Edge server power control method based on primal duality

Abstract

The present invention discloses a method for controlling edge server power based on primal duality, which obtains a set of candidate signal coverage areas according to information of edge servers and terminal devices in an edge computing system, obtains information of each candidate signal coverage area at the same time, constructs a dual programming problem and solves the signal coverage area of ​​each edge server, thereby determining the power of each edge server. The present invention realizes power control of edge servers by designing primal dual programming for application scenarios with limited edge server resources.

Description

Technical Field

[0001] The present invention belongs to the field of edge computing technology, and more specifically, relates to an edge server power control method based on primal duality. Background Art

[0002] With the advent of the Internet of Everything era, the linear growth of centralized cloud computing capabilities can no longer match the explosive growth of massive edge data. Single computing resources based on the cloud computing model can no longer meet the real-time, security and low energy consumption requirements of data processing. In 2014, mobile edge computing (MEC) came into being. As an important part of the 5G mobile network architecture, MEC sinks computing and resources to the edge of the network so that users can get a better quality of experience. MEC is a new computing model. From the definition of MEC, it can be seen that it is a supplement to cloud computing and provides a better computing platform for mobile computing, the Internet of Things, etc.

[0003] In the original centralized cloud computing model, the terminal device (TD) transmits its tasks to the cloud center through the Internet for processing. After MEC was proposed, the computing tasks of TD can be offloaded to the edge server (ES) with computing resources adjacent to it for processing. In the process of ES providing computing offloading services for TD, signal coverage, data transmission, task execution, etc. will consume energy. In the ten years from 2010 to 2020, the amount of data transmitted on the Internet has increased tenfold, and the energy used for data transmission has increased at the same rate. Therefore, how to reduce energy consumption during data transmission has always been a research focus of academia.

[0004] Generally speaking, data is transmitted between TD and ES in an MEC environment through wireless communication technology. This requires the signal formed by the ES to cover the TD and establish a connection between the two. Due to the attenuation of the signal, an ES needs to provide it with greater power to form a larger signal coverage range. However, in the same time period, greater power consumes more energy. How to reduce the power used by ES for signal coverage as much as possible in an MEC environment while considering the limited resources of ES and provide computing offloading services for all TDs is an urgent problem to be solved. In this process, the management decisions of the system should include: 1) What is the power size to be set for each ES; 2) Which TDs an ES provides computing offloading services for.

[0005] However, the above joint optimization of power control for capacity-constrained ESs is extremely challenging. The following problems need to be overcome:

[0006] 1) In the process of power control for each ES, the goal must be to minimize the total power of all ESs. Both ES and TD facilities in the system are distributed on a plane. According to the relationship between ES power and the radius of the signal coverage area it forms, the signal coverage area formed by ES must cover all TDs. This problem can be abstracted as a geometric set covering problem, which is a rather complex combinatorial optimization problem.

[0007] 2) In model construction, we must fully consider that the computing resources of each ES are limited, that is, the number of TDs that a certain ES can serve is limited; this will further increase the complexity of the optimization problem.

[0008] At present, all related studies mainly focus on signal interference between ESs or quality of service (QoS) for power control, while ignoring the location relationship between various facilities in the MEC environment and the resource limitations of ESs. Summary of the invention

[0009] The purpose of the present invention is to overcome the shortcomings of the prior art and provide an edge server power control method based on primal duality, designing primal duality planning for application scenarios with limited edge server resources, thereby realizing power control of the edge server.

[0010] In order to achieve the above-mentioned object of the invention, the edge server power control method based on the primal duality of the present invention comprises the following steps:

[0011] S1: The edge server set S in the edge computing system = {s1,s2,...,s M}, where s i represents the i-th edge server, i = 1, 2, ..., M, M represents the number of edge servers, and obtains the resource capacity k of each edge server i ; Let the terminal device set U in the edge computing system be U = {u1,u2,...,u N}, where u j represents the jth terminal device, j = 1, 2, ..., N, N represents the number of terminal devices;

[0012] S2: For each edge server s i , record the edge server s i is the center of the circle and the terminal device u j The signal coverage area with the boundary as D ij , construct a signal coverage area set containing M×N signal coverage areas At the same time, according to each edge server s i and each terminal device u j The position coordinates of each signal are calculated to obtain the coverage area D of each signal ijThe radius R ij , the edge server s is calculated according to the preset method i To maintain signal coverage area D ij The power required p ij ;

[0013] S3: Solve the following dual programming problem [D] to obtain the power of each edge server:

[0014]

[0015]

[0016]

[0017]

[0018]

[0019] Among them, θ h , μ i , β ij , γ h,ij are dual programming variables, θ h Indicates that the edge server covers the terminal device u h Available compensation, μ i represents the penalty for edge servers selecting two or more signal coverage areas, both are 0, φ ij Indicates that it is located in the signal coverage area D ij The collection of terminal devices within range, β ij Indicates that when the terminal device u h Located at the terminal device set φ ij But edge servers i The resources are not enough to cover the terminal device u h Terminal device u h The price to be paid, γ h,ij Indicates that when the terminal device u h Located at the terminal device set φ ij And edge servers i The resources are sufficient to cover the terminal device u h Terminal device u h The actual price paid.

[0020] The present invention is based on the primal-dual edge server power control method, obtains a set of candidate signal coverage areas according to the information of edge servers and terminal devices in the edge computing system, and simultaneously obtains the information of each candidate signal coverage area, constructs a dual programming problem and solves the signal coverage area of ​​each edge server, thereby determining the power of each edge server. The present invention realizes power control of edge servers by designing primal-dual programming for application scenarios with limited edge server resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flowchart of a specific implementation method of the edge server power control method based on the primal duality of the present invention;

[0022] Figure 2 is a flow chart of dual programming solution in this embodiment;

[0023] Figure 3 It is a schematic diagram of the edge computing system in this embodiment. DETAILED DESCRIPTION

[0024] The specific implementation of the present invention is described below in conjunction with the accompanying drawings so that those skilled in the art can better understand the present invention. It should be noted that in the following description, when the detailed description of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted here.

[0025] Example

[0026] In order to better illustrate the technical solution of the present invention, the derivation process of the optimization problem of the present invention is first briefly described.

[0027] Let the edge server set S in the edge computing system be {s1,s2,...,s M}, where s i represents the i-th edge server, i=1,2,...,M, M represents the number of edge servers, and the terminal device set U={u1,u2,...,u N}, where u j represents the jth terminal device, j=1,2,...,N, and N represents the number of terminal devices.

[0028] Generally speaking, the resource capacity of the entire edge computing system is sufficient to meet the needs of all terminal devices within the current range. The problem to be solved by the present invention is how to provide a power allocation so that the signal coverage area formed by all edge servers in the edge computing system that are limited by resource capacity can cover all terminal devices and meet their resource needs, so that the sum of the power of all edge servers is minimized. The signal coverage area of ​​the edge server is a circular area with the edge server as the center (which can be regarded as a signal disk). If the terminal device u j On edge servers i The signal coverage area is called the edge server s i "Includes" terminal device u j ; If the terminal device u j On edge servers i The signal coverage area is formed, and the edge server s i Allocate its resources to terminal device u j , then the edge server s i "Covered" terminal device u j .

[0029] The signal coverage area is determined by the center (edge ​​server) and the radius, so each edge server s i and each terminal device u j A signal coverage area D can be determined ij , at this time the radius of this area is exactly equal to the edge server s i and terminal device u j The distance between the terminal devices u j Just located in signal coverage area D ij Therefore, there are a total of M×N optional signal coverage areas to consider, and the set of these signal coverage areas is recorded as For edge servers i The set of signal coverage areas with centered at A subset of Indicated by φ ij Indicates that it is located in the signal coverage area D ij The collection of terminal devices within range, represented by r ij and p ij To represent the signal coverage area D ij The radius and the power required to maintain this radius are expressed as k i Represents edge servers i The resource capacity (i.e. the maximum number of terminal devices that can be covered). One point that needs to be emphasized here is that the terminal device u h (u h ∈U) is the area covered by the signal Dij Containment means that the terminal device is in the signal coverage area D ij Within the range (i.e. u h ∈D ij );u h Edge Servers i Coverage refers to u h Occupancy i A resource of the edge server is connected to the edge server, while the former does not occupy the resources of the corresponding edge server.

[0030] The goal of the present invention is to obtain a power decision scheme under the condition of limited edge server capacity, so that the signal coverage area set formed by the scheme can cover all terminal devices and minimize the total power. (Right now Is a collection A subset of ) needs to meet the following conditions:

[0031] (1) Full coverage conditions. To represent the signal coverage area set All terminal devices that can be covered, then The condition must be met;

[0032] (2) Limited capacity condition. Since the resource capacity of each edge server is limited, the number of terminal devices allocated to a certain edge server must be less than its capacity;

[0033] (3) Power uniqueness. For any edge server s i ∈S, it has a total of N power strategies (i.e., N signal coverage areas), but at most only one strategy is selected, i.e., the set There is at most one signal coverage area with i as the center.

[0034] According to the above constraints, the planning model of the present invention can be established. Set the variable x ij ∈{0,1} represents the signal coverage area D ij Whether it is selected, that is, when x ij =1 when signal coverage area D ij is selected when x ij = 0 when signal coverage area D ij Not selected. Set variable y h,ij ∈{0,1} represents user u h ∈U Whether the signal covers area D ij Cover, when y h,ij =1 indicates that user u h Signal coverage area D ij Cover, and at this time x ij=1, when y h,ij =1 indicates that user u h Area not covered by signal D ij Cover. Thus, the following integer programming problem [P] is constructed:

[0035]

[0036]

[0037]

[0038] x ij ≥y h,ij ,u h ∈φ ij

[0039]

[0040]

[0041] By relaxing the variables x and y, we can get the linear programming form of the plan [P], that is, the original plan. In this way, we can get the dual plan [D] of the original plan as follows:

[0042]

[0043]

[0044]

[0045]

[0046]

[0047] Among them, θ h , μ i , β ij , γ h,ij are dual programming variables, θ h Indicates that the edge server covers the terminal device u h Available compensation, μ i represents the penalty for the edge server to select two or more signal coverage areas. However, since the edge server in the present invention only selects one signal coverage area, μ i Both are 0, β ij Indicates that when the terminal device u h Located at the terminal device set φ ij But edge servers i The resources are not enough to cover the terminal device u h Terminal device u h The price to be paid, γh,ij Indicates that when the terminal device u h Located at the terminal device set φ ij And edge servers i The resources are sufficient to cover the terminal device u h Terminal device u h The actual price paid.

[0048] The primal-dual method is an important mathematical tool in operations research that can solve many complex problems. The main idea is to continuously add dual variables and select the object that corresponds to a dual constraint when it is tightened in the iteration process. Therefore, by solving the above dual planning problem, the power of each edge server can be determined.

[0049] Based on the above derivation process, the edge server power control method based on primal duality of the present invention can be obtained. Figure 1 FIG. 1 is a flow chart of a specific implementation method of the edge server power control method based on the primal duality of the present invention. Figure 1 As shown, the specific steps of the edge server power control method based on the original duality of the present invention include:

[0050] S101: Obtain system parameters:

[0051] Let the edge server set S in the edge computing system be {s1,s2,...,s M}, where s i represents the i-th edge server, i = 1, 2, ..., M, M represents the number of edge servers, and obtains the number of resources k of each edge server i ; Let the terminal device set U in the edge computing system be U = {u1,u2,...,u N}, where u j represents the jth terminal device, j=1,2,...,N, and N represents the number of terminal devices.

[0052] S102: Obtain a set of candidate signal coverage areas:

[0053] For each edge server i , record the edge server s i is the center of the circle and the terminal device u j The signal coverage area with the boundary as D ij , construct a signal coverage area set containing M×N signal coverage areas At the same time, according to each edge server s i and each terminal device u j The position coordinates of each signal are calculated to obtain the coverage area D of each signal ij The radius R ij, according to the preset method, the radius R ij Calculate the edge server s i To maintain signal coverage area D ij The power required p ij .

[0054] In this embodiment, the power p of the edge server is calculated as p=c·R·α, where R represents the radius of the signal coverage area, c and α≥1 are two preset constant parameters, and α is usually called an attenuation factor.

[0055] S103: Power control based on dual programming:

[0056] Solve the following dual programming problem [D] to obtain the power of each edge server:

[0057]

[0058]

[0059]

[0060]

[0061]

[0062] The solution method of the dual programming problem can be selected according to actual needs, but the traditional solution method is time-consuming, so this embodiment also proposes a solution method of the dual programming. Figure 2 is the flow chart of the dual programming solution in this embodiment. Figure 2 As shown, the specific steps of solving the object planning in this embodiment include:

[0063] S201: Initialize dual variables:

[0064] All dual variables θ h , μ i , β ij , γ h,ij All are initialized to 0.

[0065] S202: Initialize the terminal device set:

[0066] For each signal coverage area D ij , get the terminal device set φ included in its coverage ij , initialize the signal coverage area D ij The corresponding terminal device set V ij =φ ij , initialize the undetermined terminal device set U′=U.

[0067] S203: Increase the duality parameter:

[0068] For each terminal device u in the currently undetermined terminal device set U′ h , let its reward θ h =θ h +Δ θ , Δ θ Indicates the preset reward growth step, which can be set according to actual needs.

[0069] Then for each current terminal device set V ij , find the set A ij =V ij ∩U′, that is, the set A ij Contains the terminal device set V ij If the terminal device is not determined in the set A ij The number of terminal devices in is greater than the edge server s i The resource capacity k i , then the required price β ij =β ij +Δ β , Δ β Represents the preset growth step of the required payment cost, otherwise let the terminal device set V ij Each terminal device u h The corresponding actual payment cost γ h,ij =γ h,ij +Δ γ , Δ γ Indicates the growth step of the preset actual payment cost.

[0070] S204: For each signal coverage area D ij Perform power constraint judgment, that is, determine whether If there is signal coverage area D ij If the power constraint is judged to be established, the process proceeds to step S205, otherwise, the process returns to step S203.

[0071] S205: Determine a signal coverage area:

[0072] Signal coverage area D determined by power constraints ij Select any signal coverage area as the signal coverage area D determined in this round ab , let the edge server s a The actual signal coverage area L a Signal coverage area D ab , let the edge server s a The covering terminal set F a V is the terminal device set ab .

[0073] S206: Update terminal device set:

[0074] The terminal device set V ab The terminal devices in are deleted from the current undetermined terminal device set U′.

[0075] S207: Update the signal coverage area set:

[0076] Gather in the current signal coverage area In the a The center of the circle is the radius less than or equal to the signal coverage area D ab The radius R ab That is, the signal coverage area of ​​the edge server s a The center of the circle is greater than R. ab The signal coverage area will further participate in the judgment, then the edge server s a The signal coverage area L a May be updated.

[0077] S208: Update service terminal device set:

[0078] For the current signal coverage area set Each signal in covers area D ij , in its corresponding service terminal device set V ij The terminal device set V ab Delete the terminal device in.

[0079] S209: Determine whether the current undetermined terminal device set U′ is empty. If not, return to step S203; otherwise, proceed to step S210.

[0080] S210: Determine edge server power:

[0081] According to each edge server s i The final actual signal coverage area L i The radius of the edge server s is calculated according to the preset method i To maintain signal coverage area L i The power required p i , which is the edge server s i The current power that should be set.

[0082] In order to better illustrate the technical solution of the present invention, a specific example is used to simulate and verify the present invention. Figure 3 Schematic diagram of the edge computing system in this embodiment. Figure 3As shown, this embodiment includes 5 edge servers (ES) and 20 terminal devices (TD), and the resource capacity of each edge server is 5, so the total capacity of the system is 25. Figure 3 The signal coverage area drawn by the dotted line is temporarily selected for each edge server during the implementation of the present invention, and the signal coverage area drawn by the solid line is the signal coverage area finally determined by each edge server. 3,10 The terminal device 10 is selected first and is assigned to the edge server 3 for coverage. 4,19 , D 5,7 , D 3,8 , D 1,4 , D 2,5 , D 4,20 and D 2,16 For a terminal device that is simultaneously covered by multiple signal coverage areas, it will be covered by the signal coverage area selected first. For example, terminal device 10 will be covered by D 3,8 Cover instead of D 4,20 . Then the final coverage result is: The power of the edge server can be calculated based on the radius of each actual signal coverage area.

[0083] Although the above describes the illustrative specific embodiments of the present invention to facilitate those skilled in the art to understand the present invention, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations using the concept of the present invention are protected.

Claims

1. A primal-dual based edge server power control method, characterized in that: The following steps are involved: S1: The edge server set S in the edge computing system = {s1,s2,...,s M }, where s i represents the i-th edge server, i = 1, 2, ..., M, M represents the number of edge servers, and obtains the resource capacity k of each edge server i ; Let the terminal device set U in the edge computing system be U = {u1,u2,...,u N }, where u j represents the nth terminal device, j = 1, 2, ..., N, N represents the number of terminal devices; S2: For each edge server s i , record the edge server s i is the center of the circle and the terminal device u j The signal coverage area with the boundary as D ij , construct a signal coverage area set containing M×N signal coverage areas At the same time, according to each edge server s i and each terminal device u j The position coordinates of each signal are calculated to obtain the coverage area D of each signal ij The radius R ij , the edge server s is calculated according to the preset method i To maintain signal coverage area D ij The power required p ij ; S3: Solve the following dual programming problem [D] to obtain the power of each edge server: [D]： Among them, θ h , μ i , β ij , γ h,ij are dual programming variables, θ h Indicates that the edge server covers the terminal device u h Available compensation, μ i represents the penalty for edge servers selecting two or more signal coverage areas, both are 0, φ ij Indicates that it is located in the signal coverage area D ij The collection of terminal devices within range, β ij Indicates that when the terminal device u h Located at the terminal device set φ ij But edge servers i The resources are not enough to cover the terminal device u h Terminal device u h The price to be paid, γ h,ij Indicates that when the terminal device u h Located at the terminal device set φ ij And edge servers i The resources are sufficient to cover the terminal device u h Terminal device u h actual consideration paid; The dual programming problem is solved by the following method: S3.1: All dual variables θ h , μ i , β ij , γ h,ij All are initialized to 0; S3.2: For each signal coverage area D ij , get the terminal device set φ included in its coverage ij , initialize the signal coverage area D ij The corresponding terminal device set V ij =φ ij , initialize the undetermined terminal device set U′=U; S3.3: For each terminal device u in the currently undetermined terminal device set U′ h , let its reward θ h =θ h +Δ θ , Δ θ Indicates the preset reward growth step; Then for each current terminal device set V ij , find the set A ij =V ij ∩U′, if the set A ij The number of terminal devices in is greater than the edge server s i The resource capacity k i , then the required price β ij =β ij +Δ β , Δ β Represents the preset growth step of the required payment cost, otherwise let the terminal device set V ij Each terminal device u h The corresponding actual payment cost γ h,ij =γ h,ij +Δ γ , Δ γ Indicates the growth step of the preset actual payment price; S3.4: For each signal coverage area D ij Perform power constraint judgment, that is, determine whether If there is signal coverage area D ij If the power constraint is established, the process goes to step S3.5, otherwise it returns to step S3.3; S3.5: Signal coverage area D determined from power constraints ij Select any signal coverage area as the signal coverage area D determined in this round ab , let the edge server s a The actual signal coverage area L a Signal coverage area D ab , let the edge server s a The covering terminal set F a V is the terminal device set ab ; S3.6: Set the terminal device V ab The terminal devices in are deleted from the current undetermined terminal device set U′; S3.7: Gather in the current signal coverage area In the a The center of the circle is the radius less than or equal to the signal coverage area D ab The radius R ab The signal coverage area is deleted; S3.8: For the current signal coverage area set Each signal in covers area D ij , in its corresponding service terminal device set V ij The terminal device set V ab Delete the terminal device in; S3.9: Determine whether the current undetermined terminal device set U′ is empty. If not, return to step S3.3; otherwise, proceed to step S3.10; S3.10: According to each edge server s i The final actual signal coverage area L i The radius of the edge server s is calculated according to the preset method i To maintain the actual signal coverage area L i The power required p i , which is the edge server s i The current power that should be set.

2. The edge server power control method according to claim 1, characterized in that: The power p of the edge server is calculated as p=c·R·α, where R represents the radius of the signal coverage area, and c and α≥1 are two preset constant parameters.

Images