Optimal Selection Method for Task Offloading Strategies in Cloud-Edge-Endpoint Collaborative Internet of Things Edge Computing Systems

Abstract

The present invention provides a method for optimizing a task offloading strategy for a cloud-edge-end collaborative Internet of Things edge computing system, comprising the following steps: constructing a model of the cloud-edge-end collaborative Internet of Things edge computing system, in which the tasks of each terminal device are divided into three parts and processed in parallel on a cloud server, an edge server, and a terminal device; obtaining a minimum weight matching based on the distance between the terminal device and each edge server according to the model of the cloud-edge-end collaborative Internet of Things edge computing system, and determining the obtained result as the offloading target S; after determining the offloading target S, performing w iterative operations with the offloading rate α and the computing resource f as variables to optimize the offloading rate α and the computing resource f; after the wth iteration, calculating the system effect generated by utilizing the current task offloading strategy according to the system effect function, and performing convergence verification. The present invention effectively reduces task processing delay, improves task processing efficiency, and increases system effect.

Description

Technical Field

[0001] The present invention belongs to the field of edge computing of the Internet of Things, and specifically relates to a method for optimizing task offloading strategies in a cloud-edge-end collaborative Internet of Things edge computing system. Background Art

[0002] Edge computing (EC) systems typically consist of cloud servers, multiple edge servers, and a large number of end users. End users communicate with servers within their communication range via wireless channels. Based on the attributes of end-user tasks and information such as computing resources, the system designs a rational offloading strategy to improve task processing efficiency and enhance system effectiveness. The most straightforward task offloading strategy is to set a fixed offloading target and offloading rate, and allocate a fixed amount of computing resources to process tasks from the end-users. This approach offers advantages in terms of scalability and dynamic adaptability. Another common offloading strategy is to offload tasks to the server closest to the device and optimize the offloading rate and computing resources. This strategy allocates different offloading rates and computing resources based on the attributes of the end-user tasks, effectively reducing task processing latency and improving system quality of service. However, task offloading in IoT edge computing systems still faces several key challenges: 1) how to select a rational offloading target to avoid server overload; and 2) how to optimize the offloading rate and computing resources to improve task processing efficiency and system effectiveness.

[0003] In edge computing systems, edge servers are typically small, private servers. Compared to cloud servers, edge servers have limited resources. During edge computing task offloading, some tasks may be complex and difficult to handle solely on resource-constrained edge servers. A viable solution is to introduce cloud servers and design an edge computing task offloading strategy based on cloud-edge-device collaboration.

[0004] Currently, many studies focus on edge computing task offloading strategies for cloud-edge-device collaboration. Zhao et al. (see J. Zhao, Q. Li, Y. Gong, and K. Zhang. Computation offloading and resource allocation for cloud-assisted mobile edge computing in vehicular networks. IEEE Transactions on Vehicular Technology, vol. 68, no. 8, pp. 7944–7956, 2019) proposed a task offloading strategy based on the collaboration of mobile edge computing and cloud computing, and further proposed a collaborative optimization algorithm to optimize the offloading rate and computing resources to improve system effectiveness. Shen et al. (see B. Shen, X. Xu, F. Dai, L. Qi, X. Zhang, and W. Dou. Dynamic task offloading with minority game for internet of vehicles in cloud-edge computing. in 2020 IEEE International Conference on Web Services (ICWS), 2020, pp. 372–379) proposed a dynamic task offloading strategy based on game theory to optimize the offloading target, offloading rate, and computing resources.

[0005] However, the overall defect of the above-mentioned IoT edge computing task offloading strategy is that the model is relatively simple, that is, it assumes that during the offloading process, the task can only select one of the cloud server and the edge server as the offloading target, and the system effect still needs to be improved. Summary of the Invention

[0006] In order to solve at least some of the above-mentioned problems existing in the prior art, the present invention provides a method for optimizing task offloading strategies in a cloud-edge-end collaborative Internet of Things edge computing system.

[0007] The present invention is achieved in that:

[0008] The present invention provides a method for optimizing a task offloading strategy for a cloud-edge-device collaborative Internet of Things edge computing system, the method comprising the following steps:

[0009] Build a model of a cloud-edge-device collaborative IoT edge computing system. The model includes a central cloud server, M edge servers, and N terminal devices. In the model, the tasks of each terminal device are divided into three parts and processed in parallel on the cloud server, edge server, and terminal device.

[0010] According to the model of the cloud-edge-device collaborative IoT edge computing system, a weighted bipartite graph is constructed based on the distance between the terminal device and each edge server. The KM matching method is used to find the minimum weight matching, and the result is determined as the offloading target S. After the offloading target S is determined, w iterative operations are performed with the offloading rate α and the computing resource f as variables to optimize the offloading rate α and the computing resource f.

[0011] After the wth iteration, the system effect generated by the current task offloading strategy is calculated according to the system effect function, and the convergence is verified. If the system effect converges to a relatively stable value, the iteration process ends and the current task offloading strategy is the preferred task offloading strategy; otherwise, the iteration continues until the system effect converges.

[0012] Furthermore, in the model, each terminal device V i Contains a calculation task, whose properties are Among them D i Indicates the task size, R i represents the computing resources required to process the task, Represents the maximum tolerable delay in completing the task, then regarding the terminal device V i The system effect function is:

[0013] U i =β P (c v (α i,e +α i,c )D i R i -c s (f i,j +f i,CS ))+β Q log(1+ε-T i )

[0014] Among them, U i Indicates processing terminal device V i The system effect of the task,

[0015] c v (α i,e +α i,c )D i R i -c s (f i,j +f i,CS ) represents the profit, f i,j , f i,CS They represent the edge servers assigned to the terminal device V i Computing resources and cloud servers are allocated to terminal devices V i Computational resources, βP represents the weight of the income, c v , c s They represent the unit price that the system charges the terminal device for processing tasks and the unit price of the system renting the server, α i,e , α i,c Represents the terminal device V i The offloading rate when offloading tasks to edge servers and cloud servers, log(1+ε-T i ) indicates the terminal device V i Quality of service β Q Represents the weight of service quality, ε is used to ensure T i Indicates terminal device V i Task processing delay;

[0016] The system effect generated by utilizing the current task offloading strategy is the sum of the system effects generated by processing tasks of each terminal device.

[0017] Furthermore, during the parallel processing of tasks on the cloud server, edge server, and terminal device, the task processing delay includes three parts: communication delay, calculation delay, and movement delay; the terminal device V i The task processing delay is determined as the larger value of the local processing delay, the cloud server processing delay, and the edge server processing delay.

[0018] Furthermore, the uninstall rate α includes the terminal device V i The offloading rate α when offloading tasks to edge servers i,e and terminal equipment V i Offloading rate α when offloading tasks to the cloud server i,c The method of performing w iterative operations with the offload rate α and the computing resource f as variables to optimize the offload rate α and the computing resource f specifically includes:

[0019] 1) Fixed α i,e ,f, using genetic algorithm to optimize α i,c ;

[0020] 2) Fixed α i,c , α i,e , optimize f using the interior point method;

[0021] 3) Fixed α i,c , f, using Lagrange multiplier method and KKT conditions to optimize α i,e .

[0022] Furthermore, the fixed α i,c , f, using Lagrange multiplier method and KKT conditions to optimize αi,e , specifically including:

[0023] The Lagrangian function is constructed as follows:

[0024]

[0025] According to the Lagrangian function, the KKT condition is written:

[0026]

[0027] η i (α i,e +α i,c -1)=0

[0028] θ i (0-α i,e )=0

[0029] ζ i ≥0,η i ≥0,θ i ≥0

[0030]

[0031] According to the Lagrangian function and KKT conditions, solve α i,e .

[0032] Furthermore, when determining the offloading target S, optimizing the offloading rate α and computing resources f, the following constraints must be met:

[0033] 1) The tasks of each terminal device must be completed within the maximum tolerable delay for the terminal device to complete the task;

[0034] 2) Uninstall rate is between 0-1;

[0035] 3) The terminal device can only select one edge server as the offloading target;

[0036] 4) The sum of computing resources allocated to each terminal device must be less than the total resources of the edge server.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The task offloading strategy optimization method of the cloud-edge-end collaborative Internet of Things edge computing system provided by the present invention can simultaneously utilize cloud servers and edge servers, effectively reducing task processing delays, improving task processing efficiency, and increasing system effects; in the process of task offloading, a reasonable offloading target is selected with the distance between the sensor and each edge server as the weight, and the maximum number of connections is set according to the resource margin of the edge server, which can avoid server overload. Simulation shows that the performance of the present invention is superior to existing methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A flowchart of a method for optimizing a task offloading strategy for a cloud-edge-device collaborative IoT edge computing system provided by an embodiment of the present invention;

[0040] Figure 2 A schematic diagram of a task offloading strategy for a cloud-edge-device collaborative IoT edge computing system provided by an embodiment of the present invention;

[0041] Figure 3 Schematic diagram comparing task processing delays of the task offloading strategy obtained by the present invention and the task offloading strategy obtained by the prior art when the number of iterations N=30;

[0042] Figure 4 Schematic diagram comparing the system effects of the task offloading strategy obtained by the present invention and the task offloading strategy obtained by the existing method as the number of sensors changes. DETAILED DESCRIPTION

[0043] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only 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 making creative efforts are within the scope of protection of the present invention.

[0044] like Figure 1 As shown, an embodiment of the present invention provides a method for optimizing a task offloading strategy for a cloud-edge-device collaborative IoT edge computing system, the method comprising the following steps:

[0045] S1. Build a model of the cloud-edge collaborative IoT edge computing system. The model includes a central cloud server, M edge servers, and N terminal devices, where M and N are natural numbers greater than zero, and each terminal device V i (1≤i≤N) contains a computing task whose attributes are Among them D i Indicates the task size, R i represents the computing resources required to process the task, The maximum tolerable delay for completing a task. In the model, each terminal device's task is divided into three parts and processed in parallel on the cloud server, edge server, and terminal device. During this process, the task processing delay includes communication delay, computation delay, and mobile delay.

[0046] Communication delay is the time required for the terminal device to offload the task to the server. i With edge server E jThe communication delay between The communication delay with the cloud server is where α i,e , α i,c Represents the terminal device V i The offloading rate when offloading tasks to edge servers and cloud servers, ρ i,j Indicates terminal device V i With edge server E j The data transmission rate between i,CS Indicates terminal device V i Data transfer rate to and from the cloud server;

[0047] The computational delay of the task includes local computational delay, edge server computational delay, and cloud server computational delay. i Using its own computing resources to process (1-α i,e -α i,c )D i This part of the task, and the rest is offloaded to the cloud server and edge server for processing, then the terminal device V i The local computing delay, edge server computing delay, and cloud server computing delay are expressed as f i,V , f i,j , f i,CS Represents the terminal device V i Its own computing resources and edge servers are allocated to terminal devices V i Computing resources and cloud servers are allocated to terminal devices V i computing resources;

[0048] Mobile delay is the terminal device V i Move to the edge server E selected as the offload target j The moving time within the communication range, i.e.

[0049] Terminal device V i The task processing delay is determined as the larger value among the local processing delay, cloud server processing delay and edge server processing delay, which is expressed as

[0050] For IoT edge computing systems, the system effect includes two parts: revenue and service quality. The main factor affecting service quality is task processing delay. Generally speaking, service quality should be inversely proportional to task processing delay. This paper uses the log function as the service quality model. Regarding the terminal device V i The quality of service can be expressed as β QRepresents the weight of service quality, ε is used to ensure Therefore, the processing terminal device V i The system effect of the task is:

[0051] U i =β P (c v (α i,e +α i,c )D i R i -c s (f i,j +f i,CS ))+β Q log(1+ε-T i )

[0052] Among them, U i Indicates processing terminal device V i The system effect of the task, c v (α i,e +α i,c )D i R i -c s (f i,j +f i,CS ) represents the profit, f i,j , f i,cs They represent the edge servers assigned to the terminal device V i Computing resources and cloud servers are allocated to terminal devices V i Computational resources, β P represents the weight of the income, c v , c s They represent the unit price that the system charges the terminal device for processing tasks and the unit price of the system renting the server, α i,e , α i,c Represents the terminal device V i The offloading rate when offloading tasks to edge servers and cloud servers, log(1+ε-T i ) indicates the terminal device V i Quality of service β Q Represents the weight of service quality, ε is used to ensure T i Indicates terminal device V i Task processing delay;

[0053] The system effect generated by utilizing the current task offloading strategy is the sum of the system effects generated by processing tasks of each terminal device.

[0054] S2. Based on the model of the cloud-edge collaborative IoT edge computing system, a weighted bipartite graph is constructed according to the distance between the terminal device and each edge server, and the KM matching method is used to obtain the minimum weight matching. The result is determined as the offloading target S. After determining the offloading target S, w iterative operations are performed with the offloading rate α and the computing resource f as variables to optimize the offloading rate α and the computing resource f. The offloading rate α includes the terminal device V i The offloading rate α when offloading tasks to edge servers i,e and terminal equipment V i Offloading rate α when offloading tasks to the cloud server i,c .

[0055] Since the offloading rate α and computing resources f are highly coupled, after optimizing the offloading target S, α and f are decoupled and α and f are optimized separately. i,e , α i,c and f, specifically including:

[0056] 1) Fixed α i,e ,f, using genetic algorithm to optimize α i,c ;

[0057] 2) Fixed α i,c , α i,e , optimize f using the interior point method;

[0058] 3) Fixed α i,c , f, using Lagrange multiplier method and KKT conditions to optimize α i,e ; Specifically include:

[0059] The Lagrangian function is constructed as follows:

[0060]

[0061] According to the Lagrangian function, the KKT condition is written:

[0062]

[0063] η i (α i,e +α i,c -1)=0

[0064] θ i (0-α i,e )=0

[0065] ζ i ≥0,η i ≥0,θ i ≥0

[0066]

[0067] According to the Lagrangian function and KKT conditions, solve α i,e .

[0068] Preferably, in the process of step S2, the following constraints must be met when determining the offloading target S, optimizing the offloading rate α and the computing resources f:

[0069]

[0070] Among them, constraint (1) ensures that the task of each terminal device must be completed within the maximum tolerable delay of the terminal device to complete the task; constraint (2)(3)(4) indicates that the offloading rate must be between 0 and 1, constraint (5)(6) indicates that the sensor can only select one edge server as the offloading target, and constraint (7) ensures that the sum of the computing resources allocated to each sensor must be less than the total resources of the edge server.

[0071] S3. After the wth iteration, the system effect generated by the current task offloading strategy is calculated according to the system effect function, and the convergence is verified. If the system effect converges to a relatively stable value, the iteration process is terminated and the current task offloading strategy is the preferred task offloading strategy; otherwise, the iteration is continued until the system effect converges.

[0072] The present invention implements the above-mentioned cloud-edge-device collaborative IoT edge computing system model. For given server information, sensor information, and task information, the system effect is maximized by jointly optimizing the offloading target, offloading rate, and computing resources on the basis of satisfying the above-mentioned constraints:

[0073]

[0074] Now combined with the attached Figure 2 A specific embodiment of the present invention is described.

[0075] like Figure 2 As shown in the figure, there are 3 edge servers and 7 sensors in the cloud-edge collaborative IoT edge computing system. The sensor is the terminal device. Assuming that each sensor has only one task, its attributes are In any k-th iteration, a weighted bipartite graph is constructed based on the distance between the sensor and each edge server, and the weighted maximum matching is obtained using KM matching, and then the offloading target S is obtained. Figure 2 As shown, sensor V4 selects edge server E1 as the offloading target, V3 selects edge server E2 as the offloading target, and V6 and V7 select edge server E3 as the offloading target.

[0076] After determining the offloading target, the offloading rate α and computing resources f are optimized using genetic algorithm, interior point method, Lagrangian relaxation formula and KKT condition according to information such as task attributes, sensor motion status and server resource quantity, thereby obtaining an efficient task offloading strategy.

[0077] The system effect generated by the current task offloading strategy is calculated according to the system effect function, and the convergence is verified. If the system effect converges to a relatively stable value, the iteration process ends; otherwise, the iteration will continue until the system effect converges.

[0078] like Figure 3 As shown, it is a schematic diagram comparing the task processing delay of the task offloading strategy obtained by the present invention and the task offloading strategy obtained by the existing method when the number of iterations N=30; Figure 4 The figure shows a comparison of the system effects of the task offloading strategy obtained by the present invention and the task offloading strategy obtained by the existing method as the number of sensors changes. Figure 3 、 4 It can be seen that the performance of the task offloading strategy obtained by the present invention is significantly better than that obtained by the existing method.

[0079] To sum up, the task offloading strategy optimization method of the cloud-edge-end collaborative Internet of Things edge computing system provided by the present invention can utilize cloud servers and edge servers at the same time, effectively reducing task processing delays, improving task processing efficiency, and increasing system effects; in the process of task offloading, a reasonable offloading target is selected with the distance between the sensor and each edge server as the weight, and the maximum number of connections is set according to the resource margin of the edge server, which can avoid server overload. Simulation shows that the performance of the present invention is superior to existing methods.

[0080] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for optimizing task offloading strategies in a cloud-edge-device collaborative IoT edge computing system, characterized in that: The method comprises the following steps: Build a model of a cloud-edge-device collaborative IoT edge computing system. The model includes a central cloud server, M edge servers, and N terminal devices. In the model, the tasks of each terminal device are divided into three parts and processed in parallel on the cloud server, edge server, and terminal device. According to the model of the cloud-edge-device collaborative IoT edge computing system, a weighted bipartite graph is constructed based on the distance between the terminal device and each edge server. The KM matching method is used to find the minimum weight matching, and the result is determined as the offloading target S. After the offloading target S is determined, w iterative operations are performed with the offloading rate α and the computing resource f as variables to optimize the offloading rate α and the computing resource f. After the wth iteration, the system effect generated by the current task offloading strategy is calculated according to the system effect function, and the convergence is verified. If the system effect converges to a relatively stable value, the iteration process ends and the current task offloading strategy is the preferred task offloading strategy; otherwise, the iteration continues until the system effect converges. In the model, each terminal device V i Contains a calculation task, whose properties are Among them D i Indicates the task size, R i represents the computing resources required to process the task, Indicates the maximum tolerable delay in completing the task, then the terminal device V i The system effect of the task is: U i =b P (c v (a i,e +a i,c )D i R i -c s (f i,j +f i,CS ))+b Q log(1+ε-T i ) Among them, U i Indicates processing terminal device V i The system effect of the task, c v (α i,e + α i,c )D i R i -c s (f i,j +f i,CS ) represents the profit, f i,j , f i,CS They represent the edge servers assigned to the terminal device V i Computing resources and cloud servers are allocated to terminal devices V i Computational resources, β P represents the weight of the income, c v , c s They represent the unit price that the system charges the terminal device for processing tasks and the unit price of the system renting the server, α i,e , α i,c Represents the terminal device V i The offloading rate when offloading tasks to edge servers and cloud servers, log(1+ε-T i ) indicates the terminal device V i Quality of service β Q Represents the weight of service quality, ε is used to ensure T i Indicates terminal device V i Task processing delay; The system effect generated by utilizing the current task offloading strategy is the sum of the system effects generated by processing tasks of each terminal device.

2. The method for optimizing a task offloading strategy for a cloud-edge-device collaborative IoT edge computing system according to claim 1, characterized in that: During the parallel processing of tasks on the terminal device, the task processing delay includes three parts: communication delay, calculation delay, and movement delay. i The task processing delay is determined as the larger value of the local processing delay, the cloud server processing delay, and the edge server processing delay.

3. The method for optimizing task offloading strategies for a cloud-edge-end collaborative IoT edge computing system according to claim 1, wherein: The uninstall rate α includes the terminal device V i The offloading rate α when offloading tasks to edge servers i,e and terminal equipment V i Offloading rate α when offloading tasks to the cloud server i,c The method of performing w iterative operations with the offload rate α and the computing resource f as variables to optimize the offload rate α and the computing resource f specifically includes: 1) Fixed α i,e ,f, using genetic algorithm to optimize α i,c ; 2) Fixed α i,c , α i,e , optimize f using the interior point method; 3) Fixed α i,c , f, using Lagrange multiplier method and KKT conditions to optimize α i,e .

4. The method for optimizing the task offloading strategy of the cloud-edge-end collaborative IoT edge computing system according to claim 3 is characterized in that: The fixed α i,c , f, using Lagrange multiplier method and KKT conditions to optimize α i,e , specifically including: The Lagrangian function is constructed as follows: According to the Lagrangian function, the KKT condition is written: or i (a i,e +a i,c -1)=0 i i (0-a i,e )=0 g i ≥0,η i ≥0,θ i ≥0 According to the Lagrangian function and KKT conditions, solve α i,e .

5. The method for optimizing task offloading strategies for a cloud-edge-device collaborative IoT edge computing system according to claim 1, wherein: When determining the offloading target S and optimizing the offloading rate α and computing resources f, the following constraints must be met: 1) The tasks of each terminal device must be completed within the maximum tolerable delay for the terminal device to complete the task; 2) Uninstall rate is between 0-1; 3) The terminal device can only select one edge server as the offloading target; 4) The sum of computing resources allocated to each terminal device must be less than the total resources of the edge server.

Images