KR102101120B1 - Resource allocation method for virtualized network function based on user grouping - Google Patents

Abstract

According to the present invention, in a network using a number of virtualized network functions (VNFs) deployed in mobile edge computing (MEC) servers which serve as a base station, provided is a resource allocation method comprising the steps of: analyzing characteristics of a predetermined user terminal and grouping a plurality of user terminals into at least one user group according to the analyzed characteristics of the user terminal; determining the number of clusters such that the average service time of data flow is minimized for each of the at least one user group; clustering a network with at least one cluster number determined for each user group; and placing VNFs required by the user terminals located in each clustered cluster for each user group in the MEC server included in the corresponding cluster.

Description

RESOURCE ALLOCATION METHOD FOR VIRTUALIZED NETWORK FUNCTION BASED ON USER GROUPING}

The present invention relates to a resource allocation method, and relates to a resource allocation method of a cluster-based virtual network function reflecting user characteristics in a network edge environment.

Network virtualization is to build a network through commercial high-performance servers, storage devices, and switches, and to virtually add and manage various network functions necessary for service provision, away from the existing network with a large dependency on dedicated hardware. Network services in network virtualization are provided through a service chain that is a set of a series of virtualized network functions (VNFs), and VNFs are appropriately arranged and used in nodes of a network for network virtualization.

At this time, the network node is supported through Mobile Edge Computing (MEC). MEC makes it possible to use resources in a data center near a base station at a network edge or a radio access network (RAN), and can utilize this resource to deploy VNFs near users.

Indeed, network services are becoming more diverse, and processing of data in the vicinity of users such as shorter latency and high reliability becomes more necessary, whereas the existing VNF deployment technique provides services by large users in a large data center in the center. For this, we mainly propose a resource allocation technique.

The existing method does not consider the use demand of the network using the VNF, and there is a problem in that the user does not consider the needs and individual characteristics of each other service.

Korean Open Patent No. 10-2017-0030295 (released on March 17, 2017)

An object of the present invention is to provide a method for allocating resources of a network clustering-based VNF reflecting user characteristics so as to minimize a service time of a data flow delivered through a service chain in a VNF-based network utilizing resources of a network edge.

The resource allocation method of VNF according to an embodiment of the present invention for achieving the above object is a network using a plurality of VNFs (Virtualized Network Function) deployed in MEC servers serving as a base station, the characteristics of the known user terminal Analyzing and grouping a plurality of user terminals into at least one user group according to the analyzed characteristics of the user terminal; Determining the number of clusters such that the average service time of the data flow is minimized for each of the at least one user group; Clustering the network with a determined number of clusters for each user group; And placing the VNFs required by the user terminals located in each clustered cluster for each user group in the MEC server included in the corresponding cluster. It includes.

The determining the number of clusters may determine the number of clusters such that the number of user terminals moving between clustered clusters is minimized and the number of user terminals included in the cluster is as similar as possible.

The determining of the number of clusters may include determining a set of clusters for each user group such that the average end-to-end delay time of all user groups according to the number of clusters of each of the at least one user group is the minimum, thereby determining the number of clusters of each cluster. You can decide the number.

The resource allocation method includes allocating resources of the network in proportion to the number of user terminals included in each of the at least one user group; Further comprising, the step of deploying to the MEC server may deploy the VNF according to the resources allocated to the user group including the user terminal.

Therefore, the resource allocation method of VNF according to an embodiment of the present invention is grouped in consideration of the characteristics of user terminals in a VNF-based network at a network edge, and by configuring clusters of multiple layers corresponding to the grouped user terminals, the service chain Through this, the service time of the data flow delivered to the user terminal can be minimized.

1 shows a concept of VNF migration according to user movement.
2 shows an example in which a resource allocation device according to the present embodiment clusters a network.
3 illustrates a concept of user grouping and clustering in each group according to user characteristics according to an embodiment of the present invention.
4 shows a schematic structure of a resource allocation device according to an embodiment of the present invention.
5 shows a resource allocation method according to an embodiment of the present invention.

In order to fully understand the present invention, the operational advantages of the present invention, and the objects achieved by the practice of the present invention, reference should be made to the accompanying drawings and the contents described in the accompanying drawings, which illustrate preferred embodiments of the present invention.

Hereinafter, the present invention will be described in detail by explaining preferred embodiments of the present invention with reference to the accompanying drawings. However, the present invention may be implemented in various different forms, and is not limited to the described embodiments. In addition, in order to clearly describe the present invention, parts irrelevant to the description are omitted, and the same reference numerals in the drawings indicate the same members.

Throughout the specification, when a part “includes” a certain component, this means that other components may be further included, rather than excluding other components, unless specifically stated to the contrary. In addition, terms such as "... unit", "... group", "module", and "block" described in the specification mean a unit that processes at least one function or operation, which is hardware or software or hardware. And software.

1 shows a concept of VNF migration according to user movement.

1 shows a network system including two base stations (BS1, BS2) as a simple example, and each base station (BS1, BS2) can operate as a Mobile Edge Computing (MEC) server, and at least one VNF ( By installing the Virtualized Network Function), various virtual services can be provided to the user terminal. Here, the VNF is a virtualized network module such that the node where the VNF is installed (for example, a base station functioning as a MEC) functions as a specific network equipment.

Generally, a network is composed of a large number of network equipment, and the network equipment provides various functions. For example, various functions such as firewall, deep packet inspection (DPI), network address translation (Instusion Detection System), IDS (Wdie Area Acceleration) are provided through network equipment. These network devices are implemented in the form of physical middle boxes, and they are hardware platforms manufactured by specific manufacturers for specific purposes, which are expensive, difficult to maintain, and difficult to upgrade.

In order to solve this problem, a software defined network using network virtualization (SDN) has been proposed. SDN is able to flexibly add and manage various network functions required for service provision by virtualizing it with VNFs, away from the dedicated hardware dependency of the existing network and using general purpose hardware such as commercial high-performance servers, storage devices, and switches. It is to construct a network.

Here, VNF is a module that virtualizes each network function for network virtualization, which is placed on the nodes that make up the network. When a network function is implemented using VNF, not only can a network function be implemented at a relatively low cost, but also a VNF that implements a desired function can be obtained from a server to change, confirm, or reduce the network function. In particular, the deployment of VNFs using MEC servers is essential at the present time as the number of types of services increases and is specialized for individual users.

As shown in FIG. 1, the user terminal UE is the second base station BS2 operating as the second MEC server MEC2 in the service area of the first base station BS2 operating as the first MEC server MEC1. When moving to the service area of the at least one VNF (VNF1, VNF2, VNF3) for providing a service from the first MEC server (MEC1) to the second MEC server (MEC2) according to the movement of the user terminal (UE) The migration is executed.

In SDN, it is possible to provide a service suitable for each individual user terminal (UE) through service chaining of multiple VNFs, and to maintain a quality of experience (QoE) of the user terminal (UE). ), The VNF is transferred.

However, the transfer of the VNF consumes resources of the network such as additional link resources previously consumed and computing resources for determining the location of the VNF deployment, and thus a criterion for the transfer of the VNF is required.

In this embodiment, as a criterion for the transfer of the VNF, a cluster according to the characteristics of the user terminal is configured, and the transfer of the VNF is performed only when the user terminal crosses the boundary of the cluster, so that it is delivered to the user terminal through the service chain. Minimize service time of data flow. That is, while the user terminal is located in a cluster configured with at least one base station, VNF migration is not performed while moving beyond the boundary of the cluster, VNF migration is performed to reduce resource waste.

For the deployment and transfer of VNF, the storage resources, computing resources and link resources of the MEC server are consumed. In this case, computing resources in the deployment and link resources in the past were considered as the main resources. The remaining resources of the MEC server according to the VNF arrangement may be expressed as Equation 1.

는 MEC 서버(m)의 잔여 CPU 자원,

는 MEC 서버(m)의 총 CPU 자원, C_u,i는 사용자 단말(u)의 서비스 체이닝의 i번째 VNF를 MEC 서버(m)에 배치할 때 소모되는 CPU 자원을 의미하며, x_u,i,m은 이진 변수로 사용자 단말(u)의 i번째 VNF가 MEC 서버(m)에 배치되었을 때 1이고, 아니면 0이다.here

Is the remaining CPU resource of the MEC server (m),

Is the total CPU resource of the MEC server (m), C _{u, i} means the CPU resource consumed when the i th VNF of the service chaining of the user terminal (u) is placed in the MEC server (m), x _{u, i , m} is a binary variable and is 1 when the i-th VNF of the user terminal u is placed in the MEC server m, or 0 otherwise.

Meanwhile, the residual resource of the link according to the arrangement of the VNF can be expressed as Equation (2).

는 MEC 서버들(m)과 MEC 서버(m')를 잇는 링크의 잔여 링크 자원,

는 MEC 서버(m)와 MEC 서버(m') 사이의 총 링크 자원, r_u,i는 사용자 단말(u)의 i번째 VNF를 제공하기 위해 필요한 링크 자원, r_mig,u,i는 사용자 단말(u)의 i번째 VNF를 이전할 때 소모되는 링크 자원을 의미한다. y_u,i,m,m'과 z_u,i,m,m'은 이진변수로 각각 사용자 단말(u)의 i번째 VNF를 제공 또는 이전하는 링크로 MEC 서버(m, m')가 선택될 때 1이고, 선택되지 않으면 0이다.In Equation 2

Is the remaining link resource of the link connecting the MEC servers (m) and the MEC server (m '),

Is the total link resource between the MEC server (m) and the MEC server (m '), r _{u, i} is the link resource required to provide the i-th VNF of the user terminal (u), r _{mig, u, i} is the user terminal Refers to the link resource consumed when transferring the i-th VNF in (u). y _{u, i, m, m '} and z _{u, i, m, m'} are binary variables, and the MEC server (m, m ') is selected as a link that provides or transfers the i-th VNF of the user terminal (u), respectively. 1 if it is, or 0 if not selected.

In this embodiment, as an example of the characteristics of the user terminal, the movement speed of the user terminal is used, and accordingly, the user terminals are grouped according to the movement speed of the user terminal, and the number of user terminals included in the user terminal group and server resources proportional thereto. It provides a clustering method to minimize the service time of allocation and data flow. Here, the clustering method means a method of determining the number of clusters and configuring each cluster.

2 shows an example in which a resource allocation device according to the present embodiment clusters a network.

)를 함께 표시하였다. 그리고 n번째 사용자 그룹에서 7개의 기지국(BS1 ~ BS7) 중 2개의 기지국(m, m')을 잇는 링크를 따라 이동하는 사용자 단말의 수(uⁿ _(m,m'))를 링크 상에 표시하였다. 점선으로 표시된 에지 컷(edge cut)을 통해 7개의 기지국(BS1 ~ BS7)이 제1, 제2 및 제7 기지국(BS1, BS2, BS7)를 포함하는 클러스터와 제3 내지 제6 기지국(BS3 ~ BS6)를 포함하는 클러스터로 클러스터링될 수 있다. 여기서 클러스터링 방법은 일예로 기존의 그래프 분할 알고리즘(Graph Partitioning Algorithm)을 이용하여 수행될 수 있으며, 그래프 분할 알고리즘에서도 영역 분할 도구인 METIS 기법을 이용하여 수행될 수 있다.In FIG. 2, as an example, a network in which seven base stations BS1 to BS7 capable of performing a MEC function are arranged, and the number of user terminals of the nth user group connected to each base station BS1 to BS7 (u ⁿ _m

) Together. In addition, the number of user terminals (u ⁿ _{(m, m ')} ) moving along a link connecting two base stations (m, m') among seven base stations (BS1 to BS7) in the nth user group is displayed on the link. Did. A cluster including seven base stations (BS1 to BS7) including first, second, and seventh base stations (BS1, BS2, and BS7) through an edge cut indicated by a dotted line, and third to sixth base stations (BS3 to) BS6) may be clustered. Here, the clustering method may be performed using, for example, an existing graph partitioning algorithm, and the graph partitioning algorithm may also be performed using a region partitioning tool, METIS.

로 계산될 수 있다. 따라서 사용자 단말의 이동에 따른 VNF 이전 횟수 또한

로 계산될 수 있다.As 7 base stations (BS1 to BS7) are clustered into 2 clusters, the number of user terminals moving between clusters (u ⁿ _{(m, m ')} ) is the number of user terminals moving the link located at the boundary of the cluster ( as the sum of u ⁿ _{(m, m ')} ),

Can be calculated as Therefore, the number of VNF transfers according to the movement of the user terminal is also

Can be calculated as

On the other hand, in the case of using the existing method without performing clustering, the number of VNF transfers is calculated as the sum of the number of user terminals moving between all base stations BS1 to BS7.

That is, the number of VNF transfers can be greatly reduced by clustering a plurality of base stations BS1 to BS7 into at least clusters and performing VNF transfer when the user terminal crosses the boundary of the cluster.

3 illustrates a concept of user grouping and clustering in each group according to user characteristics according to an embodiment of the present invention.

In FIG. 3, the network is composed of a plurality of base stations capable of performing the functions of the MEC server, and as an example of the characteristics of the plurality of user terminals in the service area of the network, three user groups (302, 303, 304) according to the movement speed This grouping case is shown. Here, the three user groups 302, 303, and 304, which are classified according to the movement speed of the user terminal, are pedestrian groups 302 configured in response to speeds below a person's walking speed, and vehicles, so they are configured in response to driving speeds such as railways. It was grouped into a vehicle group 204 and a bicycle group 303 configured to correspond to a moving speed using a bicycle or the like between walking speed and driving speed.

In addition, physical resources in the network 301 are allocated in proportion to the number of user terminals included in each of the three user groups 302, 303, and 304. For example, when the ratio of the pedestrian group 302 and the bicycle group 303 and the vehicle group 204 is 5: 3: 2, resources of the network may also be allocated at a ratio of 5: 3: 2. In addition, since the number of clusters for minimizing the service time of the data flow is different according to each user group 302, 303, 304, as shown in FIG. 3, each user group 302, 303, 304 in the network 201 The number of clusters for can be divided differently.

Here, as an example, the speed of one of the users included in each user group is set as a representative speed and used as a representative characteristic of the user group, but the present invention is not limited to this, and reflects various user terminal characteristics such as billing policy, traffic, etc. To set up a user group.

In addition, in the clustering of the network, a plurality of VNFs that may be required differently for each user terminal are all present in the network, and according to the movement of the user terminal, the transfer of VNFs from one base station acting as a MEC server to another base station acting as a MEC server. It is made possible. Particularly, in this embodiment, clustering is performed such that the number of user terminals moving between clusters is minimized and the number of user terminals in each cluster is similar, so that network link resources of VNFs can be minimized by minimizing movement between clusters.

4 shows a schematic structure of a resource allocation device according to an embodiment of the present invention.

Here, as an example, it is assumed that the resource allocation device 410 is implemented as a network function virtualization management and orchestrator (NFV), which is a virtualized integrated management system. NFV MONO supports VNF management and orchestration between the existing network's operations support system / business support system (OSS / BSS) and VNF. NFV MONO, like VNF, can be implemented as a software module and installed and operated on general-purpose hardware provided on the network.

And, for convenience of understanding, FIG. 4 shows a radio access network (RAN) 420 in which a resource allocation device allocates resources together with a structure of a resource allocation device, and it is assumed that the RAN is a 5G RAN. do. In the RAN 420, a plurality of base stations performing MEC functions are disposed, and a plurality of user terminals (not shown) may be located. AMF (Access and Mobility Management Function) 430 is one of the functions provided by the 5G network, and can be obtained by monitoring the movement speed of each of the at least one user terminal located in the RAN. Here, the resource allocation apparatus will be described on the assumption that resources are allocated by considering the movement speed of the user terminal as a characteristic of the user terminal.

Referring to FIG. 4, a resource allocation device implemented with NFV MONO may include an NFV orchestrator 411, a VNF manager 412, and a virtualized infrastructure manager (VIM) 413.

The NFV orchestrator 411 is a configuration for controlling and managing NFV infrastructure and software resources, and is authorized by an Access and Mobility Management Function (AMF) 430 to move each of at least one user terminal located in the RAN, From the VNF manager 412, at least one user terminal is grouped into at least one user group using service request information requested from at least one user terminal, and the number of user terminals in each grouped user group is determined. In addition, the NFV orchestrator 411 clusters each of the at least one user terminal group into at least one cluster such that the number of user terminals moving between clusters is minimized and the number of user terminals in each cluster is similar. Then, the clustering result is transmitted to the VIM 413.

The VIM 413 receives the clustering result and performs clustering on physical and virtual resources in the RAN. The VIM 413 allocates physical and virtual resources to each user group in proportion to the number of user terminals included in each user group. Then, the MEC server suitable for VNF deployment in each cluster is determined, and the VNF deployment server information is transmitted to the VNF manager 412.

The VNF manager 412 performs VNF deployment so that VNFs corresponding to service request information requested by the user terminal are installed on at least one MEC server of each cluster according to the VNF deployment server information transmitted from the VIM 413.

Hereinafter, a process in which the NFV orchestrator 411 groups at least one user terminal into a user group according to characteristics of the user terminal and clusters network resources according to the user group.

Here, it is assumed that the NFV orchestrator 411 groups the user terminals into N user groups according to the characteristics of the user terminal, and one user terminal cannot be grouped into a plurality of user groups.

이고, n번째 사용자 그룹을 위해 k개의 클러스터(k_n)가 존재할 때, 각 클러스터가 담당해야할 범위는 a_n=a/k_n로 계산될 수 있다. 기존의 연구를 참조하면, 클러스터의 경계 이전율(μ_n)은 수학식 3으로 계산될 수 있다.When the total number of user terminals is U, the number of user terminals belonging to the nth user group is U _n , and the number of base stations capable of serving as the MEC server in the network is M, the MEC allocated to the nth user group The number of servers can be said to be M _n = MU _n / U. And when the total service coverage of the network is a, the scope of the MEC server of the nth user group is

, And when there are k clusters (k _n ) for the nth user group, a range to be handled by each cluster may be calculated as a _n = a / k _n . Referring to the existing studies, the cluster boundary transfer rate (μ _n ) can be calculated by Equation 3.

When the service request rate follows the Poisson distribution of λ, and using the cluster boundary transfer rate (μ _n ), the probability (p _{mig, n} ) of each user terminal crossing the cluster boundary is calculated by Equation (4).

When the number of services requested from the n-th user group in the network (S _n ) is S _n = SU _n / U, the number of services (S _{mig, n} ) required for VNF transfer for the n-th user group is mathematical It is calculated by Equation 5.

Here, p _n is a probability that the user terminal u belongs to the n-th user group, and p _n = U _n / U.

According to Equation (5) _, the number of services (S _{mig, n} ) required for VNF migration tends to increase as the representative speed of the user group (v _n ) is faster and the number of clusters (k _n ) is larger.

When the set of the number of clusters per user group is k = (k ₁ , k ₂ , ..., k _N ), and the service time for the user terminal is T, the number of clusters per user group (k _n ) is The total service delay time (T (k)) summing the service time (T (k _n )) for each user group may be determined to be minimized (minimize T (k)) while satisfying the four constraints of Equations 6 to 9. have.

The first constraint in Equation 6 means that all user terminals can be included in only one user group.

The second constraint of Equation (7) means that the number of clusters (k _n ) for the nth user group is one or more, and is less than or equal to the total number of MEC servers (M _n ) that can be allocated to the nth user group.

)보다 클 수 없음을 의미하고, 네번째 제약조건은 사용자 단말(u)을 위한 VNF를 이전할 때 소모되는 링크 자원은 MEC 서버(m)와 서버(m') 사이의 총 링크 자원(

)을 초과할 수 없음을 의미한다.In addition, the third and fourth constraints of Equations 8 and 9 are constraints derived from Equations 1 and 2, respectively, and the third constraint is the resource of the MEC server m that the user terminal u can use is the MEC server. (m) total resources (

), And the fourth constraint is the link resource consumed when migrating the VNF for the user terminal (u), the total link resource between the MEC server (m) and the server (m ').

).

Meanwhile, influx traffic generated in addition to the link between the MEC server (m) and the server (m ') by VNF transfer flows in the network with self-similar characteristics, and the average inflow traffic is E [r _{u, i} ] = r bps, the maximum link delay time between the MEC server m and the MEC server m 'may be calculated by Equation (10).

이고, H는 자기유사성 정도를 의미하는 허스트(Hurst) 파라미터, σ는 데이터 플로우의 표준편차, ε은 자기유사성의 분수 브라운 운동(fractional Brownian motion: fBm) 모델의 오버플로우(overflow) 확률을 의미한다.here

Where H is the Hurst parameter indicating the degree of self-similarity, σ is the standard deviation of the data flow, and ε is the probability of overflow of the fractional Brownian motion (fBm) model of self-similarity. .

)으로부터 수학식 11과 같이 계산될 수 있다.Therefore, the total service delay time (T (k)) is the service time of each user group (

) From Equation (11).

At this time, Path _u is a path for delivering the number of services (S _u ) required by each user terminal, and U _n represents an n-th user group.

이라고 한다. 또한 각 사용자 그룹에서 VNF의 이전을 완료하기 위한 평균 경로의 길이는 사용자 그룹에 대한 클러스터 개수(k_n)에 따른 평균 홉 수(

)와 같다고 가정할 수 있다.The average number of VNFs required to provide one service is called f, and the average link resource consumed to move one VNF is called r _mig , and the total number of services that need to be transferred from N total user groups

It is said. Also, the average path length to complete the migration of VNFs in each user group is the average number of hops according to the number of clusters (k _n ) for the user group (

).

)은 수학식 12에 따라 계산될 수 있다.Accordingly, the average available link resource from Equations 2 and 5 (

) May be calculated according to equation (12).

)을 수학식 13으로 계산할 수 있다.And the average end-to-end delay time approximating the total service delay time (T (k)) using equations 11 and 12

) Can be calculated by Equation 13.

는 클러스터 개수(k_n)에 따른 MEC 서버 간 평균 링크 지연 시간이다.)(here

Is the average link latency between MEC servers according to the number of clusters (k _n ).)

)가 최소가 되도록, 즉 데이터 플로우의 평균 서비스 시간이 최소화되도록 하는 클러스터 개수 집합(k)은 수학식 14가 수렴할 때까지 반복 실행할 수 있다.Equation 10 can be solved by the projected gradient descent method, and the average end-to-end delay time (

), I.e., the cluster number set k that minimizes the average service time of the data flow can be repeated until Equation 14 converges.

At this time, w = [w ₁ , w ₂ , ..., w _N ] is a set of weights for each component, which can be determined according to a network policy.

Thereafter, the NFV orchestrator 411 clusters the network resources allocated for each user group by using the number of clusters (k _n ) determined for each user group. That is, the number of user terminals moving between clusters is minimized to perform clustering so that network link resources consumed by the transfer of VNFs are minimized.

5 shows a resource allocation method according to an embodiment of the present invention.

Referring to FIGS. 1 to 4, when the resource allocation method of FIG. 5 is described, first, characteristics of a preset user terminal are analyzed (S10). The characteristics of the user terminal may be variously set, but for example, the movement speed of the user terminal may be set as the characteristics of the user terminal. And the characteristics of the user terminal to be analyzed may be set or changed by a network policy or a network administrator. In some cases, characteristics of a plurality of different user terminals are set, and weights may be assigned to the characteristics of each user terminal to be analyzed.

In addition, the characteristic analysis of the user terminal may be repeatedly performed when a predetermined reference unit of time or a tendency of the user terminal characteristic is changed, and may be performed in response to a command of a network administrator.

When the characteristics of the user terminal are analyzed, a plurality of user terminals in the network are grouped into at least one user group based on the analyzed characteristics of the user terminal (S20). Then, the number of user terminals included in each of the grouped at least one user group is determined (S30).

When the number of user terminals per user group is determined, network resources are allocated to each user group according to the determined number of user terminals per user group (S40). Here, the network resources may be allocated in a proportion proportional to the number of users included in the user group.

When network resources for each user group are allocated, the number of clusters for clustering each of the at least one user group is determined (S50). Here, the number of clusters per user group is similar to the number of user terminals in each cluster in the user group, and the number of user terminals moving between clusters is determined to be minimized, so that the inter-cluster movement of the VNF is minimized, so that the average service time of data flow is Try to minimize it.

이 최소화되도록 하여, 수학식 14에 따라 획득될 수 있다.The cluster number of specific groups of users in this example (k _n) is the average of each group of users sum total service delay time (T (k)) Equation (13) the approximation of the service delay time (T (k _n)) in the End-to-end delay time

By minimizing this, it can be obtained according to Equation (14).

When the number of clusters per user group (k _n ) is determined, the network is clustered according to the determined number of clusters (k _n ). At this time, the network may be divided into a cluster of a number (k _n ) determined based on a graph segmentation algorithm, for example.

In addition, when a cluster is clustered for each user group, at least one VNF for providing a service required by a user terminal to each of the at least one cluster for each user group is disposed in one designated MEC server in the cluster. This is because, when passing through multiple MEC servers, additional delays may occur due to conversion of optical / electrical / optical signals.

The method according to the present invention can be implemented as a computer program stored in a medium for execution on a computer. Computer readable media herein can be any available media that can be accessed by a computer, and can also include any computer storage media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, and ROM (readable) Dedicated memory), RAM (random access memory), CD (compact disk) -ROM, DVD (digital video disk) -ROM, magnetic tape, floppy disk, optical data storage, and the like.

Although the present invention has been described with reference to the embodiments shown in the drawings, these are merely exemplary, and those skilled in the art will understand that various modifications and other equivalent embodiments are possible therefrom.

Therefore, the true technical protection scope of the present invention should be defined by the technical spirit of the appended claims.

Claims (5)

In a network using a number of VNF (Virtualized Network Function) deployed in MEC servers that serve as a base station,
Analyzing the characteristics of the known user terminal and grouping a plurality of user terminals into at least one user group according to the analyzed characteristics of the user terminal;
Determining the number of clusters such that the average service time of the data flow is minimized for each of the at least one user group;
Clustering the network with a determined number of clusters for each user group; And
Disposing a VNF required by a user terminal located in each clustered cluster for each user group in a MEC server included in the corresponding cluster; Resource allocation method comprising a. The method of claim 1, wherein determining the number of clusters
A resource allocation method for determining the number of clusters such that the number of user terminals moving between clustered clusters is minimized and the number of user terminals included in the cluster is as similar as possible. The method of claim 2, wherein determining the number of clusters
Average end-to-end delay time of all user groups according to the number of clusters (k _n ) of each of the at least one user group (n = {1, 2, ..., N})

Resource allocation method for determining the number of clusters by determining the set (k = {k ₁ , k ₂ , ..., k _N }) of the number of clusters per user group to minimize). The method of claim 3, wherein the average end-to-end delay time (

)silver
Equation

(here

Is the average number of hops depending on the number of clusters (k _n )

)ego,

Is the average link latency between MEC servers according to the number of clusters (k _n ), and the maximum link latency between MEC servers (m, m ')

It is calculated as the average of, p _n is the probability that the user terminal (u) belongs to the n-th user group is p _n = U _n / U. And,

Where H is the Hurst parameter indicating the degree of self-similarity, σ is the standard deviation of the data flow, and ε is the probability of overflow of the fractional Brownian motion (fBm) model of self-similarity. . Also, f is the average number of VNFs required to provide one service,

Is the average available link resource, and r is the average incoming traffic following VNF transfer.)
Is calculated according to,
The average end-to-end delay time (

The set of number of clusters per user group (k = {k ₁ , k ₂ , ..., k _N }) that minimizes) is
Equation

(Where w is a set of weights (w = [w ₁ , w ₂ , ..., w _N ]) that are specified in advance according to the network policy.)
Resource allocation method calculated according to. The method of claim 1, wherein the resource allocation method
Allocating resources of the network in proportion to the number of user terminals included in each of the at least one user group; Further comprising,
The step of deploying to the MEC server
Resource allocation method for deploying the VNF according to the resources allocated to the user group containing the user terminal.

Images