Intelligent service deployment method, system and terminal in edge computing environment

Abstract

The present invention belongs to the field of edge computing technology, and discloses a method, system and terminal for intelligent service deployment in an edge computing environment, establishes a mathematical model of the edge computing environment, determines an evaluation function that comprehensively considers migration costs and service delays; uses a resource allocation matrix to model the service deployment scheme in the edge environment, and designs an optimization scheme for edge computing service deployment based on a differential evolution algorithm. The present invention comprehensively considers various types of information on edge environment service deployment, including delay, cost, routing, available resources, etc. On this basis, a scheme for joint optimization of delay and cost is made, thereby improving edge system performance, reducing deployment costs, and reducing the time cost and workload of edge system design, solving the problem of too long request delay in centralized service deployment schemes, the problem of too high deployment costs in service deployment schemes based on service requests, and the problem of dynamically deploying services in service deployment schemes based on user trajectory prediction.

Description

Technical Field

[0001] The present invention belongs to the field of edge computing technology, and in particular, relates to a method, system and terminal for deploying intelligent services in an edge computing environment. Background Art

[0002] In recent years, the field of edge computing has received widespread attention as a hot area of ​​service computing. Compared with traditional centralized cloud computing, edge computing deploys more workloads closer to users. Due to the rapid development of services and scenarios such as 5G and IoT, and the sharp increase in the number of smart terminal devices, there are more and more demands for the sinking of edge computing services. If all services are placed in the central cloud, it is difficult to meet the growth of large-scale edge smart devices. Edge computing is currently being used on a large scale in various industries, such as automobiles, transportation, energy, etc.

[0003] The performance of an edge computing environment is generally determined by the average latency of service request processing. First, the request latency of an edge environment can be defined as: given the number of existing edge servers, deployed services, and geographic locations, the newly generated requests for each service are sent to an edge server that deploys the service for processing through a preset routing algorithm. The time spent on the entire process is the processing latency of the request. When evaluating an edge computing environment, in addition to the average latency of requests, the cost of service deployment also needs to be considered. Because the resources of an edge server are limited, and when consuming resources to deploy a service instance and run it, it also costs a certain amount of money. Therefore, it is very important to choose a deployment solution that is relatively balanced between service latency and deployment cost when deploying services.

[0004] There are many technical solutions for service deployment, such as cloud-based service deployment solutions, decentralized service deployment solutions, and service deployment solutions based on user trajectory prediction, etc. Among them, the service deployment solution based on user trajectory prediction has good applicability and performance, but it is not suitable for balancing service latency and deployment costs, and has certain risks.

[0005] The current algorithms for food delivery mainly include the following:

[0006] (1) Service deployment solution for edge-cloud collaboration

[0007] Combining the cloud computing environment with edge computing, a resource deployment algorithm based on task prediction is designed. Tasks are predicted through two-dimensional time series in the cloud service center, and the resource deployment required for edge server task operation is optimized by combining classification aggregation, delay threshold judgment, etc.

[0008] (2) Decentralized service deployment solution

[0009] When an edge server receives a request for which it does not have a corresponding type of service, if it still has available resources, it will deploy a new service instance of the corresponding type. If the resources reach the threshold, it will forward the request to other edge servers for processing. Or when the first computing cost of the target service in the first node server is not greater than the preset threshold, it will respond to the service request; if the first computing cost is greater than the preset threshold, the second node server will be determined; and the service request will be sent to the second node server, so that the second node server will respond to the service request according to the pre-deployed computing unit; another method is user-centric, and applies joint optimization algorithms such as the Lyapunov optimization method and the Benders decomposition framework to solve the service deployment problem in multi-user offloading scenarios.

[0010] (3) Service deployment solution based on user trajectory prediction:

[0011] Utilize the user's historical trajectory information and use the deep learning network to predict the user's future trajectory. By combining the user's future trajectory information, determine the possible access server of the user in the future and deploy the corresponding service instance on it. For example:

[0012] The user's future trajectory is predicted through the deep learning LSTM network, and a deployment plan is formed by combining the user's future trajectory information and the service combination to be requested.

[0013] (4) Dynamic network service deployment solution:

[0014] This type of deployment method designs dynamic service deployment solutions for dynamic edge computing environments. For example, there is another method that calculates the response delay of each current service request, builds a deployment model based on the response delay and the performance parameters of the edge computing node, and outputs the deployment data of the service instance in the edge computing node based on the deployment model; there is another method that takes the service quality assurance of user mobile cross-region as the goal, performs service scheduling decision calculations on user service requests, and generates optimization solutions for service deployment and migration; in addition, there is also a model training service deployment method for edge intelligence, which combines edge computing technology with deep learning and other technologies, and performs edge distributed model training based on the geographical distribution characteristics of edge data.

[0015] (5) Service deployment solution based on reinforcement learning:

[0016] Using reinforcement learning technology, we build a cost model for service latency and other factors, and use the reinforcement learning framework to train the deployment plan. For example:

[0017] Build a system model, determine the user energy consumption and delay calculation model, use deep reinforcement learning theory to establish an intelligent agent training model, and train the model to obtain a resource deployment strategy; another method uses Lyapunov optimization to decompose the long-term system utility maximization problem into an online Lyapunov drift plus penalty function minimization problem, introduces the service deployment probability distribution, and uses the Markov approximation model to dynamically deploy services; another method uses Q reinforcement learning method for deployment, conducts multiple trainings to obtain a long-term cumulative feedback value, and uses or to calculate this cumulative feedback value and obtain a deployment plan; some researchers have proposed a general framework for optimizing the deployment of public transportation edge applications based on deep reinforcement learning technology, which can learn the optimal deployment method from historical experience.

[0018] (6) Game-based service deployment method

[0019] First, queuing theory is used to represent the request rate and service rate of service requests configured on edge nodes. Secondly, a two-stage Stackelberg game is used to model the interaction between service providers and users' service requests. Finally, the Nash equilibrium is solved through the pruning method to obtain the deployment plan. Other researchers have proposed a service function chain deployment algorithm based on service offloading and online game to minimize latency for service deployment of latency-sensitive services. This algorithm combines effective online game and service offloading decisions to deploy service function chains in cloud fog computing networks to reduce the end-to-end latency of service function chains.

[0020] The existing service deployment methods have the following main shortcomings:

[0021] (1) Edge-cloud collaborative service deployment solution: This solution still has some defects in the cloud computing environment because it does not get rid of the use of cloud service centers. For example, due to the uncertainty of the distance between users and central servers and the large number of requests accumulated in the center waiting to be resolved, the service processing delay may be too high.

[0022] (2) Decentralized service deployment scheme: The service deployment of a server is determined by the service requests it receives, which allows the user's request to be processed by the edge server that is closer to the user, effectively reducing the service latency. However, when the user is constantly moving, the user's nearest server is constantly changing, and the continuous requests they send may cause each server to deploy a large number of service instances, which not only costs too much deployment cost, but also causes the previously deployed service instances to be idle when the user moves out, consuming computing resources.

[0023] (3) Service deployment solution based on user trajectory prediction: This method has the following two problems. On the one hand, this deployment method will dynamically add or delete service instances on each server, which will increase the deployment cost of the service accordingly. On the other hand, the accuracy of trajectory prediction is difficult to guarantee. When the prediction is wrong, the corresponding service instance will be missing on the server near the user, resulting in increased request latency.

[0024] Through the above analysis, the problems and defects of the prior art are as follows:

[0025] (1) Existing technical solutions for multiple service deployments are not suitable for balancing service latency and deployment costs, and they also have certain risks.

[0026] (2) The edge-cloud collaborative service deployment solution still has some defects in the cloud computing environment because it has not gotten rid of the use of cloud service centers. The decentralized service deployment solution not only costs too much deployment cost, but also causes previously deployed service instances to be idle when users move out, consuming computing resources.

[0027] (3) The service deployment scheme based on user trajectory prediction will increase the deployment cost of the service, and the accuracy of trajectory prediction is difficult to guarantee; when the prediction is wrong, the corresponding service instance will be missing on the server near the user, resulting in an increase in request latency. Summary of the invention

[0028] In response to the problems existing in the prior art, the present invention provides a method, system and terminal for intelligent service deployment in an edge computing environment, and more particularly, relates to a method, system, medium, device and terminal for intelligent service deployment based on differential evolution in an edge computing environment.

[0029] The present invention is implemented as follows: a method for deploying intelligent services in an edge computing environment, the method for deploying intelligent services in an edge computing environment comprising:

[0030] Establish a mathematical model of the edge computing environment and determine the evaluation function that comprehensively considers migration cost and service delay; use the resource allocation matrix to model the service deployment plan in the edge environment, and design an optimization plan for edge computing service deployment based on the differential evolution algorithm.

[0031] Furthermore, the method for deploying intelligent services in an edge computing environment includes the following steps:

[0032] Step 1: Obtain relevant information of each part in the edge computing environment: read the edge servers, access points, network topology and service cost information in the environment to obtain the parameters of the evaluation function and system model;

[0033] Step 2: Calculate the transmission path of each access point request: The transmission path of each service at each access point is obtained by a preset path-finding algorithm, so as to facilitate the subsequent calculation of the average delay of various service deployment schemes;

[0034] Step 3: Establishing the system mathematical model and objective function: Based on the information obtained in steps 1 and 2, a mathematical model for calculating the latency and cost of the deployment scheme is obtained, so that different service deployment schemes can be evaluated and compared;

[0035] Step 4: Differential evolution calculates the optimal deployment plan: Based on the differential evolution algorithm, the optimal solution for service deployment is obtained. The resource allocation matrix in this solution is a continuous variable and is suitable for solving using the differential evolution algorithm.

[0036] Further, the step 1 of obtaining relevant information of each part in the edge computing environment includes:

[0037] Obtain the available resources of each edge server and the deployment cost of each service unit resource; obtain the service type of each service, the request size of each service, and the request processing speed of each service when allocating unit resources;

[0038] Obtain the edge environment connection topology diagram of the edge computing network, the average transmission rate of each link, and the preset request transmission pathfinding algorithm; obtain the average number of requests per unit time for each type of service at the access point and the wireless transmission rate of the access point; use the resources allocated to each service by each server as the resource allocation vector of the server, and concatenate the resource allocation vectors of all edge servers by column to form a matrix Y.

[0039] Further, the transmission path calculation requested by each access point in step 2 includes:

[0040] According to the network topology map and path-finding algorithm obtained in step 1, each type of service request of each access point can find a fixed request transmission path set P; according to the path set P and the average transmission rate of each link included in the path set, the average transmission time of the service on the access point is obtained.

[0041] Furthermore, the system mathematical model and objective function establishment in step 3 include:

[0042] Establish the objective function f(Y):

[0043] f(Y)=a1×C(Y)+a2×T(Y);

[0044]

[0045] C(Y)=∑ s∈S ∑ c∈V Y s，v ×β s;

[0046] Where T(Y) is the total time consumption of each service request received by each access point per unit time, C(Y) is the capital cost of the Y matrix deployment solution, and λ s，a is the number of requests arriving per unit time for service s at access point a, d s is the data size of a single request to service s, v a is the receiving speed of access point a, μ s is the request processing speed of the service instance s per unit resource, v p is the average transmission speed of link p, β s is the deployment cost per unit resource of service s.

[0047] Set constraints on the objective function:

[0048] Constraint 1 is ∑ s∈S Y s，v ≤r v ,The sum of resources allocated by edge server v to each service cannot exceed the total available resources of v;

[0049] Constraint 2 is ∑ v∈V Y s，v ≥0, each service has at least one instance of allocated resources.

[0050] Furthermore, the optimal deployment solution of differential evolution calculation in step 4 includes:

[0051] After obtaining the mathematical model of the system, the service deployment matrix is ​​used as the solution to the problem, and the feasible solution is the one that meets the constraints in step 3. A population of feasible solutions of the service deployment matrix Y of a reasonable size is randomly generated, and the f(Y) value is obtained for each feasible solution through the established objective function. As the fitness of the solution, it participates in the subsequent differential evolution process.

[0052] The differential evolution calculation optimal deployment scheme also includes: connecting the columns of the feasible solution matrix into a vector to form a real-number coded chromosome; obtaining the fitness through a mathematical model and system data. Use the differential evolution algorithm to obtain the optimal feasible solution to the problem, including:

[0053] 1) Set the control parameters of the differential evolution algorithm, the maximum evolutionary generation T, the crossover probability, the mutation probability, and the initial feasible solution population P obtained according to the constraints; 2) Calculate the fitness of each feasible solution in the population P; 3) Select individuals according to the fitness size according to the roulette algorithm to be inherited to the next generation of the population; 4) Perform crossover and mutation operations to generate new individuals, and inherit the new individuals to the next generation of the population; 5) Repeat steps 1) to 4) until the genetic generation is T, and select the service deployment solution with the best fitness as the optimal solution.

[0054] Another object of the present invention is to provide an intelligent service deployment system in an edge computing environment applying the intelligent service deployment method in an edge computing environment, the intelligent service deployment system in an edge computing environment comprising:

[0055] The information acquisition module is used to obtain relevant information of various parts in the edge computing environment, including reading the edge servers, access points, network topology and service cost information in the environment;

[0056] A transmission path calculation module, used to obtain the transmission path of each service at each access point by a preset path finding algorithm;

[0057] A system mathematical model building module, used to obtain a mathematical model for calculating the delay and cost of the deployment solution based on the information obtained by the information acquisition module and the transmission path calculation module requested by each access point;

[0058] The optimal deployment solution acquisition module is used to obtain the optimal solution for service deployment based on the differential evolution algorithm.

[0059] Another object of the present invention is to provide a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for deploying intelligent services in an edge computing environment.

[0060] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to perform the steps of the method for deploying intelligent services in an edge computing environment.

[0061] Another object of the present invention is to provide an information data processing terminal, which is used to implement the intelligent service deployment system in the edge computing environment.

[0062] In combination with the above technical solutions and the technical problems solved, please analyze the advantages and positive effects of the technical solutions to be protected by the present invention from the following aspects:

[0063] First, in view of the technical problems existing in the above-mentioned prior art and the difficulty of solving the problems, the technical solutions to be protected by the present invention and the results and data during the research and development process are closely combined to analyze in detail and deeply how the technical solutions of the present invention solve the technical problems, and some creative technical effects brought about after solving the problems. The specific description is as follows:

[0064] The present invention establishes a mathematical model of edge computing environment and provides an evaluation function that comprehensively considers migration cost and service delay. The present invention uses resource allocation matrix to model the service deployment scheme in edge environment. The present invention designs an optimization scheme for edge computing service deployment based on differential evolution algorithm.

[0065] The present invention mainly solves the following three problems: 1) It solves the problem of long request delay in centralized service deployment scheme; 2) It solves the problem of high deployment cost in service deployment scheme based on service request; 3) It solves the problem of dynamic deployment of services in service deployment scheme based on user trajectory prediction.

[0066] Second, considering the technical solution as a whole or from the perspective of the product, the technical effects and advantages of the technical solution to be protected by the present invention are described in detail as follows:

[0067] The present invention comprehensively considers various information of edge environment service deployment, including latency, cost, routing, available resources, etc., and on this basis, makes a solution for joint optimization of latency and cost, thereby improving edge system performance, reducing deployment costs, and reducing the time cost and workload of edge system design.

[0068] Third, as auxiliary evidence of the inventiveness of the claims of the present invention, it is also reflected in the following important aspects:

[0069] The technical solution of the present invention fills the technical gap in the industry at home and abroad:

[0070] Existing service deployment solutions basically solve the problem from the user's perspective, focusing mainly on task processing delay, load balancing, and user satisfaction. However, in actual service deployment projects, only considering service quality without calculating deployment costs often leads to excessive deployment plan overhead and does not meet the project budget. The present invention targets edge environments where most of the system information is known in advance, comprehensively considers various aspects of the edge system, and provides a joint optimization solution for delay and cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0072] Figure 1 It is a flow chart of a method for deploying intelligent services in an edge computing environment provided by an embodiment of the present invention;

[0073] Figure 2 It is a schematic diagram of a method for deploying intelligent services in an edge computing environment provided by an embodiment of the present invention;

[0074] Figure 3 is an example diagram of a feasible solution matrix provided by an embodiment of the present invention;

[0075] Figure 4 It is a basic flow chart of using a differential evolution algorithm to obtain an optimal feasible solution to a problem provided by an embodiment of the present invention;

[0076] Figure 5 is an edge environment topology diagram provided by an embodiment of the present invention;

[0077] Figure 6 is a topological structure diagram of an embodiment provided by an embodiment of the present invention;

[0078] Figure 7 It is a diagram showing the effect of performing 100 iterations of optimization on the solution population generated by the above embodiment using three algorithms provided in the embodiment of the present invention. DETAILED DESCRIPTION

[0079] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0080] In view of the problems existing in the prior art, the present invention provides a method, system and terminal for intelligent service deployment in an edge computing environment. The present invention is described in detail below with reference to the accompanying drawings.

[0081] 1. Explanatory Examples In order to enable those skilled in the art to fully understand how to implement the present invention, this section provides an illustrative example that expands and describes the technical solution of the claims.

[0082] Explanation of terms: Request latency: the total time from the issuance of a service request to its processing; Server available resources: the currently idle CPU resources of a server; Pathfinding algorithm: the rules followed by a request to find an available server after it is issued.

[0083] like Figure 1 As shown, the intelligent service deployment method in the edge computing environment provided by the embodiment of the present invention includes the following steps:

[0084] S101, obtaining relevant information of various parts in the edge computing environment: reading information of each edge server, access point, network topology and service cost in the environment;

[0085] S102, each access point requests transmission path calculation: a transmission path for each service of each access point is obtained by a preset path finding algorithm;

[0086] S103, establishing a mathematical model of the system and an objective function: obtaining a mathematical model for calculating the delay and cost of the deployment solution based on the information obtained in S101 to S102;

[0087] S104, differential evolution calculation of optimal deployment solution: obtaining the optimal solution for service deployment based on the differential evolution algorithm.

[0088] As a preferred embodiment, Figure 2 As shown, the intelligent service deployment method in the edge computing environment provided by the embodiment of the present invention specifically includes the following steps:

[0089] Step 1: Obtain the available resources of each edge server and the deployment cost of each service unit resource; obtain the service type of each service, the request size of each service, and the request processing speed of each service when allocating unit resources;

[0090] Obtain the edge environment connection topology of the edge computing network, the average transmission rate of each link, and the preset request transmission pathfinding algorithm; obtain the average number of requests per unit time for each service of the access point and the wireless transmission rate of the access point. The resources allocated to each service by each server are used as the resource allocation vector of the server, and the resource allocation vectors of all edge servers are spliced ​​by column to form a matrix Y.

[0091] The data parameters used in the embodiment of the present invention are shown in Table 1.

[0092] Table 1 Data parameter description table used in the present invention

[0093] parameter significance Server available resources Maximum load on edge servers Service deployment cost The monetary cost of deploying a unit resource instance of a certain type of service Service request size The data size of a service request for a certain type of service Service processing speed The request processing speed of a certain type of service instance when allocating unit resources System topology diagram Connection topology of each edge server in the system Link transmission rate The average transfer rate between two servers Request pathfinding method System default request forwarding routing method Request access rate The average number of requests received per unit time for a service at a certain access point Wireless transmission rate The average rate at which the access point receives requests within range Resource Allocation Vector A server allocates resources to n services. Service deployment matrix It is formed by concatenating the resource allocation vectors of all servers in columns Request transfer path set Contains all node numbers of a request transmitted from the access point to the destination server

[0094] Step 2: According to the network topology and routing algorithm obtained in S1, each type of service request at each access point can find a fixed request transmission path set P. According to the path set P and the average transmission rate of each link contained in the path set, the average transmission time of this type of service on this access point can be obtained.

[0095] Step 3: Establish the objective function f(Y):

[0096] f(Y)=a1×C(Y)+a2×T(Y)

[0097]

[0098] C(Y)=∑ s∈S ∑ v∈V Y s，v ×β s

[0099] Where T(Y) is the total time consumption of each service request received by each access point per unit time, C(Y) is the capital cost of the Y matrix deployment solution, and λ s，a is the number of requests arriving per unit time for service s at access point a, d s is the data size of a single request to service s, v a is the receiving speed of access point a, μ s is the request processing speed of the service instance s per unit resource, v p is the average transmission speed of link p, β s is the deployment cost per unit resource of service s.

[0100] Set constraints on the objective function:

[0101] Constraint 1 is ∑ s∈S Y s，v ≤r v , that is, the sum of resources allocated by edge server v to each service cannot exceed the total available resources of v.

[0102] Constraint 2 is ∑ v∈V Y s，v ≥0, that is, each service must have at least one instance of allocated resources to prevent a certain type of service request from not being processed.

[0103] Step 4: Differential evolution calculates the optimal deployment solution. After obtaining the mathematical model of the system, the service deployment matrix is ​​used as the solution to the problem. The feasible solution is the one that meets the constraints in the above steps. First, a population of feasible solutions of the service deployment matrix Y of a reasonable size is randomly generated, and the f(Y) value of each feasible solution is obtained through the established objective function. As the fitness of the solution, it participates in the subsequent differential evolution process. Figure 3 shown.

[0104] Figure 3It represents a feasible solution for service deployment in an edge system with 4 edge servers and 5 service types. Taking the first column as an example, it means that edge server v1 deploys instances with 0, 0, 0.7, 1.3, and 1 unit resources for services s1, s2, s3, s4, and s5, respectively.

[0105] After connecting the columns of the feasible solution matrix into a vector, a real-coded chromosome can be formed. The fitness can be obtained through the mathematical model and system data in the above steps. The differential evolution algorithm is then used to obtain the optimal feasible solution to the problem. The algorithm is described as follows: (i) Set the differential evolution algorithm control parameters, the maximum evolutionary generation T, the crossover probability, the mutation probability, and the initial feasible solution population P obtained according to the constraints; (ii) Calculate the fitness of each feasible solution in the population P; (iii) Select individuals according to the fitness size according to the roulette algorithm to be inherited to the next generation of the population; (iv) Perform crossover and mutation operations to generate new individuals, and inherit the new individuals to the next generation of the population. Repeat (i) to (iv) until the genetic generation is T, and select the service deployment solution with the best fitness as the optimal solution. The basic process is as follows Figure 4 shown.

[0106] Figure 5 0, 1, and 2 are access points, and the other points are edge servers ( Figure 5 There are 7 in total), and then assuming that there are a total of 10 service requests in the environment, the problem is how to deploy these 10 services on these 7 servers and how many resources to allocate to each service so that the deployment cost is low while the request processing delay is short. This is the problem that this method aims to solve.

[0107] The intelligent service deployment system for edge computing environment provided by the embodiment of the present invention includes:

[0108] The information acquisition module is used to obtain relevant information of various parts in the edge computing environment, including reading the edge servers, access points, network topology and service cost information in the environment;

[0109] A transmission path calculation module, used to obtain the transmission path of each service at each access point by a preset path finding algorithm;

[0110] A system mathematical model building module, used to obtain a mathematical model for calculating the delay and cost of the deployment solution based on the information obtained by the information acquisition module and the transmission path calculation module requested by each access point;

[0111] The optimal deployment solution acquisition module is used to obtain the optimal solution for service deployment based on the differential evolution algorithm.

[0112] 2. Application Examples: In order to prove the creativity and technical value of the technical solution of the present invention, this section provides application examples of the technical solution of the claims on specific products or related technologies.

[0113] The present invention uses Python to implement an edge computing simulation environment to demonstrate the technical value of the present invention. Figure 6 is a topology of one embodiment.

[0114] Among them, 0, 1, and 2 are access points of the edge environment, 3, 4, 5, and 6 are edge servers where service instances can be deployed, and other nodes are routing nodes. Various types of information about the edge environment have been known in advance. According to the steps of the present invention, first obtain the relevant information of each part in the edge computing environment, then generate the mathematical model and objective function described in the previous section, and then randomly generate a population of deployment solutions that meet the requirements according to the rules, and finally use the differential evolution algorithm to solve the population.

[0115] 3. Evidence of the effects of the embodiments. The embodiments of the present invention have achieved some positive effects during the development or use process, and indeed have great advantages over the prior art. The following content is described in conjunction with the data, charts, etc. of the test process.

[0116] The experimental results of the above embodiments are compared in two aspects: performance comparison and efficiency comparison. The main purpose is to compare the differential evolution algorithm used in the present invention with the genetic algorithm and the particle swarm algorithm.

[0117] Performance comparison: The three algorithms were used to perform 100 iterations of optimization on the solution population generated by the above embodiment. The results are as follows: Figure 7 shown.

[0118] After 100 iterations, the objective function values ​​of the optimal solutions of each algorithm are

[0119] Differential evolution algorithm: 0.008826183

[0120] Genetic Algorithm: 0.008984239

[0121] Particle Swarm Optimization: 0.009040488

[0122] It can be seen that the optimization effect of the present invention using the differential evolution algorithm is 6.05% higher than that of other classical optimization algorithms.

[0123] Efficiency comparison: the time it takes to achieve N% optimization effect

[0124] 20％ 40％ 60％ 80％ 100％ Differential Evolution Algorithm 0.427196 0.427196 0.427196 3.282562 125.56 Genetic Algorithms 0.41986 0.41986 2.23266 2.82313 Unable to reach Particle Swarm Optimization 0.261621 0.261621 1.255416 4.5183 Unable to reach

[0125] It can be seen that the speed of the present invention in the process of optimizing to 60% is much faster than other methods; and when other methods converge after optimizing to more than 80%, the optimization method of the present invention can still spend more time to continue the optimization.

[0126] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0127] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A method for deploying intelligent services in an edge computing environment, characterized in that: The method for deploying intelligent services in an edge computing environment includes: Establish a mathematical model of the edge computing environment and determine the evaluation function that comprehensively considers migration cost and service latency; use the resource allocation matrix to model the service deployment scheme in the edge environment and design an optimization scheme for edge computing service deployment based on the differential evolution algorithm; The method for deploying intelligent services in an edge computing environment comprises the following steps: Step 1: Obtain relevant information about each part of the edge computing environment: read the edge servers, access points, network topology, and service cost information in the environment; Step 2: Each access point requests transmission path calculation: the transmission path of each service of each access point is obtained by a preset path finding algorithm; Step 3: Establishing the system mathematical model and objective function: Obtaining the mathematical model for calculating the delay and cost of the deployment solution based on the information obtained in steps 1 and 2; Step 4: Differential evolution calculates the optimal deployment solution: obtains the optimal solution for service deployment based on the differential evolution algorithm; The step 1 of obtaining relevant information of each part in the edge computing environment includes: Obtain the available resources of each edge server and the deployment cost of each service unit resource; obtain the service type of each service, the request size of each service, and the request processing speed of each service when allocating unit resources; Obtain the edge environment connection topology diagram of the edge computing network, the average transmission rate of each link, and the preset request transmission pathfinding algorithm; obtain the average number of requests per unit time for each type of service at the access point and the wireless transmission rate of the access point; use the resources allocated to each service by each server as the resource allocation vector of the server, and concatenate the resource allocation vectors of all edge servers by column to form a matrix Y.

2. The method for deploying intelligent services in an edge computing environment according to claim 1, characterized in that: The calculation of the transmission path requested by each access point in step 2 includes: According to the network topology map and path-finding algorithm obtained in step 1, each type of service request of each access point can find a fixed request transmission path set P; according to the path set P and the average transmission rate of each link included in the path set, the average transmission time of the service on the access point is obtained.

3. The method for deploying intelligent services in an edge computing environment according to claim 1, characterized in that: The system mathematical model and objective function establishment in step 3 include: Establish the objective function f(Y): f(Y)=a1×C(Y)+a2×T(Y); C(Y)=∑ s∈S ∑ v∈V AND s，v ×β s ; Where T(Y) is the total time consumption of each service request received by each access point per unit time, C(Y) is the capital cost of the Y matrix deployment solution, and λ s,a is the number of requests arriving per unit time for service s at access point a, d s is the data size of a single request to service s, v a is the receiving speed of access point a, μ s is the request processing speed of the service instance s per unit resource, v p is the average transmission speed of link p, β s is the deployment cost of service s per resource; Set constraints on the objective function: Constraint 1 is Σ s∈S Y s,v ≤r v ,The sum of resources allocated by edge server v to each service cannot exceed the total available resources of v; Constraint 2 is ∑ v∈V Y s,v ≥0, each service has at least one instance of allocated resources.

4. The method for deploying intelligent services in an edge computing environment according to claim 1, characterized in that: The optimal deployment solution of differential evolution calculation in step 4 includes: After obtaining the mathematical model of the system, the service deployment matrix is ​​used as the solution to the problem, and the feasible solution is the one that meets the constraints in step 3. A population of feasible solutions of the service deployment matrix Y of a reasonable size is randomly generated, and the f(Y) value is obtained for each feasible solution through the established objective function. As the fitness of the solution, it participates in the subsequent differential evolution process; The differential evolution calculation optimal deployment scheme also includes: connecting the columns of the feasible solution matrix into a vector to form a real-number coded chromosome; obtaining the fitness through a mathematical model and system data. Use the differential evolution algorithm to obtain the optimal feasible solution to the problem, including: 1) Set the control parameters of the differential evolution algorithm, the maximum evolutionary number T, the crossover probability, the mutation probability, and the initial feasible solution population Q obtained according to the constraints; 2) Calculate the fitness of each feasible solution in the population Q; 3) Select individuals according to the fitness size according to the roulette algorithm to be inherited to the next generation of the population; 4) Perform crossover and mutation operations to generate new individuals, and inherit the new individuals to the next generation of the population; 5) Repeat steps 1) to 4) until the genetic number is T, and select the service deployment solution with the best fitness as the optimal solution.

5. A system for deploying intelligent services in an edge computing environment using the method for deploying intelligent services in an edge computing environment as claimed in any one of claims 1 to 4, characterized in that: The intelligent service deployment system in the edge computing environment includes: The information acquisition module is used to obtain relevant information of various parts in the edge computing environment, including reading the edge servers, access points, network topology and service cost information in the environment; A transmission path calculation module, used to obtain the transmission path of each service at each access point by a preset path finding algorithm; A system mathematical model building module, used to obtain a mathematical model for calculating the delay and cost of the deployment solution based on the information obtained by the information acquisition module and the transmission path calculation module requested by each access point; The optimal deployment solution acquisition module is used to obtain the optimal solution for service deployment based on the differential evolution algorithm.

6. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for deploying intelligent services in an edge computing environment as described in any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the method for deploying intelligent services in an edge computing environment as described in any one of claims 1 to 4.

Images