Mobile edge computing network resource deployment and load balancing method

Abstract

The invention provides a mobile edge computing network resource deployment and load balancing method, which comprises the following steps that edge servers and users are respectively modeled as independent uniform Poisson point processes, the users are associated with the nearest edge server based on a distance criterion, and the service area of each edge server is represented by a Voronoi unit; considering calculation rate attenuation caused by resource sharing when frequency band multiplexing and edge servers execute tasks in parallel, and establishing a service time delay model; an effective workload index is provided, the service quality and the resource multiplexing gain are considered in a compromise, and the resource utilization rate of the whole network is quantified. Edge server resource deployment and a multi-edge server load balancing strategy are jointly designed, and the effective workload level of the whole network is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wireless communication and digital information technology, and in particular to a mobile edge computing network resource deployment and load balancing method. BACKGROUND

[0002] Mobile edge computing can achieve fast service response by sinking cloud service capabilities to the network edge. It supports users to offload part or all computing tasks to nearby edge servers to obtain related application services. After receiving the offloaded tasks from the users, the edge servers immediately complete the computing processing and feed back the results to the corresponding users. Mobile edge computing significantly reduces the service response time by processing task offloading in the wireless access network, which can guarantee the smooth deployment of emerging applications such as online games, virtual reality, and autonomous driving.

[0003] Generally, mobile edge computing can be integrated into existing ground mobile communication systems, such as 4G / 5G systems. Although this integration is a widely accepted low-cost method to implement mobile edge computing, in this deployment mode, the performance of the mobile edge computing network is highly dependent on the infrastructure resources of the underlying mobile communication system. Analyzing the influence of the infrastructure resources of the underlying mobile communication system on the service performance of the mobile edge computing network is the key to fully exploiting the technical potential of the limited resources of the edge computing network. However, existing research on mobile edge computing server deployment is mostly limited to site selection design, and lacks in-depth research on the above problems.

[0004] In summary, the existing technology has the following problems: how to accurately analyze the influence of the infrastructure resources of the underlying mobile communication system on the mobile edge network, optimize the network resource deployment plan under the condition of considering the service reliability and resource utilization, and introduce multi-edge server cooperation under the given resource deployment to improve the utilization efficiency of limited edge network resources. SUMMARY

[0005] The purpose of the present application is to solve the problem of how to realize server deployment and multi-edge server autonomous cooperation.

[0006] To this end, the embodiments of the present application provide a mobile edge computing network resource deployment and load balancing method, which comprises the following steps:

[0007] Model the network topology as a Poisson point process and define the effective workload;

[0008] According to the full-band multiplexing communication model and the parallel computing model, a communication delay model and a computing delay model are established;

[0009] The effective workload is derived using the communication delay model and the computing delay model;

[0010] Analyze optimal edge server deployment density aiming at deduced effective workload maximization;

[0011] Design multi-edge server load balancing strategy based on the optimal edge server deployment density to improve network resource utilization.

[0012] Specifically, the network topology is modeled by a Poisson point process, and the effective workload is defined, which specifically includes: the edge server and the user are respectively modeled as independent uniform Poisson point processes, the user is associated with the nearest edge server, and the service area of each edge server is represented by a Voronoi cell.

[0013] Specifically, considering the inherent localized service mode of mobile edge computing network, it is assumed that each user is associated with the nearest edge server, and the service area of a single edge server can be modeled as a Voronoi cell, and the expression is: Where X i is the position of the i-th edge server, and Φ s is a uniform Poisson point process spatial model composed of all edge servers.

[0014] Specifically, the communication delay model is:

[0015]

[0016] Where {h Y ~exp(1):Y∈Φ u} represents the power gain of small-scale channel fading, ||·|| represents the Euclidean distance, α>2 represents the path loss coefficient, and B represents the transmission bandwidth; Y j represents the position of the interfering user, represents the power gain of small-scale channel fading of the interfering user Y j ; L represents the number of task bits offloaded by the user.

[0017] Specifically, the computing delay model is:

[0018]

[0019] Where C represents the total CPU required for each computing task, f0 represents the computing frequency allocated when a virtual machine is executed on an edge server, d is the performance degradation coefficient caused by I / O interference; and M represents the number of virtual machines carried on a single edge server.

[0020] Specifically, the multi-edge server load balancing strategy is designed based on the optimal edge server deployment density to improve network resource utilization, which includes:

[0021] Derive the optimal edge server deployment density using effective workload;

[0022] Design a load balancing strategy among multiple servers based on the optimal edge server deployment density.

[0023] The above technical solution has the following beneficial effects:

[0024] This invention studies network resource deployment and load balancing in mobile edge computing networks. It models edge servers and users as independent, uniform Poisson point processes, characterizing the randomness of their spatial distribution, and defines an effective workload index to comprehensively measure service quality and resource utilization. Theoretical analysis proves that, given user distribution density and business requirements, there exists an optimal edge server density that maximizes the effective workload performance index, and this optimal edge server density can be quickly determined through binary search. Given an optimal resource deployment scheme, a collaborative strategy involving the nearest idle edge servers is proposed. This strategy migrates some tasks from overloaded servers to idle edge servers within the collaborative area, simultaneously improving user service quality and overall network resource utilization efficiency. Simulations verify that this scheme achieves higher resource utilization than existing schemes. Attached Figure Description

[0025] Figure 1 This is a flowchart of a method for deploying mobile edge computing network resources and load balancing provided in an embodiment of the present invention;

[0026] Figure 2 This is a performance comparison chart of a mobile edge computing network resource deployment and load balancing method provided in an embodiment of the present invention. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] In embodiments of the present invention, such as Figure 1 This paper provides a method for deploying and load balancing mobile edge computing network resources, the method comprising the following steps:

[0029] S101: Model the network topology using a Poisson point process and define the effective workload; model the spatial distribution of edge servers with a density of λ. s Uniform Poisson point process Φ s Mobile user modeling is related to Φ s The other independent density is λ uUniform poisson point process Φ u . Considering the local service mode of edge computing network, it is assumed that each user is associated with the nearest server, and the service area of the i-th edge server X i ∈Φ s can be modeled as a Voronoi cell, and the expression is: Where X i is the location of the i-th edge server, and X k is any edge server different from X i in the uniform poisson point process Φ s .

[0030] S102: Establish a communication delay model and a computing delay model according to a full-band multiplexing communication model and a parallel computing model;

[0031] S103: Derive the effective workload using the communication delay model and the computing delay model;

[0032] S104: Analyze the optimal edge server deployment density with the goal of maximizing the derived effective workload.

[0033] S105: Based on the optimal edge server deployment density, design a multi-edge server load balancing strategy to improve the overall network resource utilization.

[0034] The network topology modeled by the poisson point process defines the effective workload, which specifically includes: the edge server and the user are respectively modeled as independent uniform poisson point processes, the user is associated with the nearest edge server, and the service area of each edge server is represented by a Voronoi cell.

[0035] Considering the inherent local service mode of mobile edge computing network, it is assumed that each user is associated with the nearest edge server, and the service area of a single edge server can be modeled as a Voronoi cell, and the expression is: Where X i is the location of the i-th edge server, and Φ s is a uniform poisson point process spatial model composed of all edge servers.

[0036] The communication delay model is:

[0037]

[0038] Where {h Y ~exp(1): Y∈Φ u} represents the power gain of small-scale channel fading, ||·|| represents the Euclidean distance, α>2 represents the path loss coefficient, and B represents the transmission bandwidth; Yj denotes the location of the interfering user, denotes the interfering user Y j power gain of small-scale channel fading; L denotes the number of task bits offloaded by the user.

[0039] The calculation delay model is:

[0040]

[0041] wherein C denotes the total CPU required by each computing task, f0 denotes the computing frequency allocated when an edge server only executes one virtual machine, d is the performance degradation coefficient caused by I / O interference; M denotes the number of virtual machines carried on a single edge server.

[0042] The effective workload performance indicator is defined as:

[0043]

[0044] wherein τ is the delay requirement of the computing task, describes the average amount of traffic carried by a single edge server in the network, denotes the total service delay T comp +T comm satisfies the delay requirement τ.

[0045] According to the probability mass function of the random workload and the Poisson point process spatial distribution characteristics of the edge server and the user, the effective workload theoretical expression is derived as:

[0046]

[0047] wherein the shape factor a = 3.575, Γ(·) is the Gamma function, is the average load of the network, is the maximum number of virtual machines that a single edge server can bear at a time determined by the running condition T comp ≤τ.

[0048] The first derivative of the effective workload η with respect to the network average load ζ shows that the effective workload η increases first and then decreases with the network average load ζ. Therefore, it can be proved that there is a unique and optimal edge server density for a given user space distribution and computing task performance requirement, which maximizes the effective workload performance index, and the optimal edge server density varies linearly with the user density. Meanwhile, the optimal edge server density value can be quickly determined by binary search using the change relationship of the effective workload η with the network average load ζ; on the other hand, based on the obtained optimal resource deployment scheme, a nearest idle edge server load balancing strategy is further introduced. Given the optimal network deployment scheme, the edge server deployment density and deployment location usually remain unchanged throughout the service life cycle. However, considering the dynamic characteristics of the computing load carried by each edge server in the network, such as burstiness and randomness, a real-time intra-layer cooperation mechanism needs to be designed for the short-term instantaneous fluctuation characteristics to achieve cross-cell load balancing, thereby improving the overall user service quality in the network and further improving the utilization efficiency of limited resources in the mobile edge computing network.

[0049] In the proposed nearest idle edge server load balancing strategy, when an edge server is overloaded, part of the computing tasks can be migrated to its adjacent idle edge server in proportion, thereby improving the service quality of users and the resource utilization of the whole network. Let the task migration ratio be ω, which can be selected as a fixed value or adaptively adjusted according to the network load at different times and the adjacent available idle resources. Considering the distributed deployment architecture of the mobile edge computing network and the low latency requirement of user tasks, the selected cooperation edge server of the overloaded edge server needs to meet certain distance requirements, that is, the spatial distance between the overloaded edge server and the cooperation edge server does not exceed the maximum distance threshold R c . Given N s edge servers, each edge server carries a set of computing tasks M s , i = 1, …, N i , and the total amount of computing tasks is |M s |. The specific process of the proposed nearest idle edge server cooperation strategy is as follows:

[0050] Step 1: Divide the time slot into N s sub-slots according to the number of edge servers N c in the current service coverage area. Determine the overload threshold Θ of the edge server according to the computing efficiency of the edge server, and give the load distribution ratio coefficient ω of the overloaded server and the geographic area radius R i of multi-server cooperation.

[0051] Step 2: Given a time slot, each edge server polls and broadcasts the current load to surrounding edge servers within its assigned sub-time slot.

[0052] Step 3: In the subsequent time slot, each edge server sequentially performs load migration within its allocated sub-time slot. Specifically, considering all edge servers i = 1, ..., Ns in sequence, if the load currently carried by edge server i exceeds the threshold Θ, i.e., |M i If |>Θ, then the edge server triggers a load migration policy.

[0053] Step 4: Edge server i randomly selects a set of loads to be migrated according to the load splitting ratio coefficient ω. and to set W i User-distributed load migration control signaling corresponding to network load.

[0054] Step 5: Set W i After a user receives a load migration signaling message, if the user's location is the center and the cooperation radius R is... c If other idle edge servers exist within the range, the user randomly selects an edge server from the available idle edge servers in the collaborative area and migrates the computing task to that edge server; otherwise, the computing task data packet is discarded.

[0055] As one implementation method, a mobile edge computing network resource deployment and load balancing method specifically includes:

[0056] I. System Model

[0057] Consider a mobile edge computing network consisting of a large number of edge servers and mobile users. All servers constitute a uniformly labeled Poisson point process Φ. s Its density is λ s The expression is:

[0058]

[0059] Among them, X i This indicates the location of the i-th edge server. This indicates its service area.

[0060] Similarly, the location of mobile users is modeled by another independent uniform Poisson process with density λ. u The expression is:

[0061]

[0062] Among them, Y j Let X represent the location of the j-th mobile user. Without loss of generality, this application applies to a typical active edge server X located at the origin.o The following analysis is performed, the edge server is associated with at least one user to perform task offloading.

[0063] Specifically, each user's computing task can be described by a triple <L, C, τ>, where L represents offloaded data size (bits), C represents computing intensity (CPU cycles), and τ represents deadline requirement (seconds). When a user generates a computing task to be offloaded, considering the local service mode (i.e., the user accesses a nearby edge server to obtain services), the present application assumes that each user is associated with its nearest edge server. At this time, the service area of any edge server can be modeled as a Voronoi cell, expressed as:

[0064]

[0065] The present application adopts a full-frequency multiplexing license-free access strategy to provide conservative evaluation results for the communication performance of the edge computing network. In the configuration of each user being equipped with a single antenna, the communication delay can be represented as:

[0066]

[0067] where {h Y ~ exp(1): Y ∈ Φ u represents the power gain of small-scale channel fading, ||·|| represents the Euclidean distance, α>2 represents the path loss coefficient, B represents the transmission bandwidth; Y j represents the location of an interfering user, represents the location of an interfering user Y j the power gain of small-scale channel fading; L represents the number of bits of the task offloaded by the user.

[0068] To improve the service response speed, the edge server can adopt a parallel computing mode, generate a plurality of virtual machines that are logically isolated from each other for a plurality of computing tasks through a resource virtualization technology, and each virtual machine processes a computing task offloaded by a single user, thereby avoiding additional waiting delay caused by multi-task timing scheduling. However, a plurality of virtual machines on the same edge server share the same underlying physical resources, and the resource sharing between the virtual machines reduces the actual available computing frequency of each virtual machine. According to the measured data, the computing frequency decay phenomenon of each virtual machine can be described as:

[0069] f(M) = f0(1 + d) 1-M

[0070] where f0is the computing resource obtained by the virtual machine when the edge server runs only one virtual machine, and d>0 is a performance degradation factor, is the number of virtual machines concurrently executed on the edge server (i.e., random workloads). When the server executes M virtual machines, the computing rate of each virtual machine will decrease from f(M) to f(M+1) with each additional virtual machine.

[0071] Given the number of CPUs C required for a computing task, the computation latency can be expressed as:

[0072]

[0073] To evaluate the resource utilization of the whole network, the effective workload indicator is defined in this application:

[0074]

[0075] where τ is the latency requirement of the computing task, describes the average amount of traffic carried by a single edge server in the network, represents the total service latency T comp +T comm satisfies the latency requirement τ.

[0076] The effective workload describes the trade-off between the quality of service and the gain of limited resource reuse in the mobile edge computing network. Specifically, when the average workload increases, the amount of computing tasks carried by each edge server on average increases, at which time the resource sharing gain brought by spectrum reuse and parallel computing increases. However, as the workload increases, the transmission interference in the communication phase and the I / O interference in the computing phase both increase, resulting in a significant decline in the service reliability that the edge computing network can support, and the success service probability decreases. Therefore, the effective workload can comprehensively evaluate the resource utilization efficiency and service reliability of the whole network, which is used to guide the optimal resource deployment and optimization design.

[0077] II. Optimal server deployment

[0078] To facilitate subsequent analysis, the theoretical expression of the effective workload indicator is first derived in this application. According to the spatial distribution characteristics of the uniform Poisson point process, the probability mass function of the average number of users served by each edge server is:

[0079]

[0080] where represents the average load of the network, a = 3.575 represents the shape factor, and Γ(·) is the Gamma function.

[0081] Subsequently, according to the effective workload definition and the expressions of the computation latency and the communication latency, the effective workload expression can be obtained as:

[0082]

[0083] wherein, is the single edge server by response time delay requirements T comp ≤τdetermined by the single maximum virtual machine number. The maximum virtual machine number depends on the server configuration and user demand. When the number of computing tasks carried by the edge server at the same time exceeds , all virtual machines will cause service timeout due to insufficient computing resources, resulting in service interruption.

[0084] The theoretical expression of the effective workload η(ζ) can be observed that, in the case of given service capacity and task configuration, the effective workload is only a function of the ratio of user density to edge server density . Therefore, without improving or reducing the processing capacity of the edge server, to maintain the network resource utilization unchanged, the number of newly added edge servers required is proportional to the number of users increased in the service area.

[0085] The first derivative of the effective workload η(ζ) with respect to ζ can be obtained to obtain the change rule of the effective workload with the deployment density of the edge server. For ease of description, define auxiliary functions g(y) and h(y) as:

[0086]

[0087] wherein, y>0.

[0088] Analyzing the first derivative information of the effective workload η(ζ) can be equivalent to analyzing the first derivative information of g(y)·h(y), that is:

[0089]

[0090] wherein,

[0091]

[0092] When y>0, y1=1 and y2>1 can be obtained as two real roots of the equation [g(y)·h(y)]′=0. Therefore, g(y)·h(y) is monotonically increasing in y∈[1,y2] and monotonically decreasing in y∈[y2,+∞]. According to the one-to-one mapping relationship, it can be directly obtained that under the condition of given user density, there is a unique and optimal edge server density to maximize the effective workload performance index.

[0093] From the above analysis process, under the condition of given edge server configuration and user computing task performance requirement, the effective workload η(ζ) is a unimodal function of the network average load ζ. Therefore, the optimal network average load ζ that maximizes the effective workload can be directly obtained by binary search * . After calculating the optimal average workload ζ * , the optimal edge server density under the given user density λ u can be obtained according to In this case, to maximize network resource utilization, the required deployed or activated server density is linearly proportional to the user density, with the proportion being

[0094] III. Load balancing strategy for the nearest idle edge server

[0095] After determining the optimal edge server deployment strategy according to the user distribution density, due to the significant randomness of user spatial distribution, there is a significant difference in the load level of edge servers at different geographic locations in space, causing load imbalance problems. Specifically, edge servers located in hot areas may be easily affected by service interruption due to excessive load currently carried, while edge servers in other areas may be idle due to low load level or no received computing task requests. To balance the computing load distribution in the network range and avoid resource waste, the application can migrate part of the computing load of the overloaded edge server from the overloaded server to the nearest idle edge server. Specifically, given the set of edge servers X i , i = 1, …, N s , to measure whether the edge server is overloaded, the edge server computing efficiency evaluation index is defined as:

[0096]

[0097] where M is the number of computing tasks carried by the edge server, T comp (M) is the computing delay required to execute each computing task.

[0098] To maximize the computing efficiency of the edge server, by analyzing the first derivative information of CR(M) with respect to the number of computing tasks M, it can be obtained that when , the optimal number of computing tasks where round(·) is the rounding operator.

[0099] When the number of tasks carried by any edge server is higher than M *If an edge server is overloaded, a load migration strategy is automatically triggered. The overloaded edge server will migrate a portion of its currently loaded computing tasks to a nearby idle edge server according to a proportional coefficient ω. Given N s There are M edge servers, and the set of computing tasks carried by each edge server is denoted as M. i i = 1, ..., N s The total number of tasks is denoted as |M i The specific process of the proposed nearest idle edge server collaboration strategy is as follows:

[0100] Step 1: Based on the number N of edge servers within the current service coverage area s Divide the time slots into N s Each time slot. The overload threshold Θ = M for the edge server is determined based on its computational efficiency. * Given the load balancing ratio ω for overloaded servers and the geographical radius R of the multi-server collaboration area. c .

[0101] Step 2: Given a time slot, each edge server polls and broadcasts its current load to surrounding edge servers within its assigned sub-time slot.

[0102] Step 3: In the subsequent time slot, each edge server sequentially performs load migration within its assigned sub-time slot. Specifically, considering all edge servers i = 1, ..., Ns in sequence, if the load currently carried by edge server i exceeds the threshold Θ, i.e., |M... i If |>Θ, then the edge server triggers a load migration policy.

[0103] Step 4: Edge server i randomly selects a set of loads to be migrated according to the load splitting ratio coefficient ω. and to set W i User-distributed load migration control signaling corresponding to network load.

[0104] Step 5 set W i After a user receives a load migration signaling message, if the user's location is the center and the cooperation radius R is... c If other idle edge servers exist within the range, the user randomly selects an edge server from the available idle edge servers in the collaborative area and migrates the computing task to that edge server; otherwise, the computing task data packet is discarded.

[0105] Step 6: The idle edge server receives the computing tasks that the user has unloaded, immediately begins processing the tasks, and immediately sends the computing results back to the corresponding user.

[0106] By using the load migration strategy, the user service performance of the access overloaded edge server can be improved, and the service performance of other users is not adversely affected. By adaptively adjusting the values of the load migration ratio ω and the cooperation radius R c , the migration cost can be limited, and the effective workload performance of the whole network can be further improved. As shown in Figure 2 , simulation analysis can see that, compared with the existing least load priority intra-layer cooperative computing migration strategy (Least Sharing load Sharing, LSLS), the nearest neighbor intra-layer cooperative computing migration strategy (Nearest Neighbor, NN), and the random intra-layer cooperative strategy (Random Sharing, RS), the proposed nearest idle edge server load balancing strategy can significantly improve the effective workload by jointly considering the cooperative edge server load and the geographical spatial distribution.

[0107] It should be understood that the specific order or hierarchy of steps in the processes disclosed should not be interpreted as reflecting an exhaustive listing of steps to be executed in the example methods. Based on design preferences, it should be understood that the specific order or hierarchy of steps in the processes could be re-arranged while remaining within the scope of the present disclosure. The accompanying method claims present elements of the various steps in a sample order, and are not meant to be limited to the specific order or hierarchy presented.

[0108] In the above detailed description, various features are grouped together in single embodiments for the purpose of streamlining the disclosure. This disclosed approach is not to be interpreted as reflecting an intention that the claimed embodiments require more features than are explicitly recited in each claim. On the contrary, as reflected by the appended claims, the inventive subject matter is directed to each individual feature individually recited in each claim and every combination of these features. Thus, the following claims are hereby expressly incorporated into this detailed description, with each claim acting as a separate embodiment of the subject matter disclosed by this document.

[0109] The disclosed embodiments have been described above to enable any person skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the disclosure. Thus, the present disclosure is not intended to be limited to the embodiments disclosed herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0110] The above description includes examples of one or more embodiments. Of course, not all possible combinations of components or methods described above will be employed to make or use the embodiments nor will all of the following described examples necessarily be realized. One of ordinary skill in the art, however, having the benefit of the present disclosure will recognize alternatives to the embodiments described herein. Therefore, the present application as described herein is intended to cover all such alternatives, modifications and equivalents as can be included within the spirit of the scope of the appended claims. Furthermore, the use of the term "including" as well as "comprising" as used herein is intended to represent a connotation of "one or more of the recited items" as well as "one or more of the recited items and possible additional items." In addition, any use of the term "or" in the claims is intended to represent a connotation of "non-exclusivity" as that term is used in the patent law.

[0111] Those of skill would further appreciate that the various illustrative logical blocks, modules, and steps described in connection with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans can implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present embodiments.

[0112] The various illustrative logical blocks, modules, and steps described in connection with the embodiments disclosed herein can be implemented or performed by a general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor can be a microprocessor, but in the alternative, the general purpose processor can be any conventional processor, controller, microcontroller, or state machine. A processor can also be implemented as a combination of computing devices, e.g., a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration.

[0113] The steps of a method or algorithm described in connection with the embodiments disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module can reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor such the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium can be integral to the processor. The processor and the storage medium can reside in an ASIC. The ASIC can reside in a user terminal. In the alternative, the processor and the storage medium can reside as discrete components in a user terminal.

[0114] In one or more exemplary designs, the functions described can be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions can be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media include both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. Storage media can be any available media that can be accessed by a general purpose or special purpose computer. By way of example, such computer-readable media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code means in the form of instructions or data structures and that can be accessed by a general-purpose or special-purpose computer, or a general-purpose or special-purpose processor. Also, any connection is properly termed a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or data

[0115] The above detailed description describes the purpose, technical solutions and advantages of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the scope of the present application.

Claims

1. A method for deploying and load balancing mobile edge computing network resources, characterized in that, The method includes the following steps: The network topology is modeled using the Poisson point process, and the effective workload is defined. Based on the full-band multiplexing communication model and the parallel computing model, establish communication delay model and computing delay model; The effective workload is derived using the communication delay model and the computation delay model; With the goal of maximizing the effective workload derived above, we analyze the optimal edge server deployment density; Based on the optimal edge server deployment density, a multi-edge server load balancing strategy is designed to improve the overall network resource utilization.

2. The method for deploying mobile edge computing network resources and load balancing according to claim 1, characterized in that, The network topology modeled using Poisson point processes defines the effective workload, specifically including: edge servers and users are each modeled as independent uniform Poisson point processes, users are associated with the nearest edge server, and the service area of ​​each edge server is represented by Voronoi cells.

3. The method for deploying mobile edge computing network resources and load balancing according to claim 1, characterized in that, Considering the inherent localized service pattern of mobile edge computing networks, assuming each user is associated with the nearest edge server, the service area of ​​a single edge server can be modeled as a Voronoi cell, expressed as: Among them, X i Let Φ be the location of the i-th edge server. s A uniform Poisson point process space model consisting of all edge servers.

4. The method for deploying mobile edge computing network resources and load balancing according to claim 1, characterized in that, The communication delay model is as follows: Among them, {h Y ~exp(1):Y∈Φ u } represents the power gain of small-scale channel fading, ||·|| represents the Euclidean distance, α>2 represents the path loss coefficient, and B represents the transmission bandwidth; Y j This indicates interference with the user's location. Indicates interference with user Y j Power gain for small-scale channel fading; L represents the number of task bits offloaded by the user.

5. The method for deploying mobile edge computing network resources and load balancing according to claim 1, characterized in that, The computational delay model is as follows: Where C represents the total number of CPUs required for each computing task, f0 represents the computing frequency allocated to an edge server when it executes only one virtual machine, d is the performance degradation coefficient caused by I / O interference, and M represents the number of virtual machines hosted on a single edge server.

6. The method for deploying mobile edge computing network resources and load balancing according to claim 1, characterized in that, The design of a multi-edge server load balancing strategy based on the optimal edge server deployment density to improve the overall network resource utilization includes: Derive the optimal edge server deployment density using effective workload; Design a load balancing strategy among multiple servers based on the optimal edge server deployment density.

Images