Efficient deployment and scheduling method for sharing virtualized network function in edge network

Abstract

The invention provides an efficient deployment and scheduling method for sharing a virtualized network function in an edge network. The method comprises the following steps: step 1, constructing a mobile edge network model; step 2, determining a target of a deployment and scheduling problem of a shared virtualized network function in the edge network; 3, determining the coverage range of each edge server, establishing the coverage capability of each edge server, and converting the deployment and scheduling problem of the shared virtualized network function into a set coverage problem; and step 4, after the deployment and scheduling problem of the shared virtualization network function is converted into a set coverage problem, an efficient deployment and scheduling algorithm of the shared virtualization network function based on harmony search is adopted to solve the problem. According to the efficient deployment and scheduling method for the shared virtualized network function in the edge network, the performance of the edge network is evaluated through wide simulation, and a simulation result shows that the performance of the method is superior to that of an existing method.

Description

Technical Field

[0001] The present invention relates to the field of edge computing and Internet of Things technologies, and in particular to a method for efficiently deploying and scheduling shared virtualized network functions in an edge network. Background Art

[0002] Mobile Edge Computing (MEC) is considered a promising technology for providing low-latency services by keeping computing and other resources where they are needed. The functions implemented in MEC through Network Function Virtualization (NFV) are called Virtual Network Function (VNF) instances. VNF instances come in many types, such as intrusion detection and prevention, load balancing, WAN acceleration, caching, and session border controllers. Sometimes, different sub-functions can be combined to form a higher-level component, such as a virtual router. In real-world applications, multiple users often request instances of the same type, allowing a single instance to concurrently serve multiple user requests. Such instances are called shared instances. Implementing shared VNF instances can effectively reduce the costs associated with service deployment. The deployment and scheduling of VNF instances have been a hot topic of research. Deployment refers to placing instances on edge servers, while scheduling refers to allocating resources to fulfill user requests. However, most existing work fails to consider the deployment and scheduling of VNF instances, resulting in inefficient and inefficient fulfillment of user requests. Furthermore, enabling users to share instances of the same type, rather than allocating a separate instance to each user, can significantly reduce deployment costs.

[0003] Currently, many studies focus on the research of edge computing task offloading strategies for cloud-edge collaboration. Cui et al. (see Y.Cui, D.Zhang, T.Zhang, P.Yang, and H.Zhu. A New Approach on Task Offloading Scheduling for Application of Mobile Edge Computing. in Proc. of IEEE WCNC, 2021, pp. 1–6.) proposed a new task offloading and scheduling method for MEC applications. In order to reduce costs, a fine-grained task scheduling and offloading strategy for multi-user MEC systems is proposed. Yan et al. (see J.Yan, S.Bi, YJZhang, and M.Tao. Optimal Task Offloading and Resource Allocation in Mobile-Edge Computing With Inter-User Task

[0004] Dependency. IEEE Transactions on Wireless Communications, vol. 19, no. 1, pp. 235–250, 2020.) studied the optimal task offloading strategy and resource allocation. This problem is challenging due to the strong coupling between task offloading decisions and resource allocation. To address this problem, they first assumed the offloading decision and then proposed an efficient binary search method to obtain the optimal solution.

[0005] However, the above works consider the deployment and scheduling strategies of VNF instances separately, without considering them jointly. The proposed VNF instance efficient deployment and scheduling strategy jointly considers the deployment and scheduling of VNF instances, from the perspectives of both network providers and users, strictly guarantees the user's latency requirements, satisfies as many user requests as possible, and reduces the deployment and scheduling costs.

[0006] The Chinese patent application with application number CN201911300727.0 involves an edge network request scheduling decision-making method based on a deep Q network, in which the current network status and request queue are obtained through the network data acquisition subsystem. Through the feature embedding subsystem, the network node information is abstracted into a feature vector using a graph embedding method. Through the micro-cloud selection subsystem and the node selection subsystem, the feature vector is read, and a decision model is established using a deep reinforcement learning method. The model is stored in a mobile edge network request scheduling decision application system. Finally, the mobile edge network request scheduling decision application system is used to obtain the current network status and request queue required for decision-making in real time. Under the condition that the decision model has fully converged, the request scheduling path of the managed network can be determined and applied to the managed network.

[0007] In the Chinese patent application with application number: CN202110074556.5, a workflow scheduling method SAWS based on deep reinforcement learning in an edge computing environment is involved. The main process of implementing the present invention is to first construct the problem into a Markov decision process, define the rewards, states and actions corresponding to the workflow scheduling problem in the edge computing environment, and then calculate the weights of the task nodes according to the execution time, transmission time and dependencies of the task nodes in the workflow, and then make decisions on the scheduling of the task nodes based on the deep Q network. The main goal of the SAWS strategy is to find a task scheduling strategy that can minimize the long-term execution delay of the workflow while ensuring the security of user information. This invention greatly improves the execution efficiency of the workflow in a mobile edge network environment while ensuring the security of user information through the learning and decision-making of the Q neural network.

[0008] In the Chinese patent application with application number CN202111112170.5, a method for intelligent resource allocation in a mobile edge network with divisible tasks is involved, including: dividing the serial tasks generated by the terminal into multiple subtasks and establishing an offloading task model; establishing delay and energy consumption models for the subtasks according to the two execution modes of local or offloading, and defining an offloading joint objective optimization function based on multi-user serial dependent tasks; in a multi-server scenario, a Markov game model is established based on the cooperative and competitive relationship of multiple users for wireless communication and computing resources, and the offloading joint objective optimization function is optimized; in a time-varying environment, each terminal executes a reinforcement learning algorithm as a separate intelligent agent based on partial system status information to solve the Markov game model, and determine the offloading strategy, sub-channel selection, transmission power, and resource allocation amount. The present invention is conducive to the reasonable allocation of server resources and the full use of fragmented resources, ensuring the terminal user experience and improving the stability of network operation.

[0009] The above existing technologies are significantly different from the present invention and fail to solve the technical problems to be solved. Therefore, a new efficient deployment and scheduling method for shared virtualized network functions in edge networks is invented. Summary of the Invention

[0010] The purpose of the present invention is to provide an efficient deployment and scheduling scheme SCEDS based on set coverage, which efficiently completes the efficient deployment and scheduling method of shared virtualized network functions in edge networks for VNF instance deployment and scheduling.

[0011] The objectives of the present invention can be achieved by the following technical measures: an efficient deployment and scheduling method for shared virtualized network functions in an edge network, the efficient deployment and scheduling method for shared virtualized network functions in an edge network comprising:

[0012] Step 1: Build a mobile edge network model;

[0013] Step 2: Determine the goal of deploying and scheduling shared virtualized network functions in edge networks.

[0014] Step 3: Determine the coverage of each edge server and establish their coverage capabilities, transforming the deployment and scheduling problem of shared virtualized network functions into a collective coverage problem.

[0015] In step 4, after converting the deployment and scheduling problem of shared virtualized network functions into a set covering problem, an efficient deployment and scheduling algorithm of shared virtualized network functions based on harmony search is adopted to solve the problem.

[0016] The purpose of the present invention can also be achieved by the following technical measures:

[0017] In step 1, the model of the mobile edge network includes the edge computing system model, user request model, cost model and delay model.

[0018] In step 1, when building the edge computing system model, an edge computing system is established as W = (A, L, U); where A represents the base station of the edge network, i.e., the access point set, L represents the link set between the access points, and U represents the user set; it is assumed that each network access point has an edge server, each user has a user request; each edge server has computing resources Cr n and storage resources Sr n .

[0019] In step 1, if a shared VNF instance needs to be deployed on an edge server, either a new shared VNF instance is instantiated or a shared VNF instance is migrated. The two cannot coexist at the same time; that is, constraint condition (1): where y i,n , a i,n , b i,n,n' Both are binary variables, indicating whether the edge server cl n The deployment type is f i VNF instances, whether to instantiate new VNFs, whether to migrate VNF instances;

[0020] In addition, the same type of shared VNF instances can be deployed on multiple edge servers in the edge network, i.e., constraint (2): Among them, x i,n is a binary variable representing f i Whether the VNF instance of this type is deployed on the edge server cl n superior.

[0021] In step 1, the user request model is where f j (i) represents the type of virtualized network function (VNF) requested by the user, r j Indicates the data packet rate transmitted by the user, represents the delay requirement of the user request; the user request is completed by the VNF instance deployed in the edge server; the computing resources and storage resources consumed by different user requests are related to the packet rate, request type, and edge server, and are defined as r j ·Cr(f i ,cl n ) and r j ·Sr(f i ,cl n); and the remaining computing resources and storage resources of the edge server are sufficient to meet the needs of users, that is, constraints (3) and (4): In addition, user requests are processed and completed by at most one and only one edge server, that is, constraint (5): where k j,n Request u on behalf of user j Is it cl by the edge server n Computational processing.

[0022] In step 1, the cost model is divided into three parts: operation cost, transmission cost, and VNF instantiation cost or VNF migration cost: where the instantiation cost or migration cost is represented by C m , the wireless transmission cost is C t , the cost of processing user requests is C p .

[0023] In step 1, the total delay in completing the user request includes communication delay, computation processing delay, VNF instantiation delay or VNF migration delay, and is defined as where d tra (u j ,cl n ) is the communication delay, d ins (f i ,cl n ) is the instantiation delay, d mig (f i ,cl n ,cl n' ) is the migration delay, d pro (u j ,cl n ) is the calculation processing delay; a i,n and b i,n,n' is a binary decision variable; and the total delay in completing the user request cannot exceed the delay requirement of the user request, that is, constraint (6):

[0024] In step 2, the goal of the present invention is to minimize the cost to the service provider while maximizing the network throughput.

[0025]

[0026] The network throughput is α and β are normalization coefficients; the instantiation cost or migration cost is expressed as C m , the wireless transmission cost is C t , the cost of processing user requests is C p .

[0027] In step 3, the delay requirement of the user request and the actual delay of the edge server to complete the user request are calculated. in, Indicates user request u j The delay requirement, Indicates that the user request u is completed j This step accurately determines which edge servers can satisfy the user's request under the specific latency requirement.

[0028] Step 4 includes:

[0029] 41) First determine the number of user requests covered by each edge server and the cost of completing these user requests, which can be expressed as follows: S i,n =|U' i,n | and Where U' i,n Represents the edge server cl n The service request type covered is f i The user set S i,n represents the number of user requests covered by each edge server, C i,n represents the cost of completing these user requests, represents the transmission cost required to complete the covered user requests, C ins (f i ,cl n )·a i,n and Represents the virtualized network function instantiation cost and migration cost, represents the running cost required to complete the covered user requests, a i,n and b i,n,n' is a binary decision variable;

[0030] 42) Since the problem has been transformed, set parameters Thus, the objective function of the problem is also transformed; the objective function is redefined as Among them, X hms is a harmony vector in the initial harmony matrix HM, hms and hvd represent the rows and columns in the harmony matrix respectively, and HVD represents the maximum number of columns;

[0031] 43) Then, new harmony vectors are generated to continuously optimize the solution space; the generation of harmony vectors is performed using the following two rules:

[0032]

[0033]

[0034] Among them, λ new,hvd represents the newly generated harmony vector, ω1, ω2 and ω3 are random numbers uniformly distributed in the interval [0,1], HM represents the harmony matrix (or harmony memory), HMCR represents the probability of randomly extracting a harmony vector from the harmony matrix, BW represents the amplitude of fine-tuning, and PAR represents the pitch adjustment rate;

[0035] 44) On the basis of ensuring the feasibility of the new harmonic vector, the initial solution space is updated; the new harmonic vector is evaluated, and if it is better than the harmonic vector in the initial solution space with the worst objective function value, the new harmonic vector is updated to the solution space; then, new harmonic vectors are continuously generated and the solution space is updated until the maximum number of iterations is reached;

[0036] 45) The optimal harmony vector of the final solution space is the final deployment and scheduling of the shared virtualized network function; due to the limited computing and storage resources of the edge server, not all user requests are completed, that is, the constraint condition (7): where k j,n is a binary variable representing the edge server cl n Whether to process user request u j ; In addition, the running time of the algorithm is fully considered to achieve real-time requirements as much as possible, and the final experimental running time also proves this.

[0037] The objectives of the present invention can also be achieved through the following technical measures: an efficient deployment and scheduling system for shared virtualized network functions in an edge network, which adopts an efficient deployment and scheduling method for shared virtualized network functions in an edge network to solve the deployment and scheduling problems of shareable virtualized network functions in a mobile edge network.

[0038] The efficient deployment and scheduling method of shared virtualized network functions in edge networks in the present invention studies the effective deployment and scheduling of VNF instances shared between different users under the constraints of user latency and network resources. First, a VNF instance deployment and scheduling model is established in the MEC network to study how to minimize costs and maximize network throughput under the constraints of user latency and cloud computing and storage resources. Then, utilizing its sharing characteristics, an efficient deployment and scheduling scheme based on set coverage, SCEDS, is proposed to efficiently complete the deployment and scheduling of VNF instances. Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1) The proposed SCEDS (Shared Virtualized Network Function Efficiency Deployment and Scheduling Strategy) in edge networks based on harmonious search can reasonably schedule and improve network throughput according to the changes in computing and storage resources on a single edge server;

[0040] 2) The proposed SCEDS (Shared Virtualized Network Function Efficiency Deployment and Scheduling Strategy) in edge networks based on harmonious search can reasonably schedule and reduce average costs according to the changes in computing and storage resources on a single edge server;

[0041] 3) The proposed SCEDS (Shared Virtualized Network Function Efficiency Deployment and Scheduling Strategy) can reasonably schedule edge servers according to the number of edge servers, thereby improving network throughput.

[0042] 4) The efficient deployment and scheduling strategy (SCEDS) of shared virtualized network functions based on harmony search in the edge network proposed in the present invention can reasonably perform scheduling according to the changes in the number of edge servers and reduce the average cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 Schematic diagram of the scenario of the efficient deployment and scheduling strategy (SCEDS) of shared virtualized network functions in edge networks based on harmony search proposed by the present invention;

[0044] Figure 2 A schematic diagram comparing the network throughput of the proposed SCEDS (Efficient Deployment and Scheduling Strategy for Shared Virtualized Network Functions) in an edge network with the changes in computing resources and storage resources on a single edge server.

[0045] Figure 3 A schematic diagram comparing the average cost of completing a single user request with the computing resources and storage resources on a single edge server between the proposed SCEDS strategy for efficient deployment and scheduling of shared virtualized network functions based on harmony search in edge networks and the existing strategy;

[0046] Figure 4 A schematic diagram comparing the network throughput of the proposed SCEDS (Efficient Deployment and Scheduling Strategy for Shared Virtualized Network Functions) in an edge network with the number of edge servers and the existing strategy;

[0047] Figure 5A schematic diagram comparing the average cost of completing a single user request with the number of edge servers between the proposed SCEDS (Efficient Deployment and Scheduling Strategy for Shared Virtualized Network Functions) and the existing strategy in an edge network;

[0048] Figure 6 A schematic diagram of an embodiment of a GA solution for virtualized network function deployment and scheduling;

[0049] Figure 7 A schematic diagram of an embodiment of a TS solution for virtualized network function deployment and scheduling;

[0050] Figure 8 A schematic diagram of an embodiment of a SCEDS strategy for virtualized network function deployment and scheduling;

[0051] Figure 9 This is a flow chart of the method for efficiently deploying and scheduling shared virtualized network functions in an edge network of the present invention. DETAILED DESCRIPTION

[0052] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0053] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations and / or combinations thereof.

[0054] This paper discloses a state-of-the-art method for the efficient deployment and scheduling of shared virtualized network functions in edge networks. Mobile edge computing (MEC) is considered a promising technology that provides low-latency services by keeping computing and other resources physically close to where they are needed. MEC has several key technologies, among which NFV leverages virtualization technology to implement network functions in software, thus no longer being limited to hardware architectures. Furthermore, by allowing users of the same type to share the same virtualized network function (VNF) instance, rather than allocating a separate instance to each user, the cost of service deployment can be significantly reduced. The purpose of this paper is to study the efficient deployment and scheduling of VNF instances shared between different services. An MEC network model is established, and VNF instances are centrally deployed on edge servers. Virtualized network function services are provided to mobile users by migrating VNF instances and instantiating new ones. An optimization problem is proposed: minimizing cost while maximizing network throughput. Then, leveraging the shared nature of VNF instances, an efficient deployment and scheduling scheme (SCEDS) based on harmony search is proposed, and its performance is evaluated through extensive simulations. Simulation results show that this method outperforms existing methods.

[0055] like Figure 9 As shown, Figure 9 This is a flowchart of the efficient deployment and scheduling method for shared virtualized network functions in the edge network of the present invention. In order to effectively solve the deployment and scheduling problems of shareable virtualized network functions in mobile edge networks, the present invention first needs to build a mobile edge network model and then clarify our objective function. Secondly, the deployment and scheduling problem we defined is converted into a set coverage problem and solved using a deployment and scheduling algorithm for a shared virtualized network based on harmony search. We call this process an efficient deployment and scheduling strategy (SCEDS) for shared virtualized network functions based on harmony search. The efficient deployment and scheduling method for shared virtualized network functions in the edge network specifically includes:

[0056] 1. Build a mobile edge network model. The mobile edge network model includes the edge computing system model, user request model, cost model, and delay model.

[0057] 1) First, let's consider the edge computing system model. We set up an edge computing system W = (A, L, U). A represents the set of base stations (i.e., access points) in the edge network, L represents the set of links between access points, and U represents the set of users. Assume that each network access point has an edge server, and each user has a user request.

[0058] Each edge server has computing resources Cr n and storage resources Sr n .

[0059] If a shared VNF instance needs to be deployed on an edge server, either a new shared VNF instance must be instantiated or a shared VNF instance must be migrated. The two cannot coexist at the same time. That is, the constraint condition (1): where y i,n , a i,n , b i,n,n' Both are binary variables, indicating whether the edge server cl n The deployment type is f i VNF instance, whether to instantiate a new VNF, whether to migrate the VNF instance.

[0060] In addition, the same type of shared VNF instances can be deployed on multiple edge servers in the edge network, i.e., constraint (2): Among them, x i,n is a binary variable representing f i Whether the VNF instance of this type is deployed on the edge server cl n superior.

[0061] 2) The user request model is where f j (i) represents the type of virtualized network function (VNF) requested by the user, r j Indicates the data packet rate transmitted by the user, represents the latency requirement of the user request. The user request is completed by the VNF instance deployed in the edge server. The computing resources and storage resources consumed by different user requests are related to the packet rate, request type, and edge server, and are defined as r j ·Cr(f i ,cl n ) and r j ·Sr(f i ,cl n ). The remaining computing resources and storage resources of the edge server are sufficient to meet the needs of users, that is, constraints (3) and (4):

[0062]

[0063] In addition, user requests are processed and completed by at most one and only one edge server, that is, constraint (5): where k j,n Request u on behalf of user j Is it cl by the edge server n Computational processing.

[0064] 3) The cost model is divided into three parts: operation cost, transmission cost, and VNF instantiation cost or VNF migration cost: the instantiation cost or migration cost is represented by C m , the wireless transmission cost is C t , the cost of processing user requests is C p .

[0065] 4) In this process, the total delay in completing the user request includes communication delay, computation processing delay, VNF instantiation delay or VNF migration delay, which is defined as where d tra (u j ,cl n ) is the communication delay, d ins (f i ,cl n ) is the instantiation delay, d mig (f i ,cl n ,cl n' ) is the migration delay, d pro (u j ,cl n ) is the calculation processing delay. a i,n and b i,n,n' is a binary decision variable. The total delay in completing a user request cannot exceed the delay requirement of the user request, that is, constraint (6):

[0066] 2. Next, we determine the goal of the deployment and scheduling problem of shared virtualized network functions in edge networks: The goal of this invention is to minimize the cost of the service provider while maximizing the network throughput.

[0067]

[0068] The network throughput is α and β are normalization coefficients.

[0069] 3. The above problem has been proven to be an NP-hard problem. In order to better solve this problem, we transform it into a set cover problem. In this invention, the delay requirement of the user request and the actual delay of the edge server to complete the user request are used to calculate the delay of the edge server. It is possible to accurately determine which edge servers can satisfy a user's request under a specific latency requirement. In other words, the coverage of each edge server can be determined, thereby establishing their coverage capabilities. This transforms the problem into a set coverage problem.

[0070] 4. After transforming the deployment and scheduling problem of shared virtualized network functions into a set covering problem, we use an efficient deployment and scheduling algorithm for shared virtualized network functions based on harmony search to solve the problem.

[0071] 1) First determine the number of user requests covered by each edge server and the cost of completing these user requests, which can be expressed as follows: S i,n =|U' i,n | and Where U' i,n Represents the edge server cl n The service request type covered is f i The user collection, represents the virtualized network function instantiation cost or migration cost, a i,n and b i,n,n' is a binary decision variable.

[0072] 2) Since the problem has been transformed, we set up parameters Thus, the objective function of the problem is also transformed. The objective function is repositioned as Among them, X hms is a harmony vector in the initial harmony matrix.

[0073] 3) Then generate new harmony vectors to continuously optimize the solution space. The generation of harmony vectors is carried out using the following two rules:

[0074]

[0075]

[0076] Among them, ω2 and ω3 are random numbers uniformly distributed in the interval [0,1], BW represents the amplitude of fine-tuning, and PAR represents the pitch adjustment rate.

[0077] 4) The initial solution space is updated while ensuring the feasibility of the new harmonic vector. The new harmonic vector is evaluated, and if it is better than the harmonic vector in the initial solution space with the worst objective function value, the new harmonic vector is updated to the solution space. Then, new harmonic vectors are generated and the solution space is updated until the maximum number of iterations is reached.

[0078] 5) The optimal harmony vector of the final solution space is the final deployment and scheduling of the shared virtualized network function. However, the present invention needs to clarify that due to the limited computing and storage resources of the edge server, not all user requests are completed, that is, the constraint condition (7): In addition, we fully consider the running time of the algorithm to achieve real-time requirements as much as possible, and the final experimental running time also proves this.

[0079] The following are several specific embodiments of the present invention:

[0080] Example 1:

[0081] In a specific embodiment 1 of the present invention, the efficient deployment and scheduling method of shared virtualized network functions in the edge network includes the following steps:

[0082] 1. First, a model of a shared virtualized network system in an edge network is constructed. The tasks are divided into two parts: deployment of shared VNF instances and scheduling of shared VNF instances. The costs of this process are divided into three parts: operation cost, transmission cost, and VNF instantiation cost or VNF migration cost. The total latency of the task includes communication latency, computational processing latency, VNF instantiation latency, and VNF migration latency. For details on the cost and latency models, see the Summary of the Invention.

[0083] 2. Based on the constructed model, the system formulates corresponding service entity placement strategies.

[0084] Input network access point A and its edge server set Cl, user request set U, where the attribute of each user request is

[0085] 1) Through the formula

[0086] Get the latency of each user's request on different servers

[0087] 2) Through constraints To determine which edge servers can complete the user request u j ,Continue to iterate until all user requests are determined.

[0088] 3) It is necessary to determine the number of user requests covered by each edge server and the cost of completing these user requests, which can be expressed by the following formulas: S i,n =|U′ i,n | and

[0089]

[0090] 4) Set parameters This transforms the problem goal into Among them, X hms is a harmony vector in the initial harmony matrix.

[0091] 5) Then generate new harmony vectors to continuously optimize the solution space. The generation of harmony vectors is carried out using the following two rules:

[0092]

[0093]

[0094] Among them, ω2 and ω3 are random numbers uniformly distributed in the interval [0,1], BW represents the amplitude of fine-tuning, and PAR represents the pitch adjustment rate.

[0095] 6) The initial solution space is updated while ensuring the feasibility of the new harmony vector. The new harmony vector is evaluated, and if it is better than the harmony vector in the initial solution space with the worst objective function value, the new harmony vector is updated to the solution space.

[0096] 7) Then, continue to generate new harmonic vectors and update the solution space until the maximum number of iterations is reached. The optimal harmonic vector in the final solution space is the required deployment and scheduling situation.

[0097] 8) In edge network systems, the goal is to minimize the cost of the service provider while maximizing the network throughput. In addition, the algorithm runtime must also be considered to achieve real-time requirements as much as possible.

[0098] like Figure 2 As shown, Figure 2 This is a comparative diagram of the network throughput of the proposed SCEDS strategy for shared virtualized network functions in edge networks and the existing strategy as the computing resources and storage resources on a single edge server change. Figure 2 It can be seen that the efficient deployment and scheduling strategy (SCEDS) of shared virtualized network functions based on harmony search in the edge network proposed in the present invention can reasonably perform scheduling according to the changes in computing resources and storage resources on a single edge server, thereby improving the network throughput.

[0099] like Figure 3 As shown, Figure 3 This is a comparative diagram of the average cost of completing a single user request in the edge network based on the efficient deployment and scheduling strategy (SCEDS) of the proposed invention and the existing strategy as the computing resources and storage resources on a single edge server change. Figure 3 It can be seen that the efficient deployment and scheduling strategy (SCEDS) of shared virtualized network functions based on harmony search in the edge network proposed in the present invention can reasonably perform scheduling according to the changes in computing resources and storage resources on a single edge server, thereby reducing the average cost.

[0100] like Figure 4 As shown, Figure 4 This is a comparative diagram of the network throughput of the proposed SCEDS strategy for sharing virtualized network functions in edge networks and the existing strategy as the number of edge servers changes. Figure 4 It can be seen that the efficient deployment and scheduling strategy (SCEDS) of shared virtualized network functions based on harmony search in the edge network proposed in the present invention can reasonably perform scheduling according to the changes in the number of edge servers and improve the network throughput.

[0101] like Figure 5 As shown, Figure 5 This is a comparative diagram of the average cost of completing a single user request in the edge network based on the efficient deployment and scheduling strategy (SCEDS) of the shared virtualized network function proposed by the present invention and the existing strategy as the number of edge servers changes. Figure 5 It can be seen that the efficient deployment and scheduling strategy (SCEDS) of shared virtualized network functions based on harmony search in the edge network proposed in the present invention can reasonably perform scheduling according to the changes in the number of edge servers and reduce the average cost.

[0102] Example 2:

[0103] As attached Figure 6 As shown in Figures 7 and 8, there are three APs (base stations) and 11 users in the system. Three types of shared virtualized network functions are used. Assuming each AP has an edge server and each user has a user request task, VNF instances are deployed on the edge server to fulfill user requests. Different deployment and scheduling strategies result in different costs and network throughput. Each edge server can handle up to four user requests.

[0104] The mobile edge network VNF instance system uses an efficient deployment and scheduling strategy for shared virtualized network functions based on harmony search to determine the deployment location for each VNF instance and the scheduling plan for each user request, minimizing service provider costs and maximizing network throughput. The specific implementation steps are the same as those in Example 1.

[0105] The deployment and scheduling of the GA solution are as follows: Figure 6 As shown in the figure, the user range that each edge server can cover is determined by the user's delay requirement and the actual delay of each edge server to complete the user request. User requests U2, U3, U4, U6 choose to connect to AP1; U1, U5, U7 connect to AP2; U8, U9, U10, U11, U12, U13, U14, U15, U16, U17, U18, U19, U20, U21, U22, U23, U24, U25, U26, U27, U28, U29, U30, U31, U32, U33, U34, U35, U36, U37, U38, U39, U40, U41, U42, U43, U44, U45, U46, U47, U48, U49, U50, U51, U52, U53, U5 10 、U 11 Connect AP3. The deployment and scheduling of TS solution and SCEDS strategy are as follows: Figure 7 and Figure 8 As shown, no further details are given here.

[0106] All user requests in the three solutions are completed. The costs of the three solutions are shown in Table 1:

[0107] Table 1 Cost comparison of three solutions

[0108]

[0109] It is not difficult to see from the table that the proposed SCEDS scheme has lower cost.

[0110] Example 3:

[0111] In order to more fully demonstrate the superiority of the present invention, we reset different parameters.

[0112] We set up an edge system with 4 APs (base stations) and 30 users, and set the types of shared virtualized network functions to 5. Each AP has an edge server, and each edge server can handle up to 8 user requests. The specific implementation steps are the same as in Example 1.

[0113] All user requests in the three solutions are completed. Only the costs of the three solutions are shown in Table 2:

[0114] Table 2 Cost comparison of three solutions

[0115]

[0116] Example 4:

[0117] We set up an edge system with five APs (base stations) and 50 users, with five types of shared virtualized network functions. Each AP is equipped with an edge server. The computing and storage resources of each edge server are randomly selected between 20MHz and 70MHz, and the computing and storage resources consumed by each user request are randomly selected between 2MHz and 10MHz.

[0118] Under this parameter setting, the edge server has limited resources, so not all user requests are processed by the edge server, that is, the network throughput and total cost are different under different solutions. The specific implementation steps are the same as in Example 1.

[0119] The network throughput and total cost of the three solutions are shown in Table 3:

[0120] Table 3 Comparison of network throughput and total cost of three solutions

[0121]

[0122] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art may modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features therein. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

[0123] Except for the technical features described in the specification, all other technical features are known technologies to those skilled in the art.

Claims

1. An efficient deployment and scheduling method for shared virtualized network functions in edge networks, characterized in that: The efficient deployment and scheduling method for shared virtualized network functions in the edge network includes: Step 1: Build a mobile edge network model; Step 2: Determine the goal of deploying and scheduling shared virtualized network functions in edge networks. Step 3: Determine the coverage of each edge server and establish their coverage capabilities, transforming the deployment and scheduling problem of shared virtualized network functions into a collective coverage problem. In step 4, after converting the deployment and scheduling problem of shared virtualized network functions into a set covering problem, an efficient deployment and scheduling algorithm of shared virtualized network functions based on harmony search is adopted to solve the problem.

2. The efficient deployment and scheduling method for shared virtualized network functions in an edge network according to claim 1, characterized in that: In step 1, the model of the mobile edge network includes the edge computing system model, user request model, cost model and delay model.

3. The efficient deployment and scheduling method for shared virtualized network functions in an edge network according to claim 2, characterized in that: In step 1, when building the edge computing system model, an edge computing system is established as W = (A, L, U); where A represents the base station of the edge network, i.e., the access point set, L represents the link set between the access points, and U represents the user set; it is assumed that each network access point has an edge server, each user has a user request; each edge server has computing resources Cr n and storage resources Sr n .

4. The efficient deployment and scheduling method for shared virtualized network functions in an edge network according to claim 3, characterized in that: In step 1, if a shared VNF instance needs to be deployed on an edge server, either a new shared VNF instance is instantiated or a shared VNF instance is migrated, and the two cannot exist at the same time; that is, constraint condition (1): y i,n ≤a i,n +b i,n,n' ≤1, where y i,n , a i,n , b i,n,n' Both are binary variables, indicating whether the edge server cl n The deployment type is f i VNF instances, whether to instantiate new VNFs, whether to migrate VNF instances; In addition, the same type of shared VNF instances can be deployed on multiple edge servers in the edge network, i.e., constraint (2): Among them, x i,n is a binary variable representing f i Whether the VNF instance of this type is deployed on the edge server cl n superior.

5. The efficient deployment and scheduling method for shared virtualized network functions in an edge network according to claim 4, characterized in that: In step 1, the user request model is u j = where f j (i) represents the type of virtualized network function (VNF) requested by the user, r j Indicates the data packet rate transmitted by the user, represents the delay requirement of the user request; the user request is completed by the VNF instance deployed in the edge server; the computing resources and storage resources consumed by different user requests are related to the packet rate, request type, and edge server, and are defined as r j ·Cr(f i ,cl n ) and r j ·Sr(f i ,cl n ); and the remaining computing resources and storage resources of the edge server are sufficient to meet the needs of users, that is, constraints (3) and (4): In addition, user requests are processed and completed by at most one and only one edge server, that is, constraint (5): where k j,n Request u on behalf of user j Is it cl by the edge server n Computational processing.

6. The efficient deployment and scheduling method for shared virtualized network functions in an edge network according to claim 5, characterized in that: In step 1, the cost model is divided into three parts: operation cost, transmission cost, and VNF instantiation cost or VNF migration cost: where the instantiation cost or migration cost is represented by C m , the wireless transmission cost is C t , the cost of processing user requests is C p .

7. The efficient deployment and scheduling method for shared virtualized network functions in an edge network according to claim 6, characterized in that: In step 1, the actual total delay in completing the user request includes communication delay, computation processing delay, VNF instantiation delay or VNF migration delay, and is defined as where d tra (u j ,cl n ) is the communication delay, d ins (f i ,cl n ) is the instantiation delay, d mig (f i ,cl n ,cl n' ) is the migration delay, d pro (u j ,cl n ) is the calculation processing delay; a i,n and b i,n,n' is a binary decision variable; and the total delay in completing the user request cannot exceed the delay requirement of the user request, that is, constraint (6):

8. The efficient deployment and scheduling method for shared virtualized network functions in an edge network according to claim 1, characterized in that: In step 2, the goal of the present invention is to minimize the cost to the service provider while maximizing the network throughput. The network throughput is α and β are normalization coefficients; the instantiation cost or migration cost is expressed as C m , the wireless transmission cost is C t , the cost of processing user requests is C p .

9. The efficient deployment and scheduling method for shared virtualized network functions in an edge network according to claim 1, characterized in that: In step 3, the delay requirement of the user request and the actual delay of the edge server to complete the user request are calculated. in, Indicates user request u j The delay requirement, Indicates that the user request u is completed j The actual total delay of the user is calculated; this step can accurately determine which edge servers can meet the user's request under the specific delay requirement.

10. The efficient deployment and scheduling method for shared virtualized network functions in an edge network according to claim 1, characterized in that: Step 4 includes: 41) First determine the number of user requests covered by each edge server and the cost of completing these user requests, which can be expressed as follows: S i,n =|U' i,n | and where U' i,n Represents the edge server cl n The service request type covered is f i The user set S i,n represents the number of user requests covered by each edge server, C i,n represents the cost of completing these user requests, represents the transmission cost required to complete the covered user requests, Represents the virtualized network function instantiation cost or migration cost, represents the running cost required to complete the covered user requests, a i,n and b i,n,n' is a binary decision variable; 42) Since the problem has been transformed, set parameters Thus, the objective function of the problem is also transformed; the objective function is repositioned as Among them, X hms is a harmony vector in the initial harmony matrix HM, hms and hvd represent the rows and columns in the harmony matrix respectively, and HVD represents the maximum number of columns; 43) Then, new harmony vectors are generated to continuously optimize the solution space; the generation of harmony vectors is performed using the following two rules: Among them, λ new,hvd represents the newly generated harmony vector, ω1, ω2 and ω3 are random numbers uniformly distributed in the interval [0,1], HM represents the harmony matrix (or harmony memory), HMCR represents the probability of randomly extracting a harmony vector from the harmony matrix, BW represents the amplitude of fine-tuning, and PAR represents the pitch adjustment rate; 44) On the basis of ensuring the feasibility of the new harmonic vector, the initial solution space is updated; the new harmonic vector is evaluated, and if it is better than the harmonic vector in the initial solution space with the worst objective function value, the new harmonic vector is updated to the solution space; then, new harmonic vectors are continuously generated and the solution space is updated until the maximum number of iterations is reached; 45) The optimal harmony vector of the final solution space is the final deployment and scheduling of the shared virtualized network function; due to the limited computing and storage resources of the edge server, not all user requests are completed, that is, the constraint condition (7): where k j,n is a binary variable representing the edge server cl n Whether to process user request u j ; In addition, the running time of the algorithm is fully considered to achieve real-time requirements as much as possible, and the final experimental running time also proves this.

11. An efficient deployment and scheduling system for shared virtualized network functions in edge networks, characterized in that: The efficient deployment and scheduling system for shared virtualized network functions in the edge network adopts the efficient deployment and scheduling method for shared virtualized network functions in the edge network described in any one of claims 1-10 to solve the deployment and scheduling problems of shareable virtualized network functions in mobile edge networks.

Images