Distributed computing power dynamic scheduling method, equipment and medium

Abstract

The invention discloses a distributed computing power dynamic scheduling method and device and a medium, and the method comprises the steps: packaging heterogeneous computing power resources of a home terminal and an edge cloud node through a lightweight containerization technology, and collecting the hardware resource state data of the home terminal and the load data of the edge cloud node in real time; generating a dynamic scheduling strategy according to the task calculation type and the real-time network state of the to-be-calculated task, and splitting the to-be-calculated task into a plurality of sub-tasks according to the dynamic scheduling strategy; distributing the sub-tasks to home terminals or edge cloud nodes according with a dynamic scheduling strategy, and aggregating calculation results of the sub-tasks; and performing end-to-end encryption and fragmentation verification on the cross-domain transmitted data stream, and dynamically adjusting the task allocation permission of the home terminal based on the equipment security score.

Description

Technical Field

[0001] This application relates to the field of resource scheduling technology, and in particular to a method, device and medium for dynamic scheduling of distributed computing power. Background Technology

[0002] With the rapid development of edge computing and smart home technologies, the number of home terminal devices has surged, forming a massive, dispersed pool of computing resources. However, the computing resources of home terminals have long been idle, while edge cloud nodes face resource overload pressure due to carrying high-concurrency real-time services. Traditional centralized cloud computing architectures are unable to effectively coordinate the resources of both, resulting in a serious imbalance between the supply and demand of computing power.

[0003] While existing technologies attempt to integrate computing power from home terminals through containerization, they are limited by the differences in hardware architecture and operating systems of heterogeneous devices, making it impossible to achieve unified management and standardized measurement of resources. The lack of a resource description model leads to low efficiency in computing power collaboration between edge nodes and home terminals. In particular, when distributing tasks across domains, static scheduling strategies struggle to dynamically perceive changes in network status, resulting in significant fluctuations in task latency and insufficient efficiency in splitting complex tasks for execution.

[0004] At the security level, the open environment of home terminals lacks a dynamic trust assessment mechanism, traditional static security strategies have a high false positive rate, and cross-domain data transmission relies on general encryption protocols, which cannot effectively defend against man-in-the-middle attacks and data tampering risks. Furthermore, the lack of anomaly migration mechanisms means that business continuity cannot be guaranteed when nodes fail, further restricting the reliability of distributed computing networks. Summary of the Invention

[0005] This application provides a method, device, and medium for dynamic scheduling of distributed computing power to solve the above-mentioned technical problems.

[0006] On one hand, embodiments of this application provide a method for dynamic scheduling of distributed computing power, including: The heterogeneous computing resources of home terminals and edge cloud nodes are encapsulated using lightweight containerization technology, and the hardware resource status data of the home terminals and the load data of the edge cloud nodes are collected in real time. Based on the task computation type and real-time network status of the task to be computed, a dynamic scheduling strategy is generated, and the task to be computed is divided into multiple sub-tasks according to the dynamic scheduling strategy. The subtasks are distributed to home terminals or edge cloud nodes that conform to the dynamic scheduling strategy, and the subtask calculation results are aggregated. End-to-end encryption and fragmentation verification are implemented for cross-domain data streams, and task allocation permissions for home terminals are dynamically adjusted based on device security scores.

[0007] In one implementation of this application, a dynamic scheduling strategy is generated based on the task computation type and real-time network status of the task to be computed, specifically including: The system acquires historical service traffic, current terminal online rate, and real-time network status, and inputs the historical service traffic and current terminal online rate into a preset computing power demand prediction model to output the computing power demand for future periods. The task computation type of the task to be computed is determined from the computing power demand in the future time period, and the scheduling path corresponding to each home terminal is calculated based on the hardware resource status data, the load data and the real-time network status. Among multiple scheduling paths, home terminals with low-latency scheduling paths are matched for latency-sensitive computing tasks, while edge cloud nodes are allocated for computationally intensive computing tasks.

[0008] In one implementation of this application, the task to be computed is divided into multiple subtasks according to the dynamic scheduling strategy, specifically including: If the load data of the edge cloud node exceeds the preset load threshold, the shortest path algorithm is invoked to calculate the cross-domain distribution path; Data-dependent subtask groups are assigned to the same home terminal cluster, and intermediate computation results are synchronized through a two-phase commit protocol.

[0009] In one implementation of this application, after distributing subtasks to home terminals or edge cloud nodes that conform to the dynamic scheduling strategy, the method further includes: During the execution of subtasks, the isolated forest algorithm is used to detect whether there are any abnormal states in the corresponding home terminals or edge cloud nodes. If an abnormal state is detected, it is further determined whether the terminal occupancy rate exceeds the preset occupancy threshold. If the terminal resource occupancy rate exceeds a preset occupancy threshold, a migration mechanism is triggered to migrate unfinished subtasks to a backup node. Record the migration path of the subtask and update the task state machine corresponding to the subtask.

[0010] In one implementation of this application, end-to-end encryption and fragmentation verification are performed on the cross-domain transmitted data stream, specifically including: A two-way authentication channel is established between the edge management layer and the home terminal, and the session key is dynamically updated using the national cryptographic algorithm to achieve end-to-segment encryption. Establish a two-way authentication channel between the edge management layer and home terminals; The system generates hash check values ​​for the transmitted data fragments and verifies the integrity of the fragmented data based on the hash check values ​​during the aggregation phase of the subtask calculation results. Data fragments that fail verification are discarded, and a retransmission mechanism is triggered to retransmit the data fragments that fail verification.

[0011] In one implementation of this application, the task allocation permissions of the home terminal are dynamically adjusted based on the device security score, specifically including: The device's security status and historical behavior records are input into a pre-trained dynamic trust scoring model to calculate the device trust score. If the device trust score is lower than a preset score threshold, the task queue of the corresponding home terminal will be isolated. After the attack event ends, the isolation status of the task queue corresponding to the home terminal is reset, and the device integrity of the home terminal is re-verified.

[0012] In one implementation of this application, the heterogeneous computing resources of home terminals and edge cloud nodes are encapsulated using lightweight containerization technology, and the hardware resource status data of the home terminals and the load data of the edge cloud nodes are collected in real time, specifically including: Deploy a lightweight access agent of the KubeEdge framework on home terminals and deploy a resource isolation environment for container runtime; The system periodically acquires hardware resource status data from the home terminal and real-time load data from the edge cloud node. The hardware resource status data includes CPU / GPU computing power, memory usage, and network bandwidth data, while the load data includes load rate, unprocessed task queue depth, and network latency data.

[0013] In one implementation of this application, it further includes: The historical load data and task completion rate data of each edge cloud node are aggregated, and a computing power demand prediction model is trained using an LSTM-GRU hybrid neural network. The trained model parameters are encrypted and synchronized to the cloud coordination layer, and the weight coefficients of the scheduling strategy are dynamically adjusted based on real-time task execution feedback.

[0014] On the other hand, embodiments of this application also provide a distributed computing power dynamic scheduling device, the device comprising: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which are then executed by the at least one processor to enable the at least one processor to perform a distributed computing power dynamic scheduling method as described above.

[0015] On the other hand, this application also provides a non-volatile computer storage medium storing computer-executable instructions, which, when executed, implement the distributed computing power dynamic scheduling method described above.

[0016] This application provides a method, device, and medium for dynamic scheduling of distributed computing power, which has at least the following beneficial effects: By unifying the heterogeneous computing resources of home terminals and edge cloud nodes, standardized management of devices with multiple architectures and operating systems is achieved, solving the resource fragmentation problem caused by hardware differences, improving the utilization rate of idle computing power of home terminals, and effectively alleviating the resource overload pressure of edge nodes. Through intelligent task splitting and optimal node matching mechanisms, latency-sensitive tasks are prioritized to low-latency home terminals, while computationally intensive tasks are allocated to edge cloud nodes. Combined with data fragmentation hash verification and retransmission mechanisms, cross-domain transmission reliability is ensured, the average task response latency is reduced, and signaling transmission latency is stabilized. End-to-end encrypted transmission and data fragmentation verification technology are adopted to resist man-in-the-middle attacks and data tampering risks. A dynamic trust scoring model is used to evaluate the security status of devices in real time, and low-scoring devices are automatically isolated from the task queue, reducing the false positive rate of unauthorized device access. During the execution of subtasks, abnormal node status is monitored in real time, and when resource consumption exceeds the threshold, task migration is automatically triggered and the state machine is updated to ensure business continuity in fault scenarios. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating a distributed computing power dynamic scheduling method provided in an embodiment of this application; Figure 2 This is a schematic diagram of the internal structure of a distributed computing power dynamic scheduling device provided in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0020] Figure 1This is a flowchart illustrating a distributed computing power dynamic scheduling method provided in an embodiment of this application.

[0021] The analysis method involved in the embodiments of this application can be implemented by a terminal device or a server, and this application does not impose any special limitations on it. For ease of understanding and description, the following embodiments are all described in detail using a server as an example.

[0022] It should be noted that the server can be a single device or a system composed of multiple devices, i.e., a distributed server. This application does not make any specific limitations on this.

[0023] like Figure 1 As shown in the figure, an embodiment of this application provides a method for dynamic scheduling of distributed computing power, including: Step 101: Encapsulate the heterogeneous computing resources of home terminals and edge cloud nodes using lightweight containerization technology, and collect hardware resource status data of home terminals and load data of edge cloud nodes in real time.

[0024] This embodiment primarily addresses the issue of unified management of heterogeneous computing resources between home terminals and edge cloud nodes. It is understood that home terminal devices, such as smart TVs, NAS devices, and personal computers, typically have diverse hardware architectures and operating systems, making efficient resource integration difficult with traditional virtualization technologies.

[0025] It should be noted that this application employs lightweight containerization technology as the core encapsulation method. Specifically, a lightweight access proxy based on the KubeEdge framework is deployed on the home terminal. This proxy manages the container runtime environment through a standard OCI interface. For example, when the home terminal is an ARM-based smart gateway, the system automatically loads the adapted container image to ensure cross-platform compatibility of the computing power service.

[0026] Understandably, resource isolation is crucial to ensuring local business operations are not affected. In this embodiment, CPU and memory quotas are dynamically allocated through the cgroups mechanism. When high local load is detected on the terminal, the resource usage ratio of computing tasks is automatically reduced. It should be noted that this elastic isolation mechanism avoids resource conflicts while maximizing the utilization of idle computing power.

[0027] Specifically, real-time data acquisition includes two aspects. For home terminals, it involves periodically acquiring hardware resource status data such as CPU / GPU computing power and memory usage. For edge cloud nodes, it involves collecting key indicators such as load rate and task queue depth through efficient communication protocols. For example, this data is standardized to form a unified resource description model, providing a basis for subsequent dynamic scheduling decisions.

[0028] In this embodiment, a three-tiered distributed architecture is constructed, consisting of a home terminal layer, an edge management layer, and a cloud coordination layer, to achieve full-domain interconnection and hierarchical management of computing resources. For the home terminal layer, a lightweight computing power access module is deployed to support access from heterogeneous terminal devices such as PCs, smart TVs, and NAS. The home computing power unit is encapsulated through Docker containerization technology, with standardized computing power output interfaces, and real-time collection of resource status data such as device CPU / GPU / memory.

[0029] For the edge management layer, an edge computing hub node is deployed, which includes a resource management center, an intelligent scheduling engine, and a security protection module. The resource management center builds a global computing resource pool through a unified resource description language (such as the YAML extension protocol) to achieve unified modeling and status monitoring of computing power between home terminals and edge nodes; the intelligent scheduling engine executes localized scheduling strategies based on real-time load data to complete task splitting, computing power matching, and result aggregation. For the cloud coordination layer, a global computing power scheduling brain is built. Through a federated learning mechanism, the computing power demand prediction models of edge nodes are aggregated to generate cross-domain collaborative scheduling strategies to solve the problem of load imbalance among edge nodes. At the same time, upper-layer application interfaces such as computing power transaction settlement and service quality monitoring are provided.

[0030] Step 102: Generate a dynamic scheduling strategy based on the task calculation type and real-time network status of the task to be calculated, and split the task to be calculated into multiple sub-tasks according to the dynamic scheduling strategy.

[0031] This embodiment focuses on addressing the dynamic matching of tasks and resources. It is understood that different types of computing tasks have significantly different requirements for computing resources.

[0032] It should be noted that the scheduling strategy generation involves two stages: prediction and decision-making. Specifically, the system first inputs historical business traffic and terminal online rates into the computing power demand prediction model, and outputs the computing power demand trend for future periods. For example, in a smart community scenario, the model can predict that the workload of security video analysis tasks will increase significantly at night, thereby adjusting the resource allocation strategy in advance.

[0033] Understandably, task splitting strategies need to consider hardware constraints and network conditions. In this embodiment, for data-dependent tasks, such as industrial time-series data analysis, the system employs a two-phase commit protocol to ensure data consistency between subtasks. When the edge node load is too high, the shortest path algorithm is invoked to calculate the optimal cross-domain distribution path. It should be noted that this differentiated processing method significantly improves the reliability of task execution.

[0034] Specifically, the determination of task computation type is based on multi-dimensional feature analysis. For example, real-time interactive tasks are automatically marked as latency-sensitive, such as AR rendering, and are preferentially scheduled to low-latency home terminals. Batch computation tasks, on the other hand, are classified as computationally intensive, such as deep learning training, and are mainly allocated to edge cloud nodes.

[0035] In this embodiment, a computing power demand prediction model based on an LSTM-GRU hybrid neural network is constructed. By inputting historical business traffic, terminal online rate, time period characteristics and other data, the model can accurately predict computing power demand in future time periods (such as 30 minutes) and trigger the resource preheating mechanism in advance.

[0036] Based on load balancing algorithms, such as the improved minimum completion time priority algorithm, home terminals with network latency <20ms and resource utilization <70% within the same area are given priority to perform lightweight computing tasks.

[0037] When the load data of the edge cloud node exceeds the preset load threshold, the cloud coordination layer calculates the optimal cross-edge domain scheduling path through the Dijkstra algorithm, and combines it with task splitting technology, such as the lightweight implementation of MapReduce, to decompose complex tasks into parallel subtasks and distribute them to different home terminal clusters.

[0038] Step 103: Distribute the subtasks to home terminals or edge cloud nodes that comply with the dynamic scheduling strategy, and aggregate the subtask calculation results.

[0039] This embodiment focuses on addressing the execution reliability issue of distributed computing. It is understandable that ensuring continuous task execution in a heterogeneous environment presents numerous challenges.

[0040] It should be noted that the task distribution process employs a tiered protection mechanism. Specifically, the system first establishes a state machine for each subtask to track its lifecycle. When a device anomaly is detected using the Isolation Forest algorithm, circuit breaker protection is immediately triggered. For example, if a home terminal frequently experiences computation timeouts, its tasks will be migrated to a backup node, and the relevant state records will be updated simultaneously.

[0041] Understandably, result aggregation requires addressing data consistency issues in a distributed environment. In this embodiment, for critical computational tasks, the system employs a version number verification mechanism to ensure the temporal correctness of intermediate results. It should be noted that this design is particularly suitable for machine learning tasks requiring multiple iterations.

[0042] Specifically, resource usage monitoring is the foundation of dynamic scheduling. For example, when the CPU usage of a terminal continuously exceeds a safe threshold, the system will not only migrate the current task, but also automatically reduce the priority of subsequent tasks assigned to that device until the resource status returns to normal.

[0043] Step 104: Implement end-to-end encryption and fragmentation verification for cross-domain data streams, and dynamically adjust the task allocation permissions of home terminals based on device security scores.

[0044] This embodiment primarily addresses security issues in distributed computing power sharing. It is understood that data protection and device trust management are crucial in open network environments.

[0045] It should be noted that secure transmission employs a layered encryption strategy. Specifically, a two-way TLS authentication channel is established between the edge management layer and the home terminal, and the session key is periodically rotated using a national cryptographic algorithm. For example, for high-volume data such as video streams, the system implements segmented encrypted transmission, with each segment accompanied by an independent hash checksum.

[0046] Understandably, a dynamic trust assessment mechanism is the core of security protection. In this embodiment, the system continuously monitors the security status and behavioral characteristics of devices, and calculates real-time trust through a pre-trained scoring model. Security status includes, for example, the antivirus software version or system patch version, while historical behavior records include, for example, task completion rate and task anomaly frequency. It should be noted that when a device's score falls below a threshold, its task queue will be immediately isolated until it passes integrity verification.

[0047] Specifically, the security policy implementation balances strictness and flexibility. For example, for devices that have been attacked but have been repaired, the system supports restoring their computing power service permissions after manual review, while simultaneously setting an observation period to monitor their subsequent behavior. This design ensures system security while avoiding resource waste.

[0048] In one embodiment of this application, the distributed computing power dynamic scheduling method further performs an update operation on the cross-domain collaborative model. It is understood that the model update mechanism constitutes a continuous optimization closed loop for the scheduling strategy.

[0049] Specifically, the aggregation operation is implemented through a log collection agent deployed at the edge management layer. Historical load data (including CPU / GPU utilization and peak memory usage) and task completion rate data (such as the success rate / timeout ratio) are periodically (e.g., every 10 minutes) pulled from each edge cloud node, and a unified dataset across nodes is generated by timestamp alignment. For example, in a smart home scenario, a surge in security video analytics tasks at night causes a periodic increase in edge node load. The load data collected at this time will contain significant time-varying characteristics, providing key temporal patterns for model training.

[0050] It's important to note that the LSTM-GRU hybrid neural network serves as the core architecture of the computing power demand prediction model. The LSTM layer is responsible for capturing long-term load patterns, such as the traffic differences between weekdays and holidays, while the GRU layer learns short-term sudden fluctuations, such as resource contention caused by the mass deployment of home terminals. During training, the input data is divided into training and validation sets in a 7:3 ratio, and the network weights are optimized through backpropagation. Specifically, the model outputs a multi-dimensional demand vector for future periods, including computing resource requirements (such as TFLOPS), storage IOPS, and network bandwidth estimates. This vector drives the scheduling engine to pre-allocate resources. For example, if it predicts that AR games are about to enter peak periods, rendering tasks can be distributed in advance to home terminal clusters with sufficient idle GPU computing power.

[0051] Furthermore, the encrypted synchronization of model parameters employs a layered security mechanism. The gradient parameters of the trained model are first encrypted using the national cryptographic algorithm SM4, and then re-encapsulated using a session key negotiated between the edge nodes and the cloud coordination layer, forming a double-encrypted data packet. The synchronization process is transmitted through a dedicated channel with two-way certificate authentication to prevent man-in-the-middle attacks on model parameters. After the cloud coordination layer decrypts the parameters, it immediately updates the global scheduling model library and controls the hot replacement of models on edge nodes through a version number mechanism.

[0052] It is important to emphasize that the dynamic adjustment of weighting coefficients is based on constructing a negative feedback loop through real-time task execution feedback. The weighting coefficients of the scheduling strategy include latency weight, energy consumption weight, and cost weight, with initial values ​​preset by operations and maintenance personnel. When aggregating subtask results, if the completion latency of a certain type of task (such as industrial real-time control) continuously exceeds the expected threshold, the latency weighting coefficient is automatically increased in priority. Conversely, if the energy consumption of home terminals surges, the energy consumption weighting factor is increased. For example, during periods of network jitter caused by heavy rain, the system prioritizes low-latency scheduling of critical services by reducing the cost weight and increasing the latency weight.

[0053] In this embodiment, in a smart home scenario, a community constructs a regional distributed computing network to distribute sub-tasks of community security video analytics to surrounding home terminals for processing. Compared to traditional cloud solutions, video analytics latency is significantly reduced, home terminal computing power utilization is significantly improved, and edge node equipment procurement costs are reduced. In an industrial IoT scenario, a smart manufacturing park utilizes home terminal resources to process industrial equipment operation and maintenance data during non-production periods, enabling distributed training of equipment fault prediction models. This not only shortens model training time and alleviates edge node computing power constraints but also brings benefits to home users.

[0054] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide a distributed computing power dynamic scheduling device, the structure of which is as follows: Figure 2 As shown.

[0055] Figure 2 This is a schematic diagram of the internal structure of a distributed computing power dynamic scheduling device provided in an embodiment of this application. Figure 2 As shown, the device includes: At least one processor; And, a memory that is communicatively connected to at least one processor; The memory stores instructions that can be executed by at least one processor, and the instructions, when executed by at least one processor, enable at least one processor to: The heterogeneous computing resources of home terminals and edge cloud nodes are encapsulated using lightweight containerization technology, and the hardware resource status data of home terminals and the load data of edge cloud nodes are collected in real time. Based on the task computation type and real-time network status of the task to be computed, a dynamic scheduling strategy is generated, and the task to be computed is split into multiple subtasks according to the dynamic scheduling strategy. Distribute subtasks to home terminals or edge cloud nodes that conform to the dynamic scheduling strategy, and aggregate the subtask calculation results; End-to-end encryption and fragmentation verification are implemented for cross-domain data streams, and task allocation permissions for home terminals are dynamically adjusted based on device security scores.

[0056] This application also provides a non-volatile computer storage medium storing computer-executable instructions, which, when executed, can: The heterogeneous computing resources of home terminals and edge cloud nodes are encapsulated using lightweight containerization technology, and the hardware resource status data of home terminals and the load data of edge cloud nodes are collected in real time. Based on the task computation type and real-time network status of the task to be computed, a dynamic scheduling strategy is generated, and the task to be computed is split into multiple subtasks according to the dynamic scheduling strategy. Distribute subtasks to home terminals or edge cloud nodes that conform to the dynamic scheduling strategy, and aggregate the subtask calculation results; End-to-end encryption and fragmentation verification are implemented for cross-domain data streams, and task allocation permissions for home terminals are dynamically adjusted based on device security scores.

[0057] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.

[0058] The devices and media provided in this application are one-to-one with the methods. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0059] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0060] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0061] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0062] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0063] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0064] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0065] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0066] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0067] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for dynamic scheduling of distributed computing power, characterized in that, The method includes: The heterogeneous computing resources of home terminals and edge cloud nodes are encapsulated using lightweight containerization technology, and the hardware resource status data of the home terminals and the load data of the edge cloud nodes are collected in real time. Based on the task computation type and real-time network status of the task to be computed, a dynamic scheduling strategy is generated, and the task to be computed is divided into multiple sub-tasks according to the dynamic scheduling strategy. The subtasks are distributed to home terminals or edge cloud nodes that conform to the dynamic scheduling strategy, and the subtask calculation results are aggregated. End-to-end encryption and fragmentation verification are implemented for cross-domain data streams, and task allocation permissions for home terminals are dynamically adjusted based on device security scores.

2. The distributed computing power dynamic scheduling method according to claim 1, characterized in that, Based on the task computation type and real-time network status, a dynamic scheduling strategy is generated, specifically including: The system acquires historical service traffic, current terminal online rate, and real-time network status, and inputs the historical service traffic and current terminal online rate into a preset computing power demand prediction model to output the computing power demand for future periods. The task computation type of the task to be computed is determined from the computing power demand in the future time period, and the scheduling path corresponding to each home terminal is calculated based on the hardware resource status data, the load data and the real-time network status. Among multiple scheduling paths, home terminals with low-latency scheduling paths are matched for latency-sensitive computing tasks, while edge cloud nodes are allocated for computationally intensive computing tasks.

3. The distributed computing power dynamic scheduling method according to claim 1, characterized in that, According to the dynamic scheduling strategy, the task to be computed is divided into multiple subtasks, specifically including: If the load data of the edge cloud node exceeds the preset load threshold, the shortest path algorithm is invoked to calculate the cross-domain distribution path; Data-dependent subtask groups are assigned to the same home terminal cluster, and intermediate computation results are synchronized through a two-phase commit protocol.

4. The distributed computing power dynamic scheduling method according to claim 1, characterized in that, After distributing subtasks to home terminals or edge cloud nodes that conform to the dynamic scheduling policy, the method further includes: During the execution of subtasks, the isolated forest algorithm is used to detect whether there are any abnormal states in the corresponding home terminals or edge cloud nodes. If an abnormal state is detected, it is further determined whether the terminal occupancy rate exceeds the preset occupancy threshold. If the terminal resource occupancy rate exceeds a preset occupancy threshold, a migration mechanism is triggered to migrate unfinished subtasks to a backup node. Record the migration path of the subtask and update the task state machine corresponding to the subtask.

5. The distributed computing power dynamic scheduling method according to claim 1, characterized in that, End-to-end encryption and fragmentation verification are implemented for cross-domain data streams, specifically including: A two-way authentication channel is established between the edge management layer and the home terminal, and the session key is dynamically updated using the national cryptographic algorithm to achieve end-to-segment encryption; Establish a two-way authentication channel between the edge management layer and home terminals; The system generates hash check values ​​for the transmitted data fragments and verifies the integrity of the fragmented data based on the hash check values ​​during the aggregation phase of the subtask calculation results. Data fragments that fail verification are discarded, and a retransmission mechanism is triggered to retransmit the data fragments that fail verification.

6. The distributed computing power dynamic scheduling method according to claim 1, characterized in that, Based on device security scores, task allocation permissions for home devices are dynamically adjusted, specifically including: The device's security status and historical behavior records are input into a pre-trained dynamic trust scoring model to calculate the device trust score. If the device trust score is lower than a preset score threshold, the task queue of the corresponding home terminal will be isolated. After the attack event ends, the isolation status of the task queue corresponding to the home terminal is reset, and the device integrity of the home terminal is re-verified.

7. The distributed computing power dynamic scheduling method according to claim 1, characterized in that, The heterogeneous computing resources of home terminals and edge cloud nodes are encapsulated using lightweight containerization technology, and the hardware resource status data of the home terminals and the load data of the edge cloud nodes are collected in real time, specifically including: Deploy a lightweight access agent of the KubeEdge framework on home terminals and deploy a resource isolation environment for container runtime; The system periodically acquires hardware resource status data from the home terminal and real-time load data from the edge cloud node. The hardware resource status data includes CPU / GPU computing power, memory usage, and network bandwidth data, while the load data includes load rate, unprocessed task queue depth, and network latency data.

8. The distributed computing power dynamic scheduling method according to claim 1, characterized in that, The method further includes: The historical load data and task completion rate data of each edge cloud node are aggregated, and a computing power demand prediction model is trained using an LSTM-GRU hybrid neural network. The trained model parameters are encrypted and synchronized to the cloud coordination layer, and the weight coefficients of the scheduling strategy are dynamically adjusted based on real-time task execution feedback.

9. A distributed computing power dynamic scheduling device, characterized in that, The device includes: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform a distributed computing power dynamic scheduling method as described in any one of claims 1-8.

10. A non-volatile computer storage medium storing computer-executable instructions, characterized in that, When the computer-executable instructions are executed, a distributed computing power dynamic scheduling method as described in any one of claims 1-8 is implemented.

Images