An edge computing fine-grained scheduling method based on an IBBO algorithm

Abstract

The application relates to an edge computing fine-grained scheduling method based on an IBBO algorithm, wherein by acquiring tasks to be processed of each terminal device in real time, tasks to be processed greater than a preset threshold are decomposed according to a specified mode to obtain subtasks to be processed; the subtasks to be processed are assigned corresponding priorities according to the importance and urgency of the tasks; a multi-objective optimization model of task scheduling is constructed based on the subtasks to be processed and the corresponding priorities; a fitness function is set for the multi-objective optimization model, and the optimal solution of the fitness function is a comprehensive optimal solution based on optimal time delay and energy consumption of task scheduling; the comprehensive optimal solution of the fitness function of the multi-objective optimization model is solved based on the IBBO algorithm; and the subtasks to be processed are scheduled based on the obtained comprehensive optimal solution. Through the adaptive migration mechanism and the improved migration strategy of the IBBO algorithm, the migration probability between the entities is dynamically adjusted, so that the optimization process is accelerated, the search efficiency is improved, and falling into a local optimal solution is avoided.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of edge computing and task scheduling, and particularly relates to an edge computing fine-grained scheduling method based on an IBBO algorithm, namely an improved biogeography-based optimization algorithm. BACKGROUND

[0002] As a distributed computing mode, edge computing pushes data processing capability to the network edge to reduce latency and bandwidth utilization and can support fast response to real-time data. In today's information society, edge computing plays an increasingly important role. However, task scheduling in the edge computing environment faces many challenges, of which the most critical is fine-grained task scheduling. This task scheduling must meet the performance requirements of application programs while maximizing resource utilization, so it is of great importance.

[0003] The original intention of edge computing is to share the task processing pressure of the cloud computing center to improve service quality and response speed. However, the resources and computing capabilities possessed by edge computing nodes are limited, which makes them only capable of processing a small part of tasks. Therefore, how to reasonably allocate task scheduling to corresponding edge nodes has become an urgent problem to be solved. This challenge mainly comes from two aspects: one is the different distances of edge nodes from the cloud computing center, which leads to the differences in time and energy consumption required for task scheduling; the other is that different tasks require different computing resources. If small tasks are scheduled to edge nodes with strong computing capabilities, it will lead to resource waste. Therefore, how to maximize the use of edge node resources while ensuring the efficiency of task scheduling has become a problem to be solved.

[0004] There are several task scheduling strategies in the prior art:

[0005] 1. Multi-user fine-grained task offloading scheduling for mobile edge computing: This scheme abstracts the computing task as a directed acyclic graph and takes system delay as the optimization target. Based on considering constraints such as deadline, priority, node completion deadline, etc., an improved NSGA-Ⅱ algorithm is used to solve the task offloading scheduling scheme.

[0006] 2. Adaptive genetic algorithm for MEC task offloading and resource allocation: This scheme takes the system total overhead as the optimization target and decomposes it into three sub-problems, including task scheduling, resource allocation, etc. These sub-problems are solved one by one by the adaptive genetic algorithm, and a comprehensive task scheduling and resource offloading scheme is obtained.

[0007] 3. Edge computing scheduling based on deep reinforcement learning: This scheme aims to maximize task satisfaction by scheduling multiple tasks into virtual machines on edge servers and using deep reinforcement learning algorithms to solve time scheduling and resource allocation problems.

[0008] The above-mentioned task scheduling strategies, although to some extent, solve some technical problems, but still have some limitations, including a huge search space: traditional optimization algorithms often need to face a huge combination of tasks and resources in the edge computing environment. This leads to the need for a large amount of computing resources and time to explore the search space, limiting its feasibility and practicality in practice. Slow convergence speed: Many existing algorithms in dealing with task scheduling problems in edge computing environments have slow convergence speed due to the complexity of the search space. Long computing process not only increases the delay of task scheduling, but also may cause the inability to meet real-time performance requirements. Prone to local optimal solution: Some algorithms are prone to local optimal solution in the search process, which cannot globally optimize the task scheduling scheme. This may be due to the limitations of algorithm design or the lack of search strategy, affecting the quality and performance of task scheduling.

[0009] Therefore, how to improve the task scheduling optimization algorithm in the prior art to solve the technical limitations of the above-mentioned task scheduling strategies is a technical problem that needs to be solved at present. SUMMARY

[0010] The purpose of the present application is to provide an edge computing fine-grained scheduling method based on IBBO algorithm, which improves the classical biogeography-based optimization algorithm by adding a decision mechanism and a dynamic migration mechanism in the migration rate of habitats to improve the convergence speed and solution quality of the algorithm, and realizes efficient and high-precision edge computing fine-grained scheduling tasks.

[0011] To solve the above technical problems, the technical scheme adopted by the present application is as follows:

[0012] An edge computing fine-grained scheduling method based on IBBO algorithm, comprising the following steps:

[0013] S1: Real-time acquisition of each terminal device's to-be-processed task, task decomposition of the to-be-processed task greater than the preset threshold according to the specified manner, obtaining a plurality of sub-tasks, and marking the to-be-processed task and the to-be-processed sub-task uniformly as to-be-processed sub-tasks;

[0014] S2: Assigning appropriate priority to the to-be-processed sub-tasks according to the importance and urgency of the tasks;

[0015] S3: based on the to-be-processed sub-tasks and their corresponding priorities, a multi-objective optimization model for task scheduling is constructed, a fitness function is set for the multi-objective optimization model, and the optimal solution of the fitness function is a comprehensive optimal solution based on the optimal time delay and energy consumption of task scheduling;

[0016] S4: based on the IBBO algorithm, the comprehensive optimal solution of the fitness function of the multi-objective optimization model is solved;

[0017] S5: based on the obtained comprehensive optimal solution, the to-be-processed sub-tasks are scheduled.

[0018] Preferably, the specific process of decomposing the to-be-processed task into multiple sub-tasks in step S1 is as follows:

[0019] S11: a task data shard key is set, and a preset hash algorithm is called to calculate the hash value of the task data shard key;

[0020] S12: based on the calculated hash value, the shard area where the task data is placed is determined to decompose the to-be-processed task.

[0021] Preferably, the specific process of assigning corresponding priorities to the to-be-processed sub-tasks according to the importance and urgency of the tasks in step S2 is as follows:

[0022] S21: labels of urgency and importance are added to each task, and the urgency and importance are set to a specified number of levels;

[0023] S22: an initial task processing queue is first created according to the level of urgency, and the initial task processing queue is adjusted according to the importance of the tasks among the tasks of the same level of urgency, to obtain an adjusted task processing queue;

[0024] S23: labels of urgency and importance are added to the newly generated to-be-processed task, and the adjusted task processing queue is dynamically adjusted based on step S22.

[0025] Preferably, the specific formula of the fitness function set for the multi-objective optimization model in step S3 is as follows:

[0026] ;

[0027] ;

[0028] ;

[0029] wherein, represents the minimum value of the comprehensive target; y 1 represents the energy consumption of fine-grained task scheduling; y2 represents the fine-grained task scheduling delay; and respectively represent the adjustment factors of the task scheduling energy consumption and the task scheduling delay; and respectively represent the energy consumption and the delay of the task i when processed locally; p i represents the scheduling decision factor, when p i =0, it represents that the task i is processed locally, when p i =1, the task i executes the scheduling processing; q ij represents the edge server allocation factor, when q ij =0, the task i is not allocated to the edge server j for scheduling, when q ij =1, the task i is allocated to the edge server j for scheduling, 、 respectively represent the energy consumption and the delay of the task i processed on the edge node; m represents the number of edge servers.

[0030] Preferably, the specific process of solving the comprehensive optimal solution of the fitness function of the multi-objective optimization model based on the IBBO algorithm in step S4 is as follows:

[0031] S41: based on the initial delay and energy consumption of the edge server node, set the initial parameters of the IBBO algorithm, create an initial population size NP and a random population individual H, each of which represents a potential task scheduling scheme, and build an initial solution space for each individual by allocating different edge server nodes;

[0032] S42: based on the calculation of the delay and energy consumption indicators, evaluate the fitness of each individual through the fitness function, obtain the performance of each individual in the current task scheduling environment, and determine the quality of each individual;

[0033] S43: adjust the disturbance degree of the differential vector based on the adaptive dynamic migration mechanism to search for new individuals;

[0034] S44: obtain the individual mutation rate m i and habitat probability P iAccording to the fitness and migration probability of the individual, an individual suitable for mutation is selected;

[0035] S45: The fitness of the individual is continuously optimized by iteration until a preset iteration number or a preset termination condition is met, population optimization is realized, a comprehensive optimal solution is converged, and an optimal habitat individual H is obtained, i.e., an optimal edge computing fine-grained task scheduling scheme.

[0036] Preferably, step S43 further comprises adjusting the migration probability between individuals according to the calculation result, accurately adjusting the migration rate by normalizing the HSI value of the individual and combining the maximum and minimum HSI values in the current population, and retaining excellent individuals.

[0037] Preferably, the specific process of step S43 for adjusting the disturbance degree of the difference vector based on the adaptive dynamic migration mechanism to search for a new individual is as follows:

[0038] S431: Calculate the HSI of each habitat x i In the population H , λ i and the immigration rate μ i and the emigration rate

[0039] ;

[0040] wherein, F i is the HSI of the habitat x i , F max and F min are the maximum and minimum values of the HSI in the current population, respectively;

[0041] S432: Calculate the migration probability of each individual based on the immigration rate λ i and the emigration rate μ i , and the specific calculation formula is as follows:

[0042] ;

[0043] wherein, x kj is the habitat to be migrated, a and b are random values between [1, NP ]; x aj andx bj respectively represent the first a and the second b habitat, the first j dimensional variable; r max and r min respectively are the maximum minimum interference degrees.

[0044] The beneficial effects of the present application include:

[0045] The edge computing fine-grained scheduling method based on the IBBO algorithm provided by the present application, real-time acquisition of the to-be-processed tasks of each terminal device, task decomposition of the to-be-processed tasks greater than the preset threshold according to the specified manner to obtain to-be-processed sub-tasks; assigning corresponding priorities to the to-be-processed sub-tasks according to the importance and urgency of the tasks; based on the to-be-processed sub-tasks and the corresponding priorities, constructing a multi-objective optimization model of task scheduling, setting a fitness function for the multi-objective optimization model, and the optimal solution of the fitness function is the comprehensive optimal solution based on the optimal time delay and energy consumption of task scheduling; solving the comprehensive optimal solution of the fitness function of the multi-objective optimization model based on the IBBO algorithm; scheduling the to-be-processed sub-tasks based on the obtained comprehensive optimal solution.

[0046] Firstly, in the existing biogeography optimization algorithm, an adaptive migration mechanism and an improved migration strategy are added, which effectively solves the task scheduling optimization problem in the edge computing environment, dynamically adjusts the migration probability between individuals according to the fitness and distribution of the population, speeds up the optimization process and improves the search efficiency.

[0047] Secondly, through the improved migration strategy, the disturbance degree of the differential vector is dynamically adjusted to search for new individuals, so as to enhance the global search ability of the algorithm, avoid falling into local optimal solution, and give a new optimization scheme for the fine-grained task scheduling problem in the edge computing environment, effectively improve the efficiency and performance of task scheduling, reduce the time delay and energy consumption, improve the optimization effect, and realize the comprehensive optimization of task scheduling in the edge computing environment. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 It is the architecture schematic diagram of the edge computing task scheduling model of the present application.

[0049] Figure 2 It is the flowchart of the edge computing fine-grained scheduling method based on the IBBO algorithm of the present application. DETAILED DESCRIPTION

[0050] The following will be combined with the attached Figures 1-2 Further detailed description of the present application:

[0051] Example 1

[0052] Referring to the accompanying drawings Figure 1 and Figure 2 As shown in the figure, an edge computing fine-grained scheduling method based on the IBBO algorithm includes the following steps:

[0053] S1: Real-time acquisition of each terminal device's to-be-processed task, task decomposition of the to-be-processed task greater than the preset threshold according to the specified manner, obtaining a plurality of sub-tasks, and marking the to-be-processed task and the to-be-decomposed to-be-processed task as to-be-processed sub-tasks. Large-scale tasks are decomposed into smaller sub-tasks, and the execution time of each sub-task is relatively short. On the one hand, the task can be assigned to different computing nodes for parallel execution, thereby improving the concurrency of the task and the overall execution efficiency, and on the other hand, the performance requirement of the server for large-scale task data is reduced. Before task decomposition, the large-scale data can only be processed through the central cloud server, and after decomposition of the large-scale data, a plurality of sub-tasks are obtained, so that the edge server node can also process the sub-tasks. Therefore, while reducing the processing performance of the server, the processing efficiency is improved through the simultaneous processing of multiple servers.

[0054] S2: Assigning a corresponding priority to the to-be-processed sub-task according to the importance and urgency of the task, and giving different priorities to different tasks according to the importance and urgency of the task. In the case of limited resources or high system load, important or urgent tasks are processed first to ensure the stability and performance of the system. Fine-grained scheduling needs to monitor the state and resource utilization of the system in real time, and dynamically adjust the allocation and scheduling strategy of the task according to the current load. Through dynamic scheduling, flexible task scheduling can be performed according to real-time demand and resource changes, maximizing the resource utilization and performance of the system. Fine-grained scheduling also includes dynamic allocation and management of computing resources. According to the requirements of the task and the load of the system, computing, storage and network resources are dynamically allocated to maximize resource utilization and meet the requirements of task execution.

[0055] S3: Based on the to-be-processed sub-tasks and their corresponding priorities, a multi-objective optimization model for task scheduling is constructed, and a fitness function is set for the multi-objective optimization model, and the optimal solution of the fitness function is the comprehensive optimal solution based on the optimal delay and energy consumption of task scheduling;

[0056] S4: Solving the comprehensive optimal solution of the fitness function of the multi-objective optimization model based on the IBBO algorithm;

[0057] S5: Scheduling the to-be-processed sub-tasks based on the obtained comprehensive optimal solution.

[0058] The edge computing task scheduling model of the application comprises a local server and a central cloud server, a communication connection is established between the local server and the central cloud server through a network, and the local server and the central cloud server establish a communication connection with a base station and an edge cloud through the network.

[0059] In the fine-grained task scheduling process, the dependency relationship between tasks needs to be considered to ensure that the tasks are executed in the correct order and the integrity and consistency of the entire task flow are guaranteed. The scheduling algorithm needs to consider the dependency relationship between tasks and reasonably arrange the execution order of the tasks to avoid deadlocks or incorrect task execution order.

[0060] The traditional biogeography optimization algorithm BBO only determines the immigration and emigration rate based on the excellent order of the individual quality, ignoring the overall evolution state of the current population. Such processing method may lead to unreasonable evaluation of the individual quality, especially in the case of large difference or uneven distribution between individuals, which is easy to make wrong evaluation of the quality of individuals, resulting in that the beneficial information of better individuals cannot be fully retained, or the poor individuals participate in the evolution process too early.

[0061] The application improves the existing biogeography optimization algorithm BBO, based on the importance of the migration rate and the population migration mechanism in the BBO algorithm to the overall search speed and optimization ability, the iteration strategy and parameter setting of the algorithm are improved, and the corresponding improvement is made to the decision mechanism of the migration rate and the migration strategy of the population.

[0062] Example 2

[0063] On the basis of example 1, the specific process of decomposing the plurality of subtasks by decomposing the task according to the specified manner in step S1 is as follows:

[0064] S11: setting a task data sharding key, and calling a preset hash algorithm to calculate the hash value of the task data sharding key;

[0065] S12: determining the sharding area where the task data is placed based on the calculated hash value to decompose the to-be-processed task.

[0066] In the process of decomposing the task data, all the task data in the database is processed by using a unified hash algorithm, so that the hash function can uniformly distribute the data and reduce the hotspot risk caused by uneven data.

[0067] In another implementation of the embodiment, the large-scale data is divided according to the range of the task data value or the key space, and adjacent shard keys are more likely to fall into the same shard. In this process, a suitable shard key needs to be selected, and the shard key does not contain repeated values, so that the candidate values are as discrete as possible.

[0068] The specific process of assigning a corresponding priority to the to-be-processed subtask according to the importance and urgency of the task in step S2 is as follows:

[0069] S21: Add labels of urgency and importance to each task, and set the urgency and importance to a specified number of levels;

[0070] S22: First, create an initial task processing queue according to the level of urgency, and then adjust the initial task processing queue according to the importance of the task among the tasks of the same level of urgency, to obtain an adjusted task processing queue;

[0071] S23: Add labels of urgency and importance to the newly generated to-be-processed task, and dynamically adjust the adjusted task processing queue based on step S22.

[0072] Embodiment 3

[0073] On the basis of Embodiment 1 or Embodiment 2, the specific formula of the fitness function of the multi-objective optimization model in step S3 is as follows:

[0074] ;

[0075] ;

[0076] ;

[0077] wherein, represents the minimum value of the comprehensive target; y 1 represents the energy consumption of fine-grained task scheduling; y 2 represents the delay of fine-grained task scheduling; and respectively represent the adjustment factors of the energy consumption and the delay of task scheduling; and respectively represent the energy consumption and the delay of the task i when processed locally; p i represents a scheduling decision factor, when p i = 0, it indicates that the task i is processed locally, and when p i = 1, the task i performs scheduling processing;q ij represent the edge server assignment factor, when q ij =0, the task i is not assigned to the edge server j for scheduling, when q ij =1, the task i is assigned to the edge server j for scheduling, , respectively represent the energy consumption and the time delay of the task i processed on the edge node; m represent the number of edge servers.

[0078] The fitness function plays a crucial role in fine-grained task scheduling, which comprehensively considers the energy consumption and time delay of task scheduling in order to find the optimal task scheduling scheme. Task scheduling not only needs to consider the efficiency of resource utilization, but also needs to consider the delay of task execution, which are often mutually restrictive. Therefore, the present application models the task scheduling problem as a multi-objective optimization problem based on the above fitness function, in order to better balance the trade-off relationship between energy consumption and time delay.

[0079] The specific process of solving the comprehensive optimal solution of the fitness function of the multi-objective optimization model based on the IBBO algorithm (Improved Biogeography-Based Optimization Algorithm) in step S4 is as follows:

[0080] S41: Set the initial parameters of the IBBO algorithm based on the initial time delay and energy consumption of the edge server node, create the initial population size NP and the random population individual H, H{ x i , i =1, 2, …, NP}, each of which represents a potential task scheduling scheme, and an initial solution space is built for each individual by assigning different edge server nodes, providing a starting point for subsequent optimization. Set the initial parameters of the IBBO algorithm, including the relevant characteristics of the edge server node, such as initial time delay and energy consumption. Accurate setting of initial parameters lays the foundation for subsequent optimization process and ensures the effective operation of the algorithm.

[0081] S42: Evaluate the fitness of each individual based on the time delay and energy consumption indicators through the fitness function, obtain the performance of each individual in the current task scheduling environment, and determine the quality of each individual. The fitness of each individual is evaluated, i.e. the performance in the current task scheduling environment. By calculating indicators such as time delay and energy consumption, the quality of each individual is determined, providing a reference standard for subsequent optimization process.

[0082] S43: Adjust the disturbance degree of the difference vector based on the adaptive dynamic migration mechanism to search for a new individual.

[0083] The migration strategy of the traditional BBO directly affects the search ability of the algorithm as an important part of its evolution mechanism, but the migration mode of directly replacing the SIV of the original solution with the SIV to be migrated may cause certain blindness in migration, the ability to mine new solutions is poor, and premature convergence is easily caused. Therefore, the improved dynamic migration mechanism is arranged, the adaptive migration rate is used to adjust the disturbance degree of the difference vector to search for a new individual, so that adaptive disturbance around the SIV to be migrated in different evolution periods is realized, and the search ability of the migration algorithm is enhanced.

[0084] The migration strategy of BBO is a crucial part of its evolution mechanism, which directly affects the search ability of the algorithm. However, in the traditional migration strategy, the SIV (Solution Improvement Vector) to be migrated is directly used to replace the SIV of the original solution, which may have certain blindness, resulting in poor ability to mine new solutions, and even easily causing premature convergence. Therefore, the improved dynamic migration mechanism is designed, and the adaptive migration rate is used to adjust the disturbance degree of the difference vector, so as to more flexibly search for new individuals. The search process around the SIV to be migrated can adaptively and moderately disturb in different evolution periods. This improved process can enhance the search ability of the migration algorithm and improve the exploration efficiency of the algorithm on the solution space.

[0085] S44: Obtain the individual mutation rate of each solution m i and habitat probability P i According to the fitness and migration probability of the individual, the individual suitable for mutation is selected to explore new possibilities in the solution space.

[0086] S45: Iteratively execute the continuous optimization of individual fitness until the preset number of iterations or the preset termination condition is met, realize population optimization, converge to a comprehensive optimal solution, obtain the optimal habitat individual H, and obtain the optimal edge computing fine-grained task scheduling scheme.

[0087] In the traditional BBO algorithm, the immigration and emigration rates are determined only based on the good order of individual quality, thus ignoring the overall evolution state of the current population. The determination of the immigration and emigration rates based on the good order of individual quality may lead to an unreasonable evaluation of the individual quality. Especially in the case of a large gap between individuals or uneven distribution, the good and bad conditions of individuals are easily evaluated incorrectly, resulting in that the beneficial information of better individuals cannot be fully retained, or that worse individuals prematurely participate in the evolution process. In order to solve this problem, the traditional BBO algorithm is improved in the present application, and the migration rate is adaptively adjusted based on the normalized HSI value of the individual.

[0088] Step S43 further comprises adjusting the migration probability between individuals according to the calculation result, accurately adjusting the migration rate by normalizing the HSI value of the individual and combining the maximum and minimum HSI values in the current population, and retaining excellent individuals.

[0089] Step S43 adjusts the disturbance degree of the difference vector based on the adaptive dynamic migration mechanism to search for new individuals, and the specific process is as follows:

[0090] S431: Calculate the HSI of each habitat x i In the population H The immigration rate λ i And the emigration rate μ i , and adjust the migration probability between individuals according to the calculation result, and the specific calculation formula is as follows:

[0091] ;

[0092] Wherein, F i is the HSI of the habitat x i F max , and F min The maximum and minimum values of HSI in the current population, respectively;

[0093] S432: Calculate the migration probability of each individual based on the immigration rate λ i And the emigration rate μ i , and the specific calculation formula is as follows:

[0094] ;

[0095] Wherein, x kj ​is a habitat to be migrated, a and b is a random value between [1, NP ] and [0, x aj and x bj respectively represent the first a and the first b dimensional variable of the first j habitat; r max and r min respectively are the maximum and minimum interference degrees.

[0096] To sum up, the edge computing fine-grained scheduling method based on the IBBO algorithm provided by the application acquires the to-be-processed tasks of each terminal device in real time, decomposes the to-be-processed tasks greater than the preset threshold according to the specified manner to obtain to-be-processed subtasks, allocates corresponding priorities to the to-be-processed subtasks according to the importance and urgency of the tasks, constructs a multi-objective optimization model of task scheduling based on the to-be-processed subtasks and the corresponding priorities, sets a fitness function for the multi-objective optimization model, and the optimal solution of the fitness function is a comprehensive optimal solution based on the optimal time delay and energy consumption of task scheduling; the comprehensive optimal solution of the fitness function of the multi-objective optimization model is solved based on the IBBO algorithm; and the to-be-processed subtasks are scheduled based on the obtained comprehensive optimal solution.

[0097] By adding an adaptive migration mechanism and an improved migration strategy on the basis of the existing biogeography optimization algorithm, the task scheduling optimization problem in the edge computing environment is effectively solved, the migration probability between individuals is dynamically adjusted according to the fitness and distribution of the population, the optimization process is accelerated, and the search efficiency is improved. Through the improved migration strategy, the disturbance degree of the differential vector is dynamically adjusted to search for new individuals, so as to enhance the global search ability of the algorithm, avoid falling into a local optimal solution, and give a new optimization scheme for the fine-grained task scheduling problem in the edge computing environment, effectively improve the efficiency and performance of task scheduling, reduce the time delay and energy consumption, improve the optimization effect, and realize the comprehensive optimization of task scheduling in the edge computing environment.

Claims

1. A fine-grained scheduling method for edge computing based on the IBBO algorithm, characterized in that: The following steps are involved: S1: Obtain pending tasks from each terminal device in real time, decompose pending tasks that are greater than a preset threshold according to a specified method to obtain multiple subtasks, and mark them together with undecomposed pending tasks as pending subtasks; S2: Assign corresponding priorities to the subtasks to be processed according to their importance and urgency; S3: Based on the pending subtasks and their corresponding priorities, a multi-objective optimization model for task scheduling is constructed, and a fitness function is set for the multi-objective optimization model. The optimal solution of the fitness function is a comprehensive optimal solution based on the optimal delay and energy consumption of task scheduling; S4: Solve the comprehensive optimal solution of the fitness function of the multi-objective optimization model based on the IBBO algorithm; S5: Schedule the subtasks to be processed based on the obtained comprehensive optimal solution; The specific formula for setting the fitness function for the multi-objective optimization model in step S3 is as follows: ； ； ； in, Represents the minimum value of the comprehensive target; y 1 represents the energy consumption of fine-grained task scheduling; y 2 represents the fine-grained task scheduling latency; and They represent the adjustment factors of task scheduling energy consumption and task scheduling delay respectively; and Represents tasks i Energy consumption and latency when processing locally; p i represents the scheduling decision factor, when p i =0, indicating the task i When processed locally, p i =1, the task i Execute scheduling processing; q ij represents the edge server allocation factor, when q ij =0, the task i Not assigned to an edge server j To schedule, q ij =1, the task i Assigned to edge servers j To schedule, 、 Represents tasks i Energy consumption and latency of processing at edge nodes; m Represents the number of edge servers; The specific process of solving the comprehensive optimal solution of the fitness function of the multi-objective optimization model based on the IBBO algorithm in step S4 is as follows: S41: Setting initial parameters of the IBBO algorithm based on the initial latency and energy consumption of the edge server nodes, creating an initial population size NP and random population individuals, each of which represents a potential task scheduling solution, and building an initial solution space for each individual by assigning different edge server nodes; S42: Evaluate the fitness of each individual using the fitness function based on the computational latency and energy consumption indicators, obtain the performance of each individual in the current task scheduling environment, and determine the quality of each individual; S43: Based on the adaptive dynamic migration mechanism, the perturbation degree of the differential vector is adjusted to search for new individuals; S44: Get the individual mutation rate of each solution m i and habitat probability P i , according to the individual’s fitness and migration probability, select individuals suitable for mutation; S45: Iterative execution continuously optimizes individual fitness until the preset number of iterations or the preset termination condition is met, achieving population optimization and converging to the comprehensive optimal solution. The optimal habitat individual is obtained, which is the optimal edge computing fine-grained task scheduling solution; The specific process of step S43, which adjusts the disturbance degree of the differential vector based on the adaptive dynamic migration mechanism to search for new individuals, is as follows: S431: Calculate each habitat x i The immigration rate in the population λ i and emigration rate μ i , and adjust the migration probability between individuals according to the calculation results. The specific calculation formula is as follows: ； in, F i It is a habitat x i HSI, F max and F min are the maximum and minimum values ​​of HSI in the current population, respectively; S432: Based on immigration rate λ i and emigration rate μ i Calculate the migration probability of each individual. The specific calculation formula is as follows: ； in, x kj It is a habitat to be migrated. a and b is between [1, NP ]’s random value; x aj and x bj Respectively represent a Hedi b The first habitat j dimensional variables; r max and r min are the maximum and minimum interference respectively.

2. The edge computing fine-grained scheduling method based on the IBBO algorithm according to claim 1 is characterized in that: In step S1, the specific process of decomposing the pending tasks that are greater than the preset threshold in a specified manner to obtain multiple subtasks is as follows: S11: Setting a task data sharding key, and calling a preset hash algorithm to calculate a hash value of the task data sharding key; S12: Determine the sharding area where the task data is placed based on the calculated hash value and decompose the task to be processed.

3. The edge computing fine-grained scheduling method based on the IBBO algorithm according to claim 1 is characterized in that: The specific process of assigning corresponding priorities to the subtasks to be processed according to the importance and urgency of the tasks in step S2 is as follows: S21: Add urgency and importance marks to each task, and set the urgency and importance to specified levels; S22: First, an initial task processing queue is created according to the level of urgency. Among tasks of the same level of urgency, the initial task processing queue is adjusted according to the importance of the tasks to obtain an adjusted task processing queue. S23: Add urgency and importance tags to the newly generated pending tasks, and dynamically adjust the adjusted task processing queue based on step S22.

4. The edge computing fine-grained scheduling method based on the IBBO algorithm according to claim 1 is characterized in that: Step S43 also adjusts the migration probability between individuals based on the calculation results. By normalizing the HSI value of the individual and combining it with the maximum and minimum HSI values ​​in the current population, the quality of individuals in the population is monitored in real time, the migration rate is accurately adjusted, and the excellent individuals are retained.

Images