Resource scheduling method and device, equipment, storage medium and computer program product

Abstract

The invention provides a resource scheduling method and device, equipment, a storage medium and a computer program product, and relates to the technical field of communication, the method comprises the following steps: in response to a task unloading request sent by terminal equipment, issuing a to-be-unloaded task carried by the task unloading request to adjacent MEC equipment; receiving an evaluation result of the to-be-unloaded task sent by the adjacent MEC equipment; the evaluation result is used for representing whether the adjacent MEC equipment has the capability of processing the to-be-unloaded task or not; and determining at least one target MEC device according to the energy consumption information and the time delay information of the MEC device under the condition of determining that the to-be-unloaded task needs to be migrated based on the evaluation result, and scheduling the to-be-unloaded task to the target MEC device. According to the invention, the task does not need to be unloaded to the edge device from the cloud center node, the local edge device receives the task request, distributes the calculation strategy and autonomously completes resource allocation, and decentralized resource scheduling is realized.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a resource scheduling method, device, equipment, storage medium and computer program product. Background Art

[0002] At present, the main method of resource scheduling is to use a centralized data processing architecture, where data is transmitted from the user to the cloud, and then processed by the cloud before being returned. This architecture has the following problems for the real-time processing needs of Internet of Things (IoT) devices: Insufficient real-time performance: The centralized architecture causes significant data transmission delays and cannot meet the real-time requirements of IoT devices; Poor robustness: If a central node fails, the entire system may collapse, lacking the robustness brought by decentralization; Unreasonable resource utilization: The central node may be overloaded due to task concentration, while the computing power of the edge nodes is not fully utilized. Therefore, how to achieve decentralized resource scheduling has become an urgent problem to be solved. Summary of the invention

[0003] The present invention provides a resource scheduling method, apparatus, device, storage medium and computer program product to solve the defects of how to realize decentralized resource scheduling in the prior art and realize decentralized resource scheduling.

[0004] The present invention provides a resource scheduling method, comprising the following steps: In response to a task offloading request sent by a terminal device, publishing the task to be offloaded carried in the task offloading request to a neighboring MEC device; Receiving an evaluation result of the task to be offloaded sent by the neighboring MEC device; the evaluation result is used to indicate whether the neighboring MEC device has the ability to process the task to be offloaded; When it is determined based on the evaluation result that the task to be offloaded needs to be migrated, at least one target MEC device is determined according to the energy consumption information and latency information of the MEC device, and the task to be offloaded is scheduled to the target MEC device.

[0005] According to a resource scheduling method provided by the present invention, the energy consumption information includes an energy consumption weight coefficient and the energy consumed by the MEC device to process the task to be offloaded; the delay information includes a time consumption weight coefficient and the total time consumed by the MEC device to process the task to be offloaded; Determine at least one target MEC device according to the energy consumption information and latency information of the MEC device, including: Determine an adaptability function according to the energy consumption weight coefficient, the energy consumed by the MEC device to process the task to be offloaded, the time consumption weight coefficient and the total time consumed by the MEC device to process the task to be offloaded; the adaptability function satisfies the balance between minimizing energy consumption and satisfying the quality of service QoS; Based on the adaptability function, searching the MEC device to determine an optimal combined offloading strategy; Based on the optimal combined offloading strategy, at least one of the target MEC devices is determined.

[0006] According to a resource scheduling method provided by the present invention, the MEC device is searched based on the adaptability function to determine the optimal combined unloading strategy, including: Encode the MEC devices in the idle state and mark them as population particles for initialization to obtain the MEC device population; Initialize a sub-population for a single MEC device in an idle state to obtain a sub-population; According to the adaptive function, updating the positions of the leader and the followers in the sub-population; According to the fitness value of each sub-individual in the sub-population after the position is updated, the food source position of the sub-population is updated; the food source position represents the optimal solution; updating the food source position of the MEC device population according to a comparison result of the food source position between the sub-population and the MEC device population; According to the fitness value of each individual in the MEC device population after the food source location is updated, the search operation of the MEC device is iterated to output the optimal combined unloading strategy.

[0007] According to a resource scheduling method provided by the present invention, updating the positions of the leader and the followers in the sub-population according to the adaptability function includes: Calculating a first fitness value of each sub-individual in the sub-population according to the adaptive function; Sorting the first fitness values ​​in descending order to obtain a sorted sequence; At least one of the sub-individuals ranked first in the sorting sequence is used as a leader of the sub-population, and the remaining sub-individuals in the sorting sequence are used as followers of the sub-population.

[0008] According to a resource scheduling method provided by the present invention, updating the food source location of the MEC device population includes updating the location of the leader and the follower in the MEC device population; The updating of the food source position of the MEC device population according to the comparison result of the food source position between the sub-population and the MEC device population includes: Determine a comparison result of the position of the leader and a comparison result of the position of the follower between the sub-population and the MEC device population; the comparison result of the food source position includes the comparison result of the position of the leader and the comparison result of the position of the follower; If the comparison result of the position of the leader is that the position of the leader of the sub-population is better than the position of the leader of the MEC device population, then the position of the leader of the MEC device population is updated to the position of the leader of the sub-population; If the comparison result of the positions of the followers is that the positions of the followers of the sub-population are better than the positions of the followers of the MEC device population, the positions of the followers of the MEC device population are updated to the positions of the followers of the sub-population.

[0009] According to a resource scheduling method provided by the present invention, the energy consumption weight coefficient and the time consumption weight coefficient of the MEC device are determined based on the terminal attribute document of the MEC device; the terminal attribute document is established based on the following method: Acquire basic device information of the MEC device; the basic device information includes at least one of energy consumption sensitive information, basic power, computing capability, storage capacity and delay sensitive information; The terminal attribute document is established according to the basic device information.

[0010] The present invention also provides a resource scheduling device, comprising the following modules: A request response module, configured to respond to a task offloading request sent by a terminal device and publish the task to be offloaded carried in the task offloading request to a neighboring MEC device; A receiving module, used for receiving the evaluation result of the task to be offloaded sent by the neighboring MEC device; the evaluation result is used to indicate whether the neighboring MEC device has the ability to process the task to be offloaded; The scheduling module is used to determine at least one target MEC device according to the energy consumption information and latency information of the MEC device when it is determined based on the evaluation result that the task to be offloaded needs to be migrated, and schedule the task to be offloaded to the target MEC device.

[0011] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any of the resource scheduling methods described above when executing the computer program.

[0012] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the resource scheduling method described in any one of the above is implemented.

[0013] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the resource scheduling methods described above.

[0014] The resource scheduling method, apparatus, device, storage medium and computer program product provided by the present invention respond to the task offloading request sent by the terminal device, publish the task to be offloaded carried by the task offloading request to the adjacent MEC device; receive the evaluation result of the task to be offloaded sent by the adjacent MEC device; the evaluation result is used to characterize whether the adjacent MEC device has the ability to process the task to be offloaded; when it is determined based on the evaluation result that the task to be offloaded needs to be migrated, at least one target MEC device is determined according to the energy consumption information and delay information of the MEC device, and the task to be offloaded is scheduled to the target MEC device. The tasks in the present invention do not need to be offloaded from the cloud center node to the edge device, but the local edge device accepts the task request, distributes the computing strategy and autonomously completes the resource allocation to achieve decentralized resource scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 It is a flow chart of the resource scheduling method provided by the present invention.

[0017] Figure 2 It is a flow chart of determining the energy consumption weight coefficient and time consumption weight coefficient of the MEC device provided by the present invention.

[0018] Figure 3 It is a schematic diagram of the framework of the decentralized MEC system provided by the present invention.

[0019] Figure 4 It is a flowchart diagram of the task offloading process provided by the present invention.

[0020] Figure 5 It is a flowchart of the resource allocation process provided by the present invention.

[0021] Figure 6 It is a structural diagram of the resource scheduling device provided by the present invention.

[0022] Figure 7 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] Combine the following Figure 1-Figure 7 The resource scheduling method, apparatus, device, storage medium and computer program product of the present invention are described.

[0025] Figure 1 It is a flow chart of the resource scheduling method provided by the present invention, such as Figure 1 As shown, the method includes the following: Step 101, in response to a task offloading request sent by a terminal device, publishing the task to be offloaded carried in the task offloading request to a neighboring MEC device; It can be understood that task offloading refers to the process of transferring computing tasks from smart terminal devices or mobile devices (such as smartphones, sensors, etc.) to mobile edge computing (MEC) devices for processing. The purpose of task offloading is to improve processing efficiency, reduce the computing burden and energy consumption of terminal devices, and reduce task processing latency.

[0026] In the embodiment of the present invention, the MEC device deployed at the edge of the network can be understood as an MEC node, which is responsible for processing data, executing tasks, and providing low-latency computing resources. Each MEC node is pre-installed with a micro-evaluation system that can quickly evaluate key parameters such as energy consumption, latency, and communication costs of tasks. At the same time, the evaluation system can optimize task allocation based on the specific requirements of the task (such as computational complexity and data volume) and the current network and device status (such as transmission distance and terminal device power). This evaluation capability enables MEC nodes to make more accurate decisions, improve resource utilization, and reduce overall energy consumption.

[0027] MEC nodes are highly autonomous and can handle task allocation and resource management independently without relying on a central control system. At the same time, MEC nodes can be freely connected to form a dynamic network structure. This free connection capability enables the MEC cluster to adjust the network topology as needed to adapt to different loads and task requirements. In this structure, there is no fixed central node or single point of control. Each node can become a central node at a specific stage to coordinate task processing and resource allocation, achieve decentralization, and thus improve the flexibility and robustness of the system.

[0028] In a distributed computing environment, after a terminal device generates a task offloading request, it sends the task offloading request to the nearest MEC device, where the task offloading request includes the task to be offloaded and its related information, such as task type, data volume, processing requirements, etc. The MEC device that receives the task offloading request will first evaluate the task requirements, including computing power, data volume, time limit, etc., and at the same time, check whether its own resources (such as computing power, memory, network bandwidth) are sufficient to handle the task. If the MEC device finds that its own resources are insufficient, or believes that the task processing may be completed more efficiently on other devices, it will publish the task to be offloaded to the neighboring MEC device; or directly forward the task offloading request to the neighboring MEC device, and the forwarded request will carry all the information of the original task, including a detailed description of the task, data, processing requirements, etc. Among them, the neighboring MEC device refers to a device that is physically close, which can reduce data transmission delays.

[0029] Step 102: Receive the evaluation result of the task to be offloaded sent by the neighboring MEC device.

[0030] After receiving the task to be offloaded or the task offloading request, each neighboring MEC device evaluates the task to be offloaded, including parameters such as computing load, energy consumption, and delay. Then, the neighboring MEC devices share the evaluation results and task parameters with each other. The evaluation results are used to characterize whether the neighboring MEC devices have the ability to process the task to be offloaded, including evaluating its computing power, resource availability, energy consumption, delay requirements and other factors; task parameters include task load, task type, processing time, energy consumption requirements, data transmission volume, delay requirements, priority and current status and other parameters.

[0031] The MEC device receives the evaluation results and task parameters shared by each neighboring MEC device.

[0032] Step 103: When it is determined based on the evaluation result that the task to be offloaded needs to be migrated, at least one target MEC device is determined according to the energy consumption information and latency information of the MEC device, and the task to be offloaded is scheduled to the target MEC device.

[0033] It can be understood that task migration refers to the process of moving computing tasks from one MEC device to another MEC device, with the aim of optimizing resource utilization, improving system performance, reducing latency and energy consumption, etc.

[0034] After receiving the evaluation results sent by each neighboring MEC device, the MEC device evaluates the task migration strategy based on the evaluation results to determine whether the task to be offloaded needs to be migrated. For example, if the resources of the current node (such as CPU, memory, storage) are close to saturation and the neighboring MEC device has the ability to process the task to be offloaded, task migration can be performed to reduce the load; or, in the case of uneven load on each node, task migration is required to achieve load balancing and improve the overall performance of the system.

[0035] After determining that the task to be offloaded requires task migration, at least one target MEC device is determined based on the energy consumption information and latency information of the MEC device, and the task to be offloaded is scheduled to the target MEC device. The energy consumption information includes the energy consumption weight coefficient and the energy consumed by the MEC device to process the task to be offloaded; the latency information includes the time consumption weight coefficient and the total time required for the MEC device to process the task to be offloaded.

[0036] For example, a distributed resource allocation model is established to perform the SSA algorithm (Salp Swarm Algorithm) on idle MEC devices (i.e., MEC devices in an idle state) to find the optimal strategy, share individual optimal values ​​with each other, and output the optimal combined unloading strategy.

[0037] The resource scheduling method provided by the embodiment of the present invention responds to the task offloading request sent by the terminal device, publishes the task to be offloaded carried by the task offloading request to the neighboring MEC device; receives the evaluation result of the task to be offloaded sent by the neighboring MEC device; the evaluation result is used to characterize whether the neighboring MEC device has the ability to process the task to be offloaded; when it is determined based on the evaluation result that the task to be offloaded needs to be migrated, at least one target MEC device is determined according to the energy consumption information and delay information of the MEC device, and the task to be offloaded is scheduled to the target MEC device. The tasks in the present invention do not need to be offloaded from the cloud center node to the edge device, but the local edge device accepts the task request, distributes the computing strategy and autonomously completes the resource allocation to realize decentralized resource scheduling.

[0038] Based on the above embodiment, at least one target MEC device is determined according to the energy consumption information and latency information of the MEC device, including: Step 1030, determining an adaptability function according to the energy consumption weight coefficient, the energy consumed by the MEC device to process the task to be offloaded, the time consumption weight coefficient, and the total time consumed by the MEC device to process the task to be offloaded; the adaptability function satisfies the balance between minimizing energy consumption and satisfying the QoS satisfaction; Step 1031, based on the adaptability function, searching for MEC devices to determine the optimal combined offloading strategy; Step 1032: Determine at least one target MEC device based on the optimal combined offloading strategy.

[0039] A distributed resource allocation model is established to determine the adaptive function that satisfies the balance between minimizing energy consumption and satisfying the quality of service (QoS): ; in, Indicates the energy consumed by the MEC device to process the task to be offloaded, that is, the MEC node Processing tasks Energy consumed; represents the energy consumption weight coefficient; Indicates the total time required for the MEC device to process the unloaded tasks, that is, the MEC node Processing tasks Total time spent Represents the time consumption weight coefficient.

[0040] After determining the adaptability function, the SSA algorithm is used to find the optimal value for all MEC devices, and the individual optimal values ​​are shared with each other to output the optimal combined unloading strategy.

[0041] For example, the task set is: ; Assume that each MEC device Processing tasks The energy consumed is And each MEC device Processing tasks The total time taken is , For MEC devices The energy consumed to perform all tasks, For MEC devices The power consumption, is the size of the task set, the task allocation matrix The size is , the task allocation matrix Each element in is represented by , To perform tasks Power required, is the total latency of MEC node executing application A, For MEC devices and MEC devices The transmission distance between them. Under the condition of weighing the delay and computing resource constraints, the optimization goal can be described as: (1) Total energy consumption: ; (2) Total delay: ; The steps of using the SSA algorithm to find the optimal combined unloading strategy include: (1) Encode the MEC devices in the idle state and mark them as population particles for initialization to obtain the MEC device population. For example, first assign a unique code to each idle MEC device to represent the state or characteristics of the device; then, initialize these codes as particles in the SSA algorithm to form the initial population of the algorithm. Each particle represents a possible solution or strategy. Through these particles, the algorithm can explore the solution space of the optimization problem.

[0042] (2) Initialize a subpopulation for a single MEC device in an idle state to obtain a subpopulation. For example, select a MEC device to initialize a subpopulation, set the algorithm parameters, initialize the subpopulation size to n, the number of MEC devices to m, and the squat tunicate population composition with a dimension of D, recorded as ,in Refers to the third particle of the MEC device with code 2.

[0043] (3) Update the positions of the leader and followers in the subpopulation according to the adaptive function. Specifically, calculate the first fitness value of each sub-individual in the subpopulation according to the adaptive function; sort the first fitness values ​​in descending order to obtain a sorted sequence; take at least one sub-individual ranked first in the sorted sequence as the leader of the subpopulation, and take the remaining sub-individuals in the sorted sequence as followers of the subpopulation.

[0044] For example, initialize each Leader's position. Calculate the fitness values ​​of N salps (i.e., sub-individuals or MEC sub-individuals), that is, calculate the value of the adaptive function; then, sort all i×j salps in descending order according to the fitness value (i.e., sort from large to small), and update the position of the salp with the best fitness in the first place as the latest food position of the node group. Further, perform position update, and sort the remaining N-1 salps in the group in descending order according to the fitness value of the salp group, and regard the first N / 2 salps as leaders, and the remaining N / 2 salps as followers.

[0045] It can be understood that the leader is the current optimal particle, representing the best solution or strategy. In each sub-population, the leader is the particle with the highest fitness, which guides the movement direction of other particles and helps optimize the search process. Followers are particles that follow the leader and are suboptimal particles with better fitness. Followers adjust their positions according to the position of the leader so that they can better follow the leader and explore the solution space during the optimization process. The goal of the followers is to gradually approach the leader, thereby improving the optimization effect of the entire population.

[0046] (4) Update the location of the food source of the sub-population according to the fitness value of each sub-individual in the sub-population after the location update. The location of the food source represents the optimal solution and represents the target location for finding the best solution. For example, each independent MEC device Prioritize updating the latest leader of your own subpopulation and followers .

[0047] It can be understood that updating the position of the leader means finding the current optimal location of the food source, that is, optimizing the current solution. Updating the position of the follower means following the leader's adjustments to help other individuals get closer to this optimized position, thereby spreading high-quality solutions in the population. Among them, changes in the leader's position will directly affect the update of the food source position, because the leader's position represents the optimal solution or a position close to the optimal solution. Therefore, updating the positions of leaders and followers can be regarded as a way to update the location of the food source, because this helps to optimize the entire search space and find a better solution.

[0048] (5) Based on the comparison result of the food source location between the sub-population and the MEC device population, the food source location of the MEC device population is updated. The comparison result of the food source location includes the comparison result of the leader's location and the comparison result of the follower's location; updating the food source location of the MEC device population includes updating the locations of the leader and the follower in the MEC device population. Specifically, the comparison result of the leader's location and the comparison result of the follower's location between the sub-population and the MEC device population are determined; if the comparison result of the leader's location is that the location of the leader of the sub-population is better than the location of the leader of the MEC device population, the location of the leader of the MEC device population is updated to the location of the leader of the sub-population; if the comparison result of the follower's location is that the location of the follower of the sub-population is better than the location of the follower of the MEC device population, the location of the follower of the MEC device population is updated to the location of the follower of the sub-population.

[0049] For example, the leader of a single MEC device subpopulation and followers , respectively, with the leaders of the MEC device population and followers Compare, if Better, then update for ;if Better, then update for ,vice versa.

[0050] (6) According to the fitness value of each individual in the MEC device population after the food source location is updated, the search operation of the MEC device is iterated to output the optimal combined unloading strategy.

[0051] For example, the fitness value of the updated MEC device population is calculated, and the fitness value of each salp individual after the update is compared with the fitness value of the current food. If the fitness value of the updated salp is better than that of the food, the position of the salp with a better fitness value is used as the new food position. Repeat the position update operation of the leader and follower of the population until the set number of iterations is reached or the fitness value meets certain requirements. After the termination condition is met, the current food position is output as the estimated position of the target.

[0052] The embodiment of the present invention uses idle MEC devices as population particles of the SSA algorithm to obtain a population composed of MEC devices. The sub-population can be reinitialized inside a single MEC device to form a two-layer heuristic computing mode. By utilizing the distributed and efficient computing characteristics of MEC, the improved SSA algorithm is improved from one-dimensional computing to two-dimensional computing, with exponential growth in capabilities, saving the delay of multi-objective optimization and improving MEC resource utilization.

[0053] Based on the above embodiment, the energy consumption weight coefficient and the time consumption weight coefficient of the MEC device are determined based on the terminal attribute document of the MEC device; the terminal attribute document is established in the following manner: Obtain basic device information of the MEC device; the basic device information includes at least one of energy consumption sensitive information, basic power, computing capability, storage capacity and delay sensitive information; establish a terminal attribute document based on the basic device information.

[0054] For example, establishing a device cloud profile based on device attributes is conducive to realizing different task offloading and resource scheduling for specific devices. Figure 2 , collect basic device information of MEC devices, including energy consumption sensitive information, basic power, computing power, storage capacity and delay sensitive information, etc., to provide a basis for MEC device decision-making. For the marked terminal attributes, establish a terminal attribute document; then, based on the device's own performance and power supply mode, computing, communication and other capabilities, mark the energy consumption weight coefficient and delay weight coefficient of the MEC device without affecting QoS. For example, the monitoring device and the alarm device are two terminal devices. The monitoring device itself requires a large bandwidth and has a heavy computing task that tends to be energy-intensive. The initial energy consumption With delay The ratio is 8:2, the alarm device is more inclined to respond promptly, so the energy consumption and delay The ratio is 2:8.

[0055] The embodiment of the present invention takes into account the different requirements of each terminal device, constructs a cloud archive of the terminal device, sets an energy consumption weight coefficient and a delay weight coefficient, and effectively improves resource allocation efficiency.

[0056] In order to further analyze and illustrate the resource scheduling method provided by the present invention, refer to the following embodiments.

[0057] The embodiment of the present invention proposes a distributed computing resource scheduling method based on edge computing, specifically a method for task unloading and reasonable resource scheduling in a decentralized regional neighbor MEC system, which comprehensively considers the transmission time, energy consumption, task dependency and communication cost and computing cost generated by the distribution distance between MEC devices. The method mainly includes: establishing a cloud archive of intelligent terminal devices, including the communication volume, location, energy consumption requirements and time consumption parameters of terminal devices, etc. Because different types of devices have different emphasis on energy consumption and delay requirements, the original path is to upload terminal tasks to the cloud center, and the cloud center calculates the unloading method and resource allocation. The new method proposed in the embodiment of the present invention distributes the original cloud center calculation to the second-layer MEC devices to reduce the excessive load of the cloud center caused by task concentration. Each MEC device is pre-set with a micro system, and the energy consumption, delay and communication cost are quickly obtained according to the task transmission distance, task set size and data volume. The dynamic task strategy management is for the MEC node to give priority to tasks suitable for itself according to its own attributes, and the multi-node information is asynchronously communicated and combined to output the optimal task combination processing strategy for the front and back dependency tasks. For tasks that require combined processing, the model calculation amount increases exponentially as the number of tasks and edge devices increases. The heuristic algorithm SSA is proposed to save processing delay. In order to improve the convergence speed, each MEC device uses the random walk method to update the location of the food source, which is equivalent to each MEC device becoming a particle, which speeds up the convergence speed of the mainstream SSA and improves the decision-making model speed. The actual problem is converted into a minimum optimization problem of the adaptability function, and the adaptive global optimal value is quickly found to output task offloading, communication resource allocation and computing resource allocation decisions.

[0058] refer to Figure 3 ,The distributed computing resource scheduling method based on edge computing mainly includes the following contents: S1 builds weighted coefficients: Establishes a MEC device cloud file for smart devices, including the device communication network status, location, energy consumption, and time parameters. Because different device types have different emphases on energy consumption and latency requirements, balance the energy consumption weight coefficient and latency weight coefficient of the device.

[0059] S2 task offloading process: A decentralized dynamic node task management strategy is proposed to reduce task spillover and reduce information communication costs. The terminal device autonomously selects to establish a regional neighbor topology relationship with a nearby MEC device based on the distance from the MEC device, and the terminal device completes hand-in-hand (information authentication) with nearby neighbors. If the MEC device cannot undertake the current task or there is a task waiting, and task migration is required, the nearby neighbors will take the initiative to carry part of the task and take over the task in turn. Multiple MEC devices form a directed acyclic graph to reduce the latency of uploading tasks to the center and the latency of unloading tasks in the cloud center.

[0060] S3 resource allocation process: The main focus of the resource allocation problem is the rationality of resource allocation. According to the characteristics of the heuristic algorithm and the characteristics of the particle group dispersedly searching for values ​​and then updating the optimal value, all idle MECs are randomly initialized as particles to find the optimal value, and the group particle position and the particle position are updated in real time. The convergence speed of the mainstream SSA is accelerated. Compared with other optimal algorithms, the SSA algorithm has high exploratory and avoids local optimality.

[0061] refer to Figure 4 ,The task offloading process mainly includes the following contents: S2-1: The terminal device sends an unloading task request; S2-2: The nearest regional MEC node receives the offloading task request; S2-3: The task is horizontally published to the regional neighbor MEC nodes. After distributed task evaluation, the neighbor MEC nodes share task parameters with each other. S2-4: Evaluate whether the current task set requires task migration, e.g. The terminal device uploads the data to the nearest MEC node. The MEC node calculates the parameters and determines whether it can be unloaded. The redundant tasks are unloaded to the neighboring MEC nodes in the area. The two nodes complement each other and cycle S2-3 according to the task volume until there is no migration task. S2-5: Establish a distributed resource allocation model to meet the balance between energy consumption minimization and QoS satisfaction (energy consumption weight coefficient and time weight coefficient ): ; The SSA algorithm is used to find the optimal value for all MEC devices, and the individual optimal values ​​are shared with each other to output the optimal combined unloading strategy; S2-6: Dynamic policy management, after the task of each MEC node is completed, it actively requests tasks when the communication cost threshold and latency requirements are met; this policy is deployed on the MEC node, and mainly performs status scanning, migration evaluation and migration decision-making on the tasks in the current queue in real time to ensure that the execution of tasks of different priorities achieves a trade-off between energy consumption and latency.

[0062] S2-7: According to the offloading strategy, the MEC devices that meet the requirements automatically receive the task and execute it; S2-8: Upload the results.

[0063] refer to Figure 5 ,The resource allocation process mainly includes the following contents: S3-1: Encode the MEC devices in the idle state and mark them as population particles for initialization; S3-2: A single MEC node initializes the subpopulation and sets the algorithm parameters. The subpopulation initialization size is n, the number of MEC devices is m, and the squat tunicate population of dimension D is composed of: ,in Refers to the third particle of the MEC device with code 2; S3-3: Initialize each The position of the leader. Calculate the fitness values ​​of N salps (i.e., sub-individuals or MEC sub-individuals), that is, calculate the value of the adaptive function; then, sort all i×j salps in descending order according to the fitness value (i.e., sort from large to small), and update the position of the salp with the best fitness at the top as the latest food position of the node group; Position update: the remaining N-1 salps in the group are sorted in descending order according to the fitness value of the salp group. The first N / 2 salps are regarded as leaders, and the remaining N / 2 salps are regarded as followers. S3-4: Each independent MEC device Prioritize updating the latest leader of your own subpopulation and followers ; S3-5: Leader of a single MEC device subpopulation and followers , respectively, with the leaders of the MEC device population and followers Compare, if Better, then update for ;if Better, then update for ,vice versa

[0064] S3-6: Calculate fitness. Calculate the fitness value of the updated MEC device population, compare the updated fitness value of each salp individual with the fitness value of the current food, and if the fitness value of the updated salp is better than that of the food, the position of the salp with the better fitness value is used as the new food position.

[0065] S3-7: Repeat S3-3 to S3-6 until the set number of iterations is reached or the fitness value meets certain requirements. After the termination condition is met, the current food position is output as the estimated position of the target.

[0066] S3-8: End.

[0067] The decentralized service request distribution method based on edge computing proposed in the embodiment of the present invention does not need to offload tasks from the cloud center node to the edge device, but the local edge device accepts the task request, and the distributed computing strategy and autonomously completes the resource allocation; at the same time, the MEC node adjacent to the stage center becomes the stage center, reducing the communication node, saving the latency of uploading the terminal task to the cloud center and unloading the cloud center task, and the horizontal distributed unloading task also improves the efficiency of task unloading. By making decisions based on different device attributes and task priority differences, the energy consumption and latency weight coefficients in the scheduling target resource allocation are generated to improve the service quality Qos and resource allocation efficiency. In addition, an optimized resource allocation algorithm SSA is proposed, combining the low latency and high bandwidth characteristics of MEC devices, and the fast information exchange between devices. All idle MEC devices are regarded as the population of the SSA algorithm, and each particle of the population can represent the population. The sub-population can be reinitialized inside a single MEC device to form a two-layer heuristic computing mode, which not only improves the convergence and speed of the SSA algorithm but also makes rational use of MEC idle resources.

[0068] The resource scheduling device provided by the present invention is described below. The resource scheduling device described below and the resource scheduling method described above can be referenced to each other.

[0069] refer to Figure 6 The resource scheduling device provided by the present invention includes a request response module 601, a receiving module 602 and a scheduling module 603.

[0070] The request response module 601 is used to respond to the task offloading request sent by the terminal device and publish the task to be offloaded carried in the task offloading request to the adjacent MEC device; The receiving module 602 is used to receive the evaluation result of the task to be offloaded sent by the neighboring MEC device; the evaluation result is used to indicate whether the neighboring MEC device has the ability to process the task to be offloaded; The scheduling module 603 is used to determine at least one target MEC device according to the energy consumption information and latency information of the MEC device when it is determined based on the evaluation result that the task to be offloaded needs to be migrated, and schedule the task to be offloaded to the target MEC device.

[0071] The resource scheduling device provided in the embodiment of the present invention responds to the task offloading request sent by the terminal device, publishes the task to be offloaded carried by the task offloading request to the neighboring MEC device; receives the evaluation result of the task to be offloaded sent by the neighboring MEC device; the evaluation result is used to characterize whether the neighboring MEC device has the ability to process the task to be offloaded; when it is determined based on the evaluation result that the task to be offloaded needs to be migrated, at least one target MEC device is determined according to the energy consumption information and delay information of the MEC device, and the task to be offloaded is scheduled to the target MEC device. The tasks in the present invention do not need to be offloaded from the cloud center node to the edge device, but the local edge device accepts the task request, distributes the computing strategy and autonomously completes the resource allocation to realize decentralized resource scheduling.

[0072] In one embodiment, the energy consumption information includes an energy consumption weight coefficient and the energy consumed by the MEC device to process the task to be offloaded; the delay information includes a time consumption weight coefficient and the total time consumed by the MEC device to process the task to be offloaded; the scheduling module 603 is specifically used to: Determine an adaptability function according to the energy consumption weight coefficient, the energy consumed by the MEC device to process the task to be offloaded, the time consumption weight coefficient and the total time consumption required by the MEC device to process the task to be offloaded; the adaptability function satisfies the balance between minimizing energy consumption and satisfying the quality of service (QoS); based on the adaptability function, search for the MEC device to determine the optimal combined offloading strategy; based on the optimal combined offloading strategy, determine at least one of the target MEC devices.

[0073] In one embodiment, the scheduling module 603 is specifically configured to: The MEC devices in the idle state are encoded and marked as population particles for initialization to obtain a MEC device population; a sub-population is initialized for a single MEC device in the idle state to obtain a sub-population; the positions of the leader and the follower in the sub-population are updated according to the adaptability function; the food source position of the sub-population is updated according to the fitness value of each sub-individual in the sub-population after the position is updated; the food source position represents the optimal solution; the food source position of the MEC device population is updated according to the comparison result of the food source position between the sub-population and the MEC device population; according to the fitness value of each individual in the MEC device population after the food source position is updated, the search operation of the MEC device is iterated to output the optimal combined unloading strategy.

[0074] In one embodiment, the scheduling module 603 is specifically configured to: According to the adaptability function, the first fitness value of each sub-individual in the sub-population is calculated; the first fitness values ​​are sorted in descending order to obtain a sorted sequence; at least one of the sub-individuals ranked first in the sorted sequence is used as the leader of the sub-population, and the remaining sub-individuals in the sorted sequence are used as followers of the sub-population.

[0075] In one embodiment, updating the food source location of the MEC device population includes updating the location of the leader and the follower in the MEC device population; the scheduling module 603 is specifically used to: Determine the comparison result of the leader's position and the comparison result of the followers' positions between the sub-population and the MEC device population; the comparison result of the food source position includes the comparison result of the leader's position and the comparison result of the followers' positions; if the comparison result of the leader's position is that the position of the leader of the sub-population is better than the position of the leader of the MEC device population, then update the position of the leader of the MEC device population to the position of the leader of the sub-population; if the comparison result of the followers' positions is that the position of the followers of the sub-population is better than the position of the followers of the MEC device population, then update the position of the followers of the MEC device population to the position of the followers of the sub-population.

[0076] In one embodiment, the energy consumption weight coefficient and the time consumption weight coefficient of the MEC device are determined based on the terminal attribute document of the MEC device; the terminal attribute document is established based on the following method: Obtain basic device information of the MEC device; the basic device information includes at least one of energy consumption sensitive information, basic power, computing capability, storage capacity and delay sensitive information; and establish the terminal attribute document based on the basic device information.

[0077] Figure 7 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 7As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communication interface 720 and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute the resource scheduling method, which includes: in response to a task offloading request sent by a terminal device, publishing the task to be offloaded carried by the task offloading request to a neighboring MEC device; receiving an evaluation result of the task to be offloaded sent by the neighboring MEC device; the evaluation result is used to characterize whether the neighboring MEC device has the ability to process the task to be offloaded; in the case where it is determined based on the evaluation result that the task to be offloaded needs to be migrated, at least one target MEC device is determined according to the energy consumption information and delay information of the MEC device, and the task to be offloaded is scheduled to the target MEC device.

[0078] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0079] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the resource scheduling method provided by the above methods, the method including: in response to a task offloading request sent by a terminal device, publishing the task to be offloaded carried by the task offloading request to a neighboring MEC device; receiving an evaluation result of the task to be offloaded sent by the neighboring MEC device; the evaluation result is used to characterize whether the neighboring MEC device has the ability to process the task to be offloaded; when it is determined based on the evaluation result that the task to be offloaded needs to be migrated, determining at least one target MEC device based on the energy consumption information and delay information of the MEC device, and scheduling the task to be offloaded to the target MEC device.

[0080] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the resource scheduling method provided by the above-mentioned methods, the method comprising: in response to a task offloading request sent by a terminal device, publishing the task to be offloaded carried by the task offloading request to a neighboring MEC device; receiving an evaluation result of the task to be offloaded sent by the neighboring MEC device; the evaluation result is used to characterize whether the neighboring MEC device has the ability to process the task to be offloaded; in a case where it is determined based on the evaluation result that the task to be offloaded needs to be migrated, determining at least one target MEC device based on the energy consumption information and delay information of the MEC device, and scheduling the task to be offloaded to the target MEC device.

[0081] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0082] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A resource scheduling method, characterized in that: MEC devices used in mobile edge computing include: In response to a task offloading request sent by a terminal device, publishing the task to be offloaded carried in the task offloading request to a neighboring MEC device; Receiving an evaluation result of the task to be offloaded sent by the neighboring MEC device; the evaluation result is used to indicate whether the neighboring MEC device has the ability to process the task to be offloaded; When it is determined based on the evaluation result that the task to be offloaded needs to be migrated, at least one target MEC device is determined according to the energy consumption information and latency information of the MEC device, and the task to be offloaded is scheduled to the target MEC device.

2. The resource scheduling method according to claim 1, characterized in that: The energy consumption information includes an energy consumption weight coefficient and the energy consumed by the MEC device to process the task to be offloaded; the delay information includes a time consumption weight coefficient and the total time consumed by the MEC device to process the task to be offloaded; Determine at least one target MEC device according to the energy consumption information and latency information of the MEC device, including: Determine an adaptability function according to the energy consumption weight coefficient, the energy consumed by the MEC device to process the task to be offloaded, the time consumption weight coefficient and the total time consumed by the MEC device to process the task to be offloaded; the adaptability function satisfies the balance between minimizing energy consumption and satisfying the quality of service QoS; Based on the adaptability function, searching the MEC device to determine an optimal combined offloading strategy; Based on the optimal combined offloading strategy, at least one of the target MEC devices is determined.

3. The resource scheduling method according to claim 2, characterized in that: The searching the MEC device based on the adaptability function to determine the optimal combined unloading strategy includes: Encode the MEC devices in the idle state and mark them as population particles for initialization to obtain the MEC device population; Initialize a sub-population for a single MEC device in an idle state to obtain a sub-population; According to the adaptive function, updating the positions of the leader and the followers in the sub-population; According to the fitness value of each sub-individual in the sub-population after the position is updated, the food source position of the sub-population is updated; the food source position represents the optimal solution; updating the food source position of the MEC device population according to a comparison result of the food source position between the sub-population and the MEC device population; According to the fitness value of each individual in the MEC device population after the food source location is updated, the search operation of the MEC device is iterated to output the optimal combined unloading strategy.

4. The resource scheduling method according to claim 3, characterized in that: The updating of the positions of the leader and the followers in the sub-population according to the adaptive function comprises: Calculating a first fitness value of each sub-individual in the sub-population according to the adaptive function; Sorting the first fitness values ​​in descending order to obtain a sorted sequence; At least one of the sub-individuals ranked first in the sorting sequence is used as a leader of the sub-population, and the remaining sub-individuals in the sorting sequence are used as followers of the sub-population.

5. The resource scheduling method according to claim 3, characterized in that: Updating the food source location of the MEC device population includes updating the location of the leader and followers in the MEC device population; The updating of the food source position of the MEC device population according to the comparison result of the food source position between the sub-population and the MEC device population includes: Determine a comparison result of the position of the leader and a comparison result of the position of the follower between the sub-population and the MEC device population; the comparison result of the food source position includes the comparison result of the position of the leader and the comparison result of the position of the follower; If the comparison result of the position of the leader is that the position of the leader of the sub-population is better than the position of the leader of the MEC device population, then the position of the leader of the MEC device population is updated to the position of the leader of the sub-population; If the comparison result of the positions of the followers is that the positions of the followers of the sub-population are better than the positions of the followers of the MEC device population, the positions of the followers of the MEC device population are updated to the positions of the followers of the sub-population.

6. The resource scheduling method according to claim 2, characterized in that: The energy consumption weight coefficient and the time consumption weight coefficient of the MEC device are determined based on the terminal attribute document of the MEC device; the terminal attribute document is established based on the following method: Acquire basic device information of the MEC device; the basic device information includes at least one of energy consumption sensitive information, basic power, computing capability, storage capacity and delay sensitive information; The terminal attribute document is established according to the basic device information.

7. A resource scheduling device, characterized in that: include: A request response module, configured to respond to a task offloading request sent by a terminal device and publish the task to be offloaded carried in the task offloading request to a neighboring MEC device; A receiving module, used for receiving the evaluation result of the task to be offloaded sent by the neighboring MEC device; the evaluation result is used to indicate whether the neighboring MEC device has the ability to process the task to be offloaded; The scheduling module is used to determine at least one target MEC device according to the energy consumption information and latency information of the MEC device when it is determined based on the evaluation result that the task to be offloaded needs to be migrated, and schedule the task to be offloaded to the target MEC device.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the resource scheduling method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the resource scheduling method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the resource scheduling method according to any one of claims 1 to 6 is implemented.

Images