An edge device deployment method

Abstract

This application discloses an edge device deployment method, comprising: acquiring relevant information about the area to be deployed and the edge devices; establishing an edge device deployment optimization model that maximizes the benefits to resource providers based on the relevant information; solving the optimization model using a two-layer deployment strategy algorithm to obtain corresponding device deployment strategies and resource allocation strategies; deploying edge devices according to the device deployment strategies and allocating resources according to the resource allocation strategies. This application constructs the edge device deployment problem as a two-layer model, decoupling the complex edge device deployment problem into two simple sub-problems, overcoming the shortcomings of high complexity and strong coupling in existing technologies, and possessing the advantages of high cohesion and low coupling. Furthermore, the method of this application can adapt to scenarios where one or more wireless networks operate in parallel, overcoming the shortcomings of existing technologies that only consider one type of wireless network, and possessing the advantages of flexibility and wide applicability.

Description

Technical Field

[0001] This application relates to the field of edge computing technology, and in particular to a method for deploying edge devices. Background Technology

[0002] With the rapid development of computer and wireless network technologies, mobile terminals have become ubiquitous across various industries. However, due to factors such as architecture, size, and heat dissipation, while the processing power of mobile terminals has improved, it remains relatively weak compared to ordinary computers. It consumes considerable time when performing computationally intensive tasks, and may even be unable to execute certain heavyweight computing tasks. Therefore, relying solely on the resources of mobile terminals can no longer meet the increasingly complex needs of users. Task migration is one method to address the resource constraints of mobile terminals, migrating some computing tasks to external platforms to expand their available resources. Leveraging cloud computing, the concept of mobile cloud computing has emerged. This combines cloud computing, mobile computing, and wireless networks, aiming to provide abundant cloud resources to mobile terminals to enhance their capabilities. In mobile cloud computing, service requests must be transmitted through the core network to the cloud for processing, making it unsuitable for application scenarios involving latency sensitivity and massive data transmission.

[0003] Building upon mobile cloud computing, enhancing and extending it by adding an edge layer closer to mobile terminals creates a new architecture called MEC (Multi-access Edge Computing). Compared to traditional mobile cloud computing, MEC offers advantages such as low latency, low bandwidth consumption, and strong security, making it adaptable to a wider range of application scenarios. However, this new architecture also brings numerous challenges. MEC comprises three basic components: resource providers, service providers, and users. Resource providers are responsible for infrastructure construction and physical resource maintenance; service providers lease virtual resources from resource providers and are responsible for the development and operation of specific services; and users purchase and use cloud services provided by service providers. Resource providers offer hardware for MEC, forming its foundation, and the deployment of edge devices is a primary challenge in MEC construction. Only through proper edge device deployment can resource providers increase their revenue, prompting them to update and maintain hardware, thus promoting the healthy operation of MEC. The geographically dispersed and numerous nature of edge devices also presents challenges for their deployment. An improper deployment of edge devices will not only waste a lot of financial and material resources, but will also increase the complexity and coupling of other problems within the system. Summary of the Invention

[0004] This application provides an edge device deployment method to address the problems of high complexity and strong coupling in the deployment of edge devices in the prior art.

[0005] On one hand, embodiments of this application provide a method for deploying edge devices, including:

[0006] Obtain relevant information about the deployment area and edge devices;

[0007] Based on relevant information, establish an optimization model for edge device deployment that maximizes the benefits for resource providers;

[0008] The optimization model is solved using a two-layer deployment strategy algorithm to obtain the corresponding device deployment strategy and resource allocation strategy;

[0009] Deploy edge devices according to the device deployment strategy, and allocate resources according to the resource allocation strategy;

[0010] The optimization model is a two-layer model. The first layer is about how to place edge devices to different deployment areas, and the second layer is about how to handle the resource requirements of each area after the edge devices are placed. When using the two-layer deployment strategy algorithm to solve the optimization model, the first layer of the optimization model is transformed into solving a 1-dimensional "bin packing" problem, and the second layer of the optimization model is transformed into solving a (1+W)-dimensional "bin packing" problem. Then, the two-layer deployment strategy algorithm based on adaptive grouping genetic algorithm is used to solve the transformed optimization model.

[0011] The edge device deployment method in this application has the following advantages:

[0012] First, this application constructs the edge device deployment problem as a two-layer model, decoupling the complex edge device deployment problem into two simple sub-problems. The first layer of the two-layer model is only responsible for placing edge devices in different areas, and the second layer of the model only needs to handle the resource requirements of each area. The two layers interact through device deployment strategy and resource allocation strategy, which overcomes the shortcomings of the existing technology that treats the problem as a whole, resulting in high complexity and strong coupling. It has the advantages of high cohesion and low coupling.

[0013] Second, this application proposes a two-layer deployment strategy algorithm based on an adaptive grouping genetic algorithm to solve the optimization model. The optimization model solution is transformed into a two-layer "boxing" problem, and the adaptive grouping genetic algorithm is reused to solve the device deployment strategy in the first layer and the resource allocation strategy in the second layer. This overcomes the disadvantages of existing technologies that directly use different structure algorithms to solve the optimization model, such as the complexity of algorithm implementation. It has the advantages of low algorithm implementation cost and high code reusability.

[0014] Third, the method proposed in this application can be adapted to scenarios where one or more wireless networks operate in parallel, overcoming the shortcomings of existing technologies that only consider one type of wireless network, and has the advantages of being flexible in use and having a wide range of applications. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating an edge device deployment method provided in this application embodiment;

[0017] Figure 2 A two-layer edge device deployment model constructed for embodiments of this application;

[0018] Figure 3 This is a flowchart of the two-layer deployment strategy algorithm proposed in the embodiments of this application;

[0019] Figure 4 The benefits obtained by using the deployment method of this application;

[0020] Figure 5 It is the deployment rate using the deployment method of this application. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] Figure 1 A flowchart illustrating an edge device deployment method provided in this application embodiment. This application embodiment provides an edge device deployment method, including:

[0023] S100 obtains relevant information about the area to be deployed and edge devices.

[0024] For example, methods for obtaining relevant information about the deployment area and edge devices include:

[0025] S101, obtain the computing and wireless resource requirements of the area to be deployed. Store the relevant information of the area to be deployed in the graph G = (S, H), where the i-th vertex s of the graph is... i (s i ∈S) is the i-th region, and its computational resource requirements are calculated. Wireless resource demand Maximum tolerable delay T i Regional electricity fee q iThis application supports W types of wireless networks, and when calculating the demand for wireless resources, the demand for W types of wireless resources needs to be calculated. Right now After statistical analysis, the information from M regions is stored in a set.

[0026] S102, Obtain the distances between the regions to be deployed. After calculating the relevant information of the regions to be deployed themselves, calculate the distance information between the regions. Edge h = (i, k) of graph G represents region s. i and s k The connection between them, with weight d i,k It refers to distance. In this application, the Wireless Access Point (WAP) and edge devices are deployed together, and will be deployed in the center of the area. The formula t(d) is used. i,k )=αd i,k +β is used to calculate region s i and deployed in the region k Latency between mid-edge devices.

[0027] S103, obtain the edge device's computing resources and wireless capacity. For a specific edge device e j Statistical analysis of its computing resource capacity and the wireless resource capacity of the WAP deployed together. Similarly, it is necessary to calculate the capacity of Class W wireless resources. Right now After statistical analysis, information on N edge devices is stored in a set.

[0028] S104: Obtain power consumption and revenue information from resource providers. Statistical analysis of edge devices. j Full load power No-load power And calculate resource utilization rate γ j And calculate its power P according to the following formula. c (γ j ):

[0029]

[0030] Statistics and edge devices e j The power information of the WAP deployed together is used to calculate the full-load power of its W-type wireless resources. No-load power and wireless resource utilization And calculate the power it provides for type w wireless resources according to the following formula:

[0031]

[0032] After collecting the above information, the power of the resource provider is calculated using the following formula. This application supports resource providers leasing wireless resources from network providers. In this case, there is no need to worry about the power consumption of wireless resources, and the parameter λ can be set to 0. If the resource provider also operates the WAP, then λ is set to 1. In this embodiment, λ is set to 1.

[0033]

[0034] function P(γ) j ,ρ j ) used for computing edge devices e j Energy consumption, parameter γ j and ρ j For edge devices e j The utilization rates of computing resources and wireless resources, and the revenue from computing resources of rental units. Revenue from wireless resources of type w Resource providers derive their revenue from leasing out computing and wireless resources, and their revenue is calculated using the following formula:

[0035]

[0036] S110, based on relevant information, establishes an edge device deployment optimization model that maximizes the benefits for resource providers.

[0037] For example, the optimization model is a two-layer model, where the first layer is how to place edge devices to different deployment areas, and the second layer is how to handle the resource requirements of each area after the edge devices are placed.

[0038] Specifically, S110 includes:

[0039] S111, see reference Figure 2 This paper decomposes the edge device deployment problem into a two-layer model, further decomposing the edge device deployment optimization problem that maximizes the resource provider's profit into a two-layer model. The first layer of the model determines the deployment location of the edge devices and passes the deployment strategy to the second layer. The second layer processes the resource requirements of each region based on the input deployment strategy, considering computational resource capacity, wireless resource capacity, and latency constraints when allocating resources to each region. The second layer outputs the optimal resource allocation strategy and feeds it back to the first layer. The deployment strategy and the resource allocation strategy together determine the resource provider's profit; that is, the resource provider's profit is optimal when both the deployment strategy of the first layer and the resource allocation strategy of the second layer are optimal.

[0040] S112, Define model parameters. Define matrix X as the device deployment strategy, whose elements x i,j =1 represents the edge device e j Deploy to region si x i,j =0 represents the edge device e j Not deployed to region s i Define matrix Y as the resource allocation strategy, whose elements y i,j =1 represents the edge device e j For region s i Allocate resources, y i,j =0 represents the edge device e j Not for region s i Allocate resources. Define a vector Z as the on / off state of the edge device, whose elements z j =1 indicates that the edge device e is enabled. j , z j =0 means turn off edge device e j .

[0041] S113, Construct a constrained optimization model for maximizing the revenue from edge device deployment. This problem is constructed as a constrained optimization model where revenue equals the income from renting out resources minus the electricity cost of running the edge devices. The objective function of the model is as follows:

[0042]

[0043] The model is constrained by the following seven conditions: 1) This constraint means that only one edge device can be deployed in a region; 2) This constraint means that an edge device can only be deployed in one region; 3) This constraint ensures that resources in a region can only be allocated by one edge device; 4) This constraint ensures that the computing resources allocated to an edge device do not exceed its computing resource capacity; 5) This constraint ensures that the type w wireless resources allocated to an edge device do not exceed the corresponding wireless resource capacity; 6) This constraint ensures that the delay in acquiring resources for a region does not exceed its maximum tolerable delay; 7) This constraint makes x i,j y i,j z j The value can only be 0 or 1.

[0044] S120 uses a two-layer deployment strategy algorithm to solve the optimization model and obtain the corresponding device deployment strategy and resource allocation strategy.

[0045] For example, when using the two-layer deployment strategy algorithm to solve the optimization model, the first layer of the optimization model is transformed into solving a 1-dimensional "boxing" problem, and the second layer of the optimization model is transformed into solving a (1+W)-dimensional "boxing" problem. Then, the two-layer deployment strategy algorithm based on the adaptive grouping genetic algorithm is used to solve the transformed optimization model.

[0046] The two-layer deployment strategy algorithm in this application consists of two layers. The first layer solves the device deployment strategy and passes it to the second layer, which then solves the resource allocation strategy based on the device deployment strategy. The first layer of the two-layer deployment strategy algorithm can call the second layer to evaluate its chromosomes. Simultaneously, the first and second layers of the algorithm reuse grouped chromosome encoding and perform selection, crossover, mutation, and inversion operations on the grouped chromosome encoding. The solution process for the optimization model can be simply summarized as follows: When solving the first layer of the transformed optimization model, an initial population is generated. Two chromosomes are selected from the initial population, and crossover, mutation, and inversion operations are performed on the selected chromosomes sequentially. The processed chromosomes are evaluated, and the evaluated chromosomes are retained to form the final population of the first layer. The optimal chromosome is obtained from the final population of the first layer. When solving the second layer of the transformed optimization model, an initial population is generated. Two chromosomes are selected from the initial population, and crossover, mutation, and inversion operations are performed on the selected chromosomes sequentially. The processed chromosomes are evaluated, and the evaluated chromosomes are retained to form the final population of the second layer. The optimal chromosome is obtained from the final population of the second layer.

[0047] Specifically, the solution process for the optimization model in this application includes:

[0048] S121 transforms the optimization model solution into a two-layer "boxing" problem solution. Specifically, the first layer of the two-layer model transforms the edge device deployment optimization model into solving a one-dimensional "boxing" problem, that is, transforming the region into a "box" and the edge device into an "item", and restricting a "box" to only contain one "item". The second layer of the two-layer model transforms the edge device deployment optimization model into solving a (1+W)-dimensional "boxing" problem, that is, transforming the edge device into a "box" and the region resource requirements into "items", with computing resources and W types of wireless resources constituting the (1+W) dimensions of the problem.

[0049] S122, see reference Figure 3 The optimization model is solved using a two-layer deployment strategy algorithm based on an adaptive grouping genetic algorithm. This step is the execution process of the first layer of the algorithm, which calls the second layer of the algorithm to evaluate its chromosomes. The inputs include: the information of the region to be deployed, G = (S, H), and the information of the edge devices, E, and the population size K set for the first layer of the algorithm. 1st The number of populations K in the second layer of the algorithm2nd The maximum number of iterations L in the first layer of the algorithm 1st The maximum number of iterations L in the second layer of the algorithm 2nd .

[0050] S122 specifically includes the following sub-steps:

[0051] S1221, use the First Modified Best Fit Algorithm (FMBFA) to generate the first-level initial population Pop. 1st (0).

[0052] The population of the first layer of the algorithm is composed of K. 1st Composed of chromosomes, using sets To store chromosomes. When placing an edge device, FMBFA selects an area that has not yet been deployed and has the lowest electricity costs.

[0053] S1222, jump to S124 to evaluate chromosome U∈Pop in the initial population. 1st (0).

[0054] S1223, Initialize variable l 1st =0,k 1st =0.

[0055] S1224, Initialize the new population NewPop 1st (l 1st ).

[0056] S1225, from population Pop 1st (l 1st Choose two chromosomes from ).

[0057] The selection method used is a round-robin selection method, which generates a random number ε∈[0,1), and the first one that satisfies the condition is selected. chromosome U i It will be selected. Represents chromosome U i The probability of being selected is calculated using the following formula:

[0058]

[0059] Represents chromosome U j (U i The fitness function value is calculated using the following formula:

[0060] f 1st (U) = max f(U,V)

[0061] Where f(U,V) is calculated using the following formula:

[0062]

[0063] S1226, the crossover probability pr is calculated using the following formula. c This determines whether to perform crossover on the two chromosomes selected in the previous step (called the parent chromosomes). Among them, pr c1 and pr c2 It is a constant, f l It is the larger of the two parent chromosome fitness function values, f max and f avg These represent the maximum fitness function value and the average fitness function value in the population, respectively.

[0064]

[0065] Generate a random number ε∈[0,1), if ε≤pr c The following steps involve a crossover operation: First, two distinct gene points are randomly generated, and a gene segment is formed from one or more consecutive genes between these two gene points. Second, the gene segment is inserted into the other chromosome. Third, to avoid duplicate edge devices on the same chromosome, genes containing the same edge device are deleted during deletion and insertion. Finally, after deleting these genes, edge devices not in the inserted gene segment are left. These remaining edge devices are processed using FMBFA and redeployed to the appropriate regions. This process utilizes the first-layer improved best-fit algorithm to handle the remaining resource requests after the first-layer grouping crossover and grouping mutation operations of the two-layer deployment strategy algorithm.

[0066] S1227, the mutation probability pr is calculated using the following formula. m This determines whether to perform mutation operations on the two chromosomes following the previous step. Among them, pr m1 and pr m2 is a constant, and f is the fitness function value of the parent chromosome.

[0067]

[0068] First, generate a random number ε∈[0,1). If ε≤pr m If the gene is selected, a group of genes is chosen, and these genes are then deleted from the chromosome. Then, FMBFA is used to process the edge devices within these genes, redeploying them to the appropriate regions.

[0069] S1228, reverse the operation on the two chromosomes after the previous step, randomly select two genes and change their order in the chromosomes.

[0070] S1229, jump to S124 to evaluate the two chromosomes after the previous step.

[0071] S1230, Add the two chromosomes from the previous step to the population NewPop. 1st (l 1st ).

[0072] S1231, k 1st The value increases by 2, that is, k 1st ←k 1st +2.

[0073] S1232, Judgment condition k 1st =K 1st Validity: If not, proceed to S1225; if valid, proceed to the next step.

[0074] S1233, the newly generated population is copied to the next generation, i.e., Pop. 1st (l 1st +1)←NewPop 1st (l 1st ).

[0075] S1234, l 1st Add 1, that is, l 1st ←l 1st +1.

[0076] S1235, Judgment condition l 1st =L 1st If not true, proceed to S1224; if true, proceed to the next step.

[0077] S1236, from the final population Pop 1st (L 1st Obtain the optimal chromosome U in ) optimal .

[0078] S124, this step is the execution process of the second layer of the algorithm. It will be called by the first layer of the algorithm to evaluate its chromosome. The inputs include: the information of the region to be deployed, G = (S, H) and the information of the edge device, E, and the population size K of the second layer of the algorithm. 2nd Maximum number of iterations L 2nd And chromosome U.

[0079] S124 specifically includes the following sub-steps:

[0080] S1241 uses the Second Modified Best Fit Algorithm (SMBFA) to generate the second-layer initial population Pop. 2nd (0).

[0081] The population of the second layer of the algorithm is composed of K. 2nd Composed of chromosomes, using sets To store chromosomes. (Condition 1) 2) and 3) To constrain chromosome V. In constraint 3), edge device e is obtained from the first-level chromosome U of the algorithm using g(U,j) (1≤g(U,j)≤M). j The deployment location.

[0082] In region s i When allocating resources, firstly, SMBFA selects a group of resources that are already enabled and in use for region s. i Edge devices that do not violate the above three constraints after resource allocation use a set. Storage. Secondly, SMBFA determines the set E. s Is it empty?: If empty, open a new edge device and set it for region s. i Allocate resources; if not empty, calculate E one by one using the following formula. s Mid-edge devices e j Δf j And select the edge device with the largest Δf as region s i Allocate resources.

[0083]

[0084] Where, q g(U,j) This represents the electricity cost of the g(U,j)-th region. The function g(U,j) is used to obtain the deployment edge device e from chromosome U. j The region number.

[0085] S1242, Given U, use the following formula to evaluate chromosome V∈Pop in the population. 2nd (0):

[0086] f 2nd (V)=f(V|U)

[0087] Where f(V|U) represents the value of f(U,V) given U, and is calculated using the following formula:

[0088]

[0089] S1243, Initialize variable l 2nd =0,k 2nd =0.

[0090] S1244, Initialize a new population NewPop 2nd (l 2nd ).

[0091] S1245, from population Pop 2nd (l 2nd Choose two chromosomes from ).

[0092] The selection method used is the same as in the first layer of the algorithm, namely, the round-robin selection method. A random number ε∈[0,1) is generated, and the first one that satisfies the condition... Chromosome V i It will be selected. Represents chromosome V i The probability of being selected is calculated using the following formula: Represents chromosome V j (V i The fitness function value of ).

[0093]

[0094] S1246, the crossover probability pr is calculated using the following formula. c This determines whether to perform crossover on the two chromosomes selected in the previous step (called the parent chromosomes). Among them, pr c1 and pr c2 It is a constant, f l It is the larger of the two parent chromosome fitness function values, f max and f avg These represent the maximum fitness function value and the average fitness function value in the population, respectively.

[0095]

[0096] Generate a random number ε∈[0,1), if ε≤pr c The crossover operation is then performed as follows: First, two distinct gene points are randomly generated, and a gene segment is formed from one or more contiguous genes between these two gene points. Second, the gene segment is inserted into the other chromosome. Third, to avoid duplicate resource requests on the same chromosome, genes with the same resource requests as those being deleted and inserted are deleted. Finally, after deleting these genes, resource requests not in the inserted gene segment are left over. These remaining resource requests are processed using SMBFA, and resource centers are reallocated to them. This is equivalent to using the second-layer improved best-fit algorithm to place the remaining edge devices after the second layer of the two-layer deployment strategy algorithm performs grouping crossover and grouping mutation operations.

[0097] S1247, the mutation probability pr is calculated using the following formula. m This determines whether to perform mutation operations on the two chromosomes following the previous step. Among them, pr m1 and pr m2is a constant, and f is the fitness function value of the parent chromosome.

[0098]

[0099] First, generate a random number ε∈[0,1). If ε≤pr m If the gene is selected, a group of genes is selected and then these genes are deleted from the chromosome. Then, SMBFA is used to process the resource requests of these genes and reallocate resource centers to them.

[0100] S1248, reverse the operation on the two chromosomes after the previous step, randomly select two genes and change their order in the chromosomes.

[0101] S1249, given U, evaluate the two chromosomes after the previous step using the following formula:

[0102] f 2nd (V)=f(V|U)

[0103] Where f(V|U) represents the value of f(U,V) given U, and is calculated using the following formula:

[0104]

[0105] S1250, Add the two chromosomes from the previous step to the population NewPop. 2nd (l 2nd ).

[0106] S1251,k 2nd The value increases by 2, that is, k 2nd ←k 2nd +2.

[0107] S1252, Judgment condition k 2nd =K 2nd Valid or invalid: If invalid, proceed to S1245; if valid, proceed to the next step.

[0108] S1253, the newly generated population is copied into the next generation, i.e., Pop. 2nd (l 2nd +1)←NewPop 2nd (l 2nd ).

[0109] S1254, l 2nd Add 1, that is, l 2nd ←l 2nd +1.

[0110] S1255, Judgment condition l 2nd =L 2ndIf not true, proceed to S1244; if true, proceed to the next step.

[0111] S1256, Return from the final population Pop 2nd (L 2nd The optimal chromosome V obtained from ) optimal And the corresponding optimal profit max f(U,V).

[0112] S130 deploys edge devices according to the device deployment strategy and allocates resources according to the resource allocation strategy.

[0113] For example, S130 specifically includes the following sub-steps:

[0114] S131, the optimal chromosome U obtained in S120 is processed according to the following rules. optimal and V optimal This translates into device deployment strategy X, resource allocation strategy Y, and edge device switching status Z.

[0115] Transformation rule: 1) If u i ={e j}, then x i,j =1, otherwise x i,j =0; 2) If s i ∈v j , then y i,j =1, otherwise y i,j =0; 3) If Then z j =1, otherwise z j =0.

[0116] S132, place edge devices according to the device deployment strategy X from the previous step, and allocate resources to each area according to the resource allocation strategy Y.

[0117] This application compares itself with five existing technologies, specifically comparing the two indicators of revenue and deployment rate, as follows: Figure 4 and Figure 5As shown in the figure. It should be noted that these comparative techniques also use the double-layer deployment strategy algorithm architecture proposed in this application, the difference being the internal algorithm used. The English abbreviations used in the figure are explained as follows: Double-Layer Adaptive Group Genetic Algorithm (DLAGGA), Double-Layer First Fit (DLFF), Double-Layer NextFit (DLNF), Double-Layer Modified Best Fit (DLMBF), Double-Layer Round Robin (DLRR), and Double-Layer Standard Genetic Algorithm (DLSGA). Figure 4 and Figure 5 This is a "box diagram" evaluating these technologies across ten scenarios, from... Figure 4 As can be seen, the returns of the DLAGGA proposed in this application are greater than those of other algorithms. The average relative differences in returns between DLFFA, DLNFA, DLMBFA, DLRR, DLHRA, DLSGA, and the DLAGGA proposed in this application are 19.78%, 32.01%, 15.81%, 36.7%, 18.83%, and 12.74%, respectively, indicating that the DLAGGA proposed in this application can achieve higher returns. Figure 5 As can be seen, the deployment rate of the DLAGGA proposed in this application is lower than that of other algorithms, indicating that the DLAGGA proposed in this application deploys fewer edge devices while meeting the requirements, and has higher deployment efficiency.

[0118] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0119] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for deploying edge devices, characterized in that, include: Obtain relevant information about the deployment area and edge devices; Based on the aforementioned relevant information, establish an edge device deployment optimization model that maximizes the benefits for resource providers; The optimization model is solved using a two-layer deployment strategy algorithm to obtain the corresponding device deployment strategy and resource allocation strategy. Deploy edge devices according to the device deployment strategy, and allocate resources according to the resource allocation strategy; The optimization model is a two-layer model. The first layer addresses how to place edge devices to different deployment areas, and the second layer addresses how to handle the resource requirements of each area after the edge devices are placed. When solving the optimization model using the two-layer deployment strategy algorithm, the first layer of the optimization model is transformed into solving a 1D "bin packing" problem, and the second layer is transformed into solving... The "binding" problem is solved, and then the transformed optimization model is solved using a two-layer deployment strategy algorithm based on an adaptive grouping genetic algorithm. The edge device deployment optimization model includes an objective function and seven constraints: The objective function represents revenue, and its value equals the income from renting out resources minus the electricity cost of operating the edge devices: Among them, matrix This indicates the edge device deployment strategy, its elements Represents edge devices Deploy to the region , Represents edge devices Not deployed to the region ;matrix This represents the resource allocation strategy, its elements Represents edge devices For the region Allocate resources, Represents edge devices Not for the region Resource allocation; vector The elements represent the on / off state of the edge device. Represents enabling edge devices , Represents turning off edge devices ;gather express One edge device to be deployed Indicates edge device The computing resource capacity, Indicates edge device of Wireless resource capacity; set express Each region Indicates the region The computing resource requirements, Indicates the area of The demand for wireless resources, Indicates the region Maximum tolerable latency, Indicates the region Electricity costs; function For computing edge devices Energy consumption, parameters and Edge devices The utilization rate of computing resources and wireless resources; function For computing edge devices Income from renting out resources; Indicates the number of types of wireless networks; The seven constraints are as follows: 1) This constraint allows only one edge device to be deployed in a region; 2) This constraint means that an edge device can only be deployed in one region; 3) This constraint ensures that resources in a region can only be allocated by one edge device; 4) This constraint ensures that the computing resources allocated to an edge device do not exceed its computing resource capacity; 5) This constraint limits the allocation of resources to an edge device. The type of wireless resource does not exceed the corresponding wireless resource capacity; 6) This constraint ensures that the delay in acquiring resources for a region does not exceed its maximum tolerable delay; 7) This constraint makes , , The value can only be 0 or 1.

2. The edge device deployment method according to claim 1, characterized in that, Methods for obtaining information about the deployment area and edge devices include: Obtain the computing and wireless resource requirements of the area to be deployed; Obtain the distance between the areas to be deployed; Acquire computing resources and wireless capacity for edge devices; Obtain power consumption and revenue information from resource suppliers.

3. The edge device deployment method according to claim 1, characterized in that, The first layer of the two-layer model is used to determine the deployment location of edge devices and pass the device deployment strategy to the second layer. The second layer of the two-layer model processes the resource requirements of each region according to the device deployment strategy, and considers computational resource capacity, wireless resource capacity and latency constraints when allocating resources to each region. The second layer of the two-layer model outputs the optimal resource allocation strategy and feeds it back to the first layer. The equipment deployment strategy and the resource allocation strategy together determine the revenue of the resource supplier. That is, when the equipment deployment strategy of the first layer of the model and the resource allocation strategy of the second layer of the model are both optimal, the revenue of the resource supplier is optimal.

4. The edge device deployment method according to claim 1, characterized in that, The first layer of the two-layer model is used to transform the edge device deployment optimization model into solving a one-dimensional "boxing" problem: transforming the region into a "box", the edge device into an "item", and restricting a "box" to only contain one "item"; The second layer of the two-layer model is used to transform the edge device deployment optimization model into solving a... The "boxing" problem involves transforming edge devices into "boxes" and regional resource requirements into "items," which are then managed by computing resources and... Wireless resources constitute the problem Each dimension.

5. The edge device deployment method according to claim 1, characterized in that, The two-layer deployment strategy algorithm consists of two layers. The first layer is used to solve the device deployment strategy and pass the device deployment strategy to the second layer. The second layer solves the resource allocation strategy according to the device deployment strategy. The first layer of the two-layer deployment strategy algorithm can call the second layer to evaluate its chromosomes. At the same time, the first and second layers of the two-layer deployment strategy algorithm reuse grouped chromosome encoding and grouped genetic operations such as selection, crossover, mutation and reverse.

6. The edge device deployment method according to claim 1, characterized in that, The first-layer improved best-fit algorithm and the second-layer improved best-fit algorithm are used to generate the first-layer initial population and the second-layer initial population, respectively. Then, the first-layer improved best-fit algorithm is used to place the remaining edge devices after the first layer of the two-layer deployment strategy algorithm has undergone grouping crossover and grouping mutation operations; at the same time, the second-layer improved best-fit algorithm is used to place the remaining resource requests after the second layer of the two-layer deployment strategy algorithm has undergone grouping crossover and grouping mutation operations.

7. The edge device deployment method according to claim 6, characterized in that, When placing an edge device using the first-layer improved best-fit algorithm, select an area that has not yet been deployed and has the lowest electricity cost; When using the second-layer improved best-fit algorithm, first select a set of regions that are already enabled and are in the region. Edge devices that do not violate computing resource capacity constraints, wireless resource capacity constraints, and latency constraints after resource allocation use a set. This indicates that, based on this, the second-layer improved best-fit algorithm determines the set. Is it empty? If empty, open a new edge device and set it for the region. Allocate resources; If not empty, calculate one by one. Mid-edge devices Value, and select The edge device with the largest value is the region. Resource allocation, including edge devices of The value is denoted as Use the following formula to calculate: in, Indicates the region The computing resource requirements, Indicates renting edge equipment Revenue per unit of computing resources; Indicates the region of The demand for wireless resources, Indicates renting edge equipment unit Revenue from wireless resources; Indicates the first Electricity cost for each region, function Used from chromosomes Obtain deployment edge devices The region number, Indicates edge device The computing resource capacity, Indicates edge device Full load power; Indicates edge device The no-load power; Indicates the number of types of wireless networks. Indicates the region of The demand for wireless resources, Indicates edge device of Wireless resource capacity, Indicates edge device of Full-load power of wireless resources; Indicates edge device of The idle power of wireless resources.

Images