such as network resources, computing resources, storage resources, etc.), migration time / priority, executable work machines, etc.).

[0077] The preselected paths are not randomly selected but are based on a comprehensive path cost metric designed to initially filter out significantly poor paths. This metric combines multiple dimensions, including: ① Available bandwidth: Paths with higher remaining bandwidth are prioritized; ② Network latency: Paths with lower end-to-end physical latency are prioritized; ③ Path hop count: When other conditions are similar, paths with fewer network devices (switches) are prioritized to reduce potential failure points and processing delays; and ④ Path stability: This is key to integrating with the TE-GCN model. The link "instability" embedding or risk score output by TE-GCN is integrated into the path cost, significantly increasing the overall cost of paths containing unstable links. During the specific implementation process, a path calculation engine can be embedded in the central controller, which can periodically execute the following steps to determine K pre-selected paths: determine the latest network topology of the target network topology based on the global development characteristics, and calculate the link cost of each network link in the latest network topology; use the calculated link cost as the weight, run the K shortest path algorithm for the "source network node-worker" and "worker-target network node" of each data migration task respectively, and obtain K different pre-selected paths from the starting point to the end point with the lowest comprehensive cost.

[0078] In the target migration decision-making process for each data migration task, the bandwidth allocation can be dynamically adjusted using the HTB (Hierarchical Token Bucket) strategy, a traffic shaping tool used to implement bandwidth management and flow control in network devices. Taking a target data migration task as an example, the specific implementation process may include: First, after allocating the required bandwidth share based on the multi-task decision model, the target migration path selected for the target data migration task is used to identify the convergence points on the path that may share bandwidth with other data migration tasks, namely the outbound ports of the switch. Next, configure an HTB queue structure on the port, consisting of a root class and subclasses: ① Create a root class with its total bandwidth set to the physical upper limit of the port (for example, ceil = 10 Gbps). All traffic originating from the port will pass through this root class. ② Under the root class, create a separate subclass for each data migration task that needs to share this bandwidth. Furthermore, for the subcategory created for the target data migration task, its HTB parameters can be configured as follows: ① orate: Set to the guaranteed bandwidth (demand bandwidth share) determined by the multi-task decision model. This is the minimum guaranteed bandwidth that the target data migration task can obtain when the network is congested. ② ceil: Usually set to the physical upper limit of the port (for example, ceil = 10Gbps). This allows the target data migration task to "borrow" idle bandwidth up to the physical limit when other data migration tasks have not fully used their quota, thereby improving bandwidth utilization. Finally, the corresponding flow table rules are generated and sent to the switch. These rules can match the characteristics of the data flow of the target data migration task (such as source IP, destination IP, TCP / UDP port number, etc.) and direct all matching data packets to the HTB subcategory created for the target data migration task for queuing and sending.

[0079] Through the above steps, for multiple data migration tasks converging on the same switch port, each task's data flow is precisely directed into a dedicated queue with a specific "guaranteed bandwidth" (rate). When aggregate demand exceeds link capacity, HTB ensures that each data migration task receives at least the bandwidth specified by its rate, achieving reliable resource isolation and guaranteed quotas. When aggregate demand is not saturated, any task can exceed its rate limit and utilize idle bandwidth, achieving efficient resource utilization. This process is completely automated by the central controller based on decisions made by the multi-task decision model, achieving a closed loop between intelligent scheduling and underlying network execution.

[0080] (2) The Critic network, including the evaluation network module, is responsible for receiving the output of the Actor network and evaluating the long-term expected return under the current multi-task state, so as to achieve the optimization update of the multi-task decision model.

[0081] (3) Reward function: a decisive indicator for evaluating the quality of the model strategy, directly reflecting the core goal of the task and constraining the behavior of the system. Its main functions are: providing learning signals to guide strategy optimization; calculating cumulative returns and advantage estimates to help update the Actor network and Critic network; and driving the Critic network to learn accurate state-value functions through target returns.

[0082] 3. Real-time monitoring: Monitor the progress, resource usage, and expected service level of each data migration task as input for the next decision cycle.

[0083] It can be seen that the data migration task scheduling method provided by the embodiment of the present invention has the following technical effects: (1) Significantly improve migration resilience and efficiency in dynamic network environments: The TE-GCN module enables the system to proactively adapt to changes in network topology and even predict potential link problems, thereby selecting more stable and efficient network paths and avoiding migration interruptions or performance degradation caused by topology mutations. (2) Achieving efficient collaboration and fair resource allocation under multi-task concurrency: The MTR-DRL module can more intelligently allocate and schedule resources among multiple concurrent migration tasks through explicit modeling and joint decision-making of the multi-task competition environment, minimize interference between tasks, improve overall resource utilization, and take into account task priority and fairness to avoid blocking high-priority tasks or starving low-priority tasks; (3) Enhanced business assurance capabilities: Through refined control of worker machine resources (including allocating different shares of CPU and bandwidth to different data migration tasks) and attention to expected service levels, the system can better ensure the stability of other key businesses during the migration process; (4) Improve the level of system automation and intelligence: reduce the need for manual intervention in network changes and configuration of multi-task parameters.

[0084] An embodiment of the present invention provides a device for scheduling data migration tasks.

[0085] Please refer to Figure 3 , Figure 3 This is a schematic diagram of the structure of a data migration task scheduling device provided by the present invention. The data migration task scheduling device is applied to a central controller, and a network topology prediction model and a multi-task decision model are deployed in the central controller. The device may include: The first acquisition module 1 is used to obtain historical statistical characteristics of the target network topology and topology change event information; The first processing module 2 is used to process historical statistical features and topology change event information using a network topology prediction model to obtain global development features of the target network topology; The second acquisition module 3 is used to obtain task information of each data migration task and a preselected path corresponding to each data migration task; The second processing module 4 is used to process each task information, each pre-selected path corresponding to each task information, and the global development characteristics using a multi-task decision model to obtain a target migration decision for each data migration task; The execution module 5 is used to execute the corresponding target migration decision for each data migration task.

[0086] It can be seen that the scheduling device for data migration tasks provided by the embodiment of the present invention deploys a network topology prediction model and a multi-task decision model in the central controller, wherein the network topology prediction model can be used to process the historical statistical characteristics and topology change event information of the target network topology to obtain its global development characteristics, thereby realizing active perception of dynamic changes in the network topology; the multi-task decision model can be used to process the task information, pre-selected paths and global development characteristics of each data migration task to obtain the target migration decision of each data migration task, thereby realizing intelligent coordination of resource competition among multiple data migration tasks. It can be seen that this technical solution can actively perceive and adapt to dynamic changes in the network topology by deploying the network topology prediction model and the multi-task decision model in the central controller, while intelligently coordinating resource competition among multiple tasks, thereby effectively ensuring the overall migration efficiency and business stability.

[0087] In one embodiment of the present invention, the above-mentioned first acquisition module 1 can be specifically used to obtain the historical working data of each network node and the historical working data of each network link in the target network topology; generate the historical status characteristics of each network node based on the historical working data of each network node; generate the historical status characteristics of each network link based on the historical working data of each network link; generate the historical statistical characteristics of the target network topology based on the historical status characteristics of each network node and the historical status characteristics of each network link.

[0088] In one embodiment of the present invention, the above-mentioned first acquisition module 1 can be specifically used to monitor topology change events in the target network topology through a software-defined network controller; count the event information of each topology change event as the topology change event information of the target network topology; wherein the event information includes a combination of one or more of the event type of the topology change event, event-related network nodes, event-related network links, and event occurrence timestamps.

[0089] In one embodiment of the present invention, the above-mentioned first processing module 2 can be specifically used to obtain the current network topology map of the target network topology; in the network topology prediction model, the current network topology map and topology change event information are used to correct the historical statistical characteristics to obtain the global development characteristics of the target network topology.

[0090] In one embodiment of the present invention, the above-mentioned second acquisition module 3 can be specifically used to obtain, for each data migration task, a combination of one or more of the source node information, target node information, migration data volume, task priority, and expected service level of the data migration task as task information of the data migration task.

[0091] In one embodiment of the present invention, the above-mentioned second acquisition module 3 can be specifically used to determine the predicted network topology map of the target network topology based on the global development characteristics; calculate the link cost of each network link in the predicted network topology map based on the global development characteristics; and use the K shortest path algorithm to process each link cost according to the source node and target node of each data migration task to obtain the pre-selected path corresponding to each data migration task.

[0092] In one embodiment of the present invention, the multi-task decision model may include a global feature processing network and a path feature processing network; accordingly, the second processing module 4 may include: A first processing unit is configured to process each task information and the global development feature using a global feature processing network to obtain a global context feature; The second processing unit is configured to process each preselected path corresponding to each task information using a path feature processing network to obtain a preselected path feature; The determination unit is used to determine the target migration decision of each data migration task based on the global context features and path features.

[0093] In one embodiment of the present invention, the multi-task decision model may further include an evaluation network; Correspondingly, the above-mentioned second processing module 4 can also include a third processing unit, which is used to use the evaluation network to evaluate the target migration decision of each data migration task to obtain each evaluation result; and optimize and update the multi-task decision model with reference to each evaluation result and the preset reward function.

[0094] In one embodiment of the present invention, the data migration task scheduling device may further include a third acquisition module for acquiring the worker cluster status information and the dependency and conflict relationships between the data migration tasks; Accordingly, the above-mentioned first processing unit can be specifically used to use the global feature processing network to process each task information, global development characteristics, work machine cluster status information, and dependency and conflict relationships between each data migration task to obtain global context features.

[0095] In one embodiment of the present invention, the above-mentioned execution module 5 can be specifically used to determine the target work machine for executing the data migration task, the target migration path and the migration requirement resources of the data migration task based on the target migration decision; allocate migration requirement resources to the data migration task, and execute the data migration task through the target work machine according to the target migration path.

[0096] In one embodiment of the present invention, the migration demand resources include the required bandwidth share of the data migration task; accordingly, the above-mentioned execution module 5 can also be used to determine the shared bandwidth aggregation point in the target migration path of the data migration task; and adjust the required bandwidth share according to the bandwidth usage and upper limit threshold of the shared bandwidth aggregation point.

[0097] For an introduction to the apparatus provided by the embodiment of the present invention, please refer to the above method embodiment, and the present invention will not be elaborated here.

[0098] An embodiment of the present invention provides an electronic device.

[0099] Please refer to Figure 4 , Figure 4 This is a schematic structural diagram of an electronic device provided by the present invention, which may include: Memory 11, for storing computer programs; The processor 10 is configured to implement the steps of any of the above-mentioned methods for scheduling data migration tasks when executing a computer program.

[0100] like Figure 4 FIG. 1 is a schematic diagram of the structure of an electronic device, which may include a processor 10, a memory 11, a communication interface 12, and a communication bus 13. The processor 10, the memory 11, and the communication interface 12 communicate with each other via the communication bus 13.

[0101] In the embodiment of the present invention, the processor 10 may be a central processing unit (CPU), an application specific integrated circuit, a digital signal processor, a field programmable gate array, or other programmable logic devices.

[0102] The processor 10 may call a program stored in the memory 11 . Specifically, the processor 10 may execute operations in an embodiment of the method for scheduling a data migration task.

[0103] The memory 11 is used to store one or more programs. The programs may include program codes, which include computer operating instructions. In the embodiment of the present invention, the memory 11 stores at least a program for implementing the following functions: Obtain historical statistical characteristics of the target network topology and topology change event information; The network topology prediction model is used to process historical statistical characteristics and topology change event information to obtain the global development characteristics of the target network topology; Obtain task information for each data migration task and the preselected path corresponding to each data migration task; The multi-task decision model is used to process each task information, each pre-selected path corresponding to each task information, and the global development characteristics to obtain the target migration decision of each data migration task; Execute corresponding target migration decisions for each data migration task.

[0104] In one possible implementation, the memory 11 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required for at least one function, etc.; the data storage area may store data created during use.

[0105] In addition, the memory 11 may include a high-speed random access memory and may also include a non-volatile memory, such as at least one disk storage device or other volatile solid-state storage device.

[0106] The communication interface 12 may be an interface of a communication module, and is used to connect to other devices or systems.

[0107] Of course, it needs to be explained that Figure 4 The structure shown does not constitute a limitation on the electronic device in the embodiment of the present invention. In actual applications, the electronic device may include Figure 4 More or fewer components than shown, or combinations of certain components.

[0108] An embodiment of the present invention provides a computer-readable storage medium.

[0109] The computer-readable storage medium provided in the embodiment of the present invention stores a computer program, and when the computer program is executed by a processor, the steps of any one of the above-mentioned methods for scheduling data migration tasks can be implemented.

[0110] Among them, computer-readable storage media can be any available medium that can be stored by a computer or a data storage device such as a server or data center that integrates one or more available media. For example, it can be magnetic media (such as floppy disks, hard disks, tapes, etc.), optical media (such as DVDs) or semiconductor media (such as solid-state drives) and other media that can store computer program codes.

[0111] For an introduction to the computer-readable storage medium provided in an embodiment of the present invention, please refer to the above method embodiment, and the present invention will not elaborate on it here.

[0112] An embodiment of the present invention provides a computer program product.

[0113] The computer program product provided by the embodiment of the present invention includes a computer program / instruction. When the computer program / instruction is executed by a processor, the steps of any one of the above-mentioned methods for scheduling data migration tasks can be implemented.

[0114] Specifically, in each of the above embodiments, all or part of the embodiments may be implemented through software, hardware, firmware, or any combination thereof. When implemented through software, all or part of the embodiments may be implemented in the form of a computer program product.

[0115] The computer program product may include one or more computer programs / instructions that, when loaded and executed on a computer, may generate, in whole or in part, the processes or functions described in accordance with the embodiments of the present invention. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line, etc.) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0116] For an introduction to the computer program product provided by the embodiment of the present invention, please refer to the above method embodiment, and the present invention will not elaborate on it here.

[0117] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0118] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0119] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0120] The technical solution provided by the present invention is described in detail above. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core ideas. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present invention, the present invention can also be improved and modified in a number of ways, and these improvements and modifications also fall within the scope of protection of the present invention.

Claims

1. A method for scheduling data migration tasks, characterized in that: Applied to a central controller, the central controller is deployed with a network topology prediction model and a multi-task decision model, and the method includes: Obtain historical statistical characteristics of the target network topology and topology change event information; Processing the historical statistical features and the topology change event information using the network topology prediction model to obtain global development features of the target network topology; Obtaining task information of each data migration task and a preselected path corresponding to each data migration task; Using the multi-task decision model, processing each of the task information, each of the pre-selected paths corresponding to the task information, and the global development characteristics to obtain a target migration decision for each of the data migration tasks; The corresponding target migration decision is executed for each data migration task.

2. The method for scheduling data migration tasks according to claim 1, characterized in that: Obtain historical statistical characteristics of the target network topology, including: Acquire historical working data of each network node and each network link in the target network topology; generating a historical status feature of each of the network nodes according to the historical working data of each of the network nodes; generating a historical status feature of each of the network links according to the historical working data of each of the network links; The historical statistical features of the target network topology are generated according to the historical status features of each of the network nodes and the historical status features of each of the network links.

3. The method for scheduling data migration tasks according to claim 1, wherein: Get topology change event information of the target network topology, including: monitoring, by a software-defined network controller, topology change events in the target network topology; Collecting event information of each of the topology change events as topology change event information of the target network topology; The event information includes one or more combinations of the event type of the topology change event, event-related network nodes, event-related network links, and event occurrence timestamp.

4. The method for scheduling data migration tasks according to claim 1, wherein: The network topology prediction model is used to process the historical statistical features and topology change event information to obtain the global development features of the target network topology, including: Obtaining a current network topology map of the target network topology; In the network topology prediction model, the historical statistical features are modified using the current network topology map and the topology change event information to obtain the global development features of the target network topology.

5. The method for scheduling data migration tasks according to claim 1, wherein: Obtain task information for each data migration task, including: For each data migration task, one or more combinations of source node information, target node information, migration data volume, task priority, and expected service level of the data migration task are obtained as task information of the data migration task.

6. The method for scheduling data migration tasks according to claim 1, wherein: Obtaining the preselected paths corresponding to each of the data migration tasks includes: Determining a predicted network topology graph of the target network topology according to the global development characteristics; Calculating the link cost of each network link in the predicted network topology diagram according to the global development characteristics; According to the source node and the target node of each data migration task, the link cost is processed using the K shortest path algorithm to obtain a preselected path corresponding to each data migration task.

7. The method for scheduling data migration tasks according to any one of claims 1 to 6, characterized in that: The multi-task decision model includes a global feature processing network and a path feature processing network; Accordingly, the multi-task decision model is used to process each task information, each pre-selected path corresponding to each task information, and the global development characteristics to obtain a target migration decision for each data migration task, including: Using the global feature processing network to process each of the task information and the global development features to obtain global context features; Processing each of the preselected paths corresponding to each of the task information using the path feature processing network to obtain preselected path features; A target migration decision for each of the data migration tasks is determined according to the global context feature and the path feature.

8. The method for scheduling data migration tasks according to claim 7, characterized in that: The multi-task decision model further includes an evaluation network; and the method further includes: Using the evaluation network to evaluate the target migration decision of each data migration task to obtain each evaluation result; The multi-task decision model is optimized and updated with reference to the evaluation results and the preset reward function.

9. The method for scheduling data migration tasks according to claim 7, wherein: Also includes: Obtaining worker cluster status information and dependencies and conflicts between the data migration tasks; Accordingly, the global feature processing network is used to process each of the task information and the global development features to obtain global context features, including: The global feature processing network is used to process the task information, the global development features, the worker cluster status information, and the dependency and conflict relationships between the data migration tasks to obtain the global context features.

10. The method for scheduling data migration tasks according to claim 1, wherein: Executing the corresponding target migration decision for each data migration task includes: Determine, according to the target migration decision, a target work machine for executing the data migration task, a target migration path for the data migration task, and migration required resources; The migration requirement resources are allocated to the data migration task, and the data migration task is executed by the target working machine according to the target migration path.

11. The method for scheduling data migration tasks according to claim 10, wherein: The migration requirement resource includes a required bandwidth share of the data migration task; and the method further includes: Determining a shared bandwidth aggregation point in a target migration path of the data migration task; The required bandwidth share is adjusted according to the bandwidth usage and the upper limit threshold of the shared bandwidth aggregation point.

12. A data migration task scheduling device, characterized in that: Applied to a central controller, the central controller is deployed with a network topology prediction model and a multi-task decision model, and the device includes: The first acquisition module is used to obtain historical statistical characteristics of the target network topology and topology change event information; A first processing module is configured to process the historical statistical features and the topology change event information using the network topology prediction model to obtain global development features of the target network topology; A second acquisition module is used to acquire task information of each data migration task and a preselected path corresponding to each data migration task; a second processing module, configured to process each of the task information, each of the pre-selected paths corresponding to each of the task information, and the global development characteristics using the multi-task decision model to obtain a target migration decision for each of the data migration tasks; An execution module is used to execute the corresponding target migration decision for each data migration task.

13. An electronic device, characterized in that: include: memory for storing computer programs; A processor, configured to implement the steps of the method for scheduling data migration tasks as described in any one of claims 1 to 11 when executing the computer program.

14. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for scheduling data migration tasks according to any one of claims 1 to 11.

15. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method for scheduling data migration tasks according to any one of claims 1 to 11 are implemented.

Images