KR102033339B1 - Method and apparatus for controlling multi-resource infrastructures - Google Patents

Abstract

Embodiments of the present invention define a set of candidate paths for each service in a network environment including a plurality of physical nodes and a plurality of links where networking resources and processing resources coexist, and a plurality of highly efficient candidate paths among the defined candidate paths. By providing a service by selecting the path of the present invention, the present invention provides a method and apparatus for managing a multi-resource infrastructure that can reduce system cost while improving service quality.

Description

METHOD AND APPARATUS FOR CONTROLLING MULTI-RESOURCE INFRASTRUCTURES}

Embodiments of the present invention relate to a method and apparatus for managing a multi-resource infrastructure, and more particularly, includes network function virtualization including general networking services in a multi-resource infrastructure environment including networking resources and processing resources. A method and apparatus for managing NFV) services and computation offloading (CO) services.

The following description merely provides background information related to the embodiments according to the present invention and does not constitute a prior art.

In a conventional network environment, network functions such as deep packet inspection (DPI), intrusion detection system (IDS), firewall and charging are only implemented on special hardware. As a result, services that require that network function must go through hardware that supports that function.

1 is a conceptual diagram illustrating a conventional network routing environment.

For example, in a cellular network, since all packets must be charged, all data packets must pass through a packet data network gateway (P-GW) that supports charging.

Recently, however, with the introduction of virtualization technology, various services using a networking resource and a processing resource are possible. One of them is network function virtualization (NFV) service.

2 is a conceptual diagram illustrating a conventional NFV service routing environment.

NFV has virtualized these network functions based on software so that they can be implemented on various nodes. Network function virtualization has the advantage of increasing resource efficiency, manageability, scalability and applicability by using network service chaining. The leading US network operators AT & T and Verizon have also begun applying NFV technology for network operations in 2016.

Another example of using networking and processing resources together is a computation offloading (CO) service.

3 is a conceptual diagram illustrating a conventional CO service routing environment.

Examples of computation functions include transcoding, audio and image processing. In the past, the compute functionality required by the end host had to be handled directly on the host node, but now CO services can be used to offload the compute capability to a nearby edge cloud or a remote remote cloud.

This technology allows end hosts to reduce the energy consumption and processing delays of their nodes and to improve the quality of service by performing computational functions at higher throughput rates. Currently, various IT companies, including Amazon, Windows, Google, and Microsoft, support CO services.

The aforementioned NFV service and CO service are services that use networking resources and processing resources together, and have the following commonalities. In a network environment where NFV service and CO service are provided, there are several physical nodes capable of performing a given virtual function, and the end host of each service selects which resource to use to perform the function. . This is similar to network routing, but considers not only networking resources, ie, link capacity, but also processing resources, ie, processor capacity. For example, no matter how rich the networking resources are, if the processing resources of the corresponding path nodes are insufficient, the quality of service will be reduced. In addition, a path defined as a collection of networking resources is defined as a collection of networking resources and processing resources. In this case, depending on which of the given paths the service is performed, characteristics such as cost, delay, and throughput will be changed, and resource efficiency or service quality will be different. In addition, if one service can simultaneously use multiple candidate paths, additional service quality or system cost can be expected.

Existing studies have been conducted in the NFV environment and the CO environment only in each environment, and none of them have been considered in one framework. However, in a real network environment, NFV service and CO service coexist as well as general network service, and these services share networking resources and processing resources.

Therefore, there is a need for a resource management method that can manage these services together in a network environment where various resources such as networking resources and processing resources coexist.

Embodiments of the present invention provide a method and apparatus for managing a multi-resource infrastructure that can reduce system cost while improving service quality in a network environment including a plurality of physical nodes and a plurality of links in which networking and processing resources coexist. It is main purpose to do it.

An embodiment of the present invention includes a multi-resource infrastructure formed of at least two nodes using processing resources and at least one link using networking resources. manage at least one service provided based on at least one path formed by connecting at least one first end node and at least one second end node selected from each of the at least two nodes in a resource infrastructure A method, comprising: a candidate path set defining step of defining a set of candidate paths for the at least one service among the at least one path; And a multipath formation process of forming a multipath by dividing an input transmission rate for the at least one service into a set of candidate paths.

An embodiment of the present invention provides at least two nodes using processing resources, the first end node and the second end node, and at least one link using networking resources. link), wherein the multi-resource infrastructure includes at least one path connecting the first end node and the second end node via the at least one link. A candidate path set definition unit defining a set of candidate paths for at least one service; A multipath forming unit configured to form a multipath by dividing an input transmission rate for the at least one service into a set of candidate paths; And a transmission rate determining unit defining a dual variable for each of the processing resource and the networking resource, and determining the rate of transmission through the multipath by updating the dual variable. Provide infrastructure management device.

According to an embodiment of the present invention, a method and apparatus for managing a network function virtualization service and a computation offloading service together may be provided.

According to another aspect of an embodiment of the present invention, there is an effect of improving the efficiency of resource use or the quality of service in a network environment where networking resources and processing resources coexist.

1 is a conceptual diagram illustrating a conventional network routing environment.
2 is a conceptual diagram illustrating a conventional NFV service routing environment.
3 is a conceptual diagram illustrating a conventional CO service routing environment.
4 is a conceptual diagram of a multi-resource infrastructure management apparatus according to an embodiment of the present invention.
5 is a flowchart illustrating a method of managing a multi-resource infrastructure according to an embodiment of the present invention.
FIG. 6 illustrates a topology of two cases for explaining the result of applying the multi-resource infrastructure management method according to an embodiment of the present invention.
FIG. 7 is a graph illustrating service utilization and system cost when the multi-resource infrastructure management method according to an embodiment of the present invention is applied to Case 1 of FIG. 6.
FIG. 8 is a graph showing the effect on resource available capacity when the multi-resource infrastructure management method according to an embodiment of the present invention is applied to Case 1 of FIG. 6.
FIG. 9 is a graph illustrating an effect on service characteristics when the multi-resource infrastructure management method according to an embodiment of the present invention is applied to Case 2 of FIG. 6.
10 illustrates a real Internet topology in the United States.
FIG. 11 is a graph showing simulation results obtained by applying a multi-resource infrastructure management method according to an embodiment of the present invention to an actual large-scale example.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In adding reference numerals to the elements of each drawing, it should be noted that the same elements are designated by the same reference numerals as much as possible even though they are shown in different drawings. In addition, in describing the embodiments of the present invention, if it is determined that the detailed description of the related well-known configuration or function may obscure the gist of the embodiments of the present invention, the detailed description thereof will be omitted.

In describing the components of the embodiments according to the present invention, symbols such as first, second, i), ii), a), and b) may be used. These symbols are only to distinguish the components from other components, and the nature, order or order of the components are not limited by the symbols. In addition, when a part of the specification is said to include or 'include' an element, this may further include other elements, not to exclude other elements unless there is an expressly opposed to the description. it means.

Hereinafter, a multi-resource infrastructure management method and apparatus according to embodiments of the present invention will be described with reference to the accompanying drawings.

4 is a conceptual diagram of a multi-resource infrastructure management apparatus according to an embodiment of the present invention.

The multi-resource infrastructure management apparatus 400 according to an embodiment of the present invention includes a candidate path set defining unit 410, a multipath forming unit 420, and a transmission rate determining unit 430.

In order to help the description of the multi-resource infrastructure management apparatus 400, first, a general network environment in which the multi-resource infrastructure management apparatus 400 may be implemented may be defined. The network environment includes at least one link l included in the set of networking resources L and at least one node k included in the set of processing resources K. The set of networking resources and the set of processing resources may be represented by Equation 1, respectively.

The available capacity of the networking resource l and the processing resource k can be represented by C _l (bits / s) and C _k (cycles / s), respectively. Each processing resource supports a set of virtualization functions, such as network functions for NFV services and computation functions for CO services. The network environment also includes a set W of at least one coexisting service, each service having a path ( s _w , d _w ) from a source node ( s _w ) to a destination node ( d _w ). ) And virtual functions to be processed. Here, the node refers to a network entity that performs or processes at least one function using a processing resource. Among the various nodes, there is a source node which is a source of a service and a destination node which is a destination of a service. There can be multiple paths between the source node and the destination node between the links and across the nodes.

The candidate path set defining unit 410 of the multi-resource infrastructure management apparatus 400 according to an embodiment of the present invention may consider the same service as the source node and the destination node. There may be a CO service in which a function in which a source node and a destination node have the same service is consumed by the source node of the service in which the processed function is an example of the same service. The candidate path set defining unit 410 defines multiple candidate paths for each service w ∈ W as P _w . Where each path p ∈ P _w includes a networking resource set L _p and a processing resource set K _p . The networking resources included in the networking resource set L _p connect the source node s _w and the destination node d _w , and the processing resources belonging to the processing resource set K _p support the virtual functions provided by the service w . That is, the candidate path set defining unit 410 defines a set of candidate paths for at least one service among at least one path (S510). The entire candidate path set P is defined as in Equation 2.

Assuming that a service can use multiple multi-path candidates at the same time, the multi-path forming unit 420 of the multi-resource infrastructure management apparatus 400 according to an embodiment of the present invention, the source node of the service is the transmission speed (Rate) R _w is limited to R _w ≥ 0 and the transmission rate is divided into multiple candidate paths (S520). Here, x _p means the transmission rate assigned to the path p . That is, the transmission rates of the service w divided by the multipath forming unit 420 are as shown in Equation 3, which is the sum of the transmission rates allocated to the candidate paths.

The processing resource k is used by multiple candidate paths of coexisting services, and the virtual functions performed by the processing resource k may be different for the corresponding path p .

For example, some paths use processing resource k for video transcoding, while others may use the same processing resource k for DPI. To reflect this heterogeneity, we define the processing density ρ _{k, p} (cycles / bit) as the CPU cycles needed for the processing resource k for processing the unit bits of the operation on the path p . Therefore, the load (cycles / s) for the processing resource k for the path p becomes ρ _{k, p} x _p . Depending on the functions performed in processing resource k for path p , the output transmission rate (bitrate) after performing processing resource k may vary. For example, if a function is only for verification of data accounting and virus scanning, etc., the output transfer rate will be equal to the input transfer rate. On the other hand, if the function includes a function for converting input data such as encryption, transcoding, packet aggregation, etc., the output transmission rate may be different from the input transmission rate. Changes in transmission speed that are identified after processing are only seen on links that are successors processing resources on that path. To consider the change in the transmission rate after processing _, we define σ _{l, p} as the bit conversion ratio found in the networking resource l . Thus, the load (bit / s) on networking resource l for path p is σ _{l, p} x _p . For the transmission rates assigned to all paths p ∈ P , the total load on the networking resource l can be represented by F _l , and the total load on the processing resource k can be represented by F _k , which is represented by equations (4) and (4), respectively. Equation 5

Here, P _l and P _k are a set of paths to which the networking resource l belongs and a set of paths to which the processing resource k belongs, respectively.

To take into account the costs, delays, energy, or throughput that affect network operations, we define the cost function D _l (F _l ) for networking resource l and the cost function D _k (F _k ) for processing resource k , respectively. . Where D _l (F _l ) and D _k (F _k ) are convex and F _l And increase as F _k increases, and assume that it has a bounded second derivative. U _w (R _w) is a function of the utilization of services w (utility function) That is, the satisfaction of the service that is determined by the input transmission rate of the service w, convex, and increases as the transmission rate R _w is increased.

Routing in the conventional network routing environment shown in FIG. 1 includes data from the source node 110 to the destination node 130 without any functional requirements, including different source nodes 110 and destination nodes 130. do. In this case, the data may pass through various hardware routers 122, 124, and 126. Thus, routing for conventional routing services is defined as determining the path from the source node 110 to the destination node 130. When applied to the multi-resource infrastructure management apparatus 400 according to an embodiment of the present invention, ρ _{k, p} = for all k ∈ K _p , l ∈ L _p , p ∈ P _w and s _w ≠ d _w 0, sigma _{l, p} = 1.

In the conventional NFV service routing environment illustrated in FIG. 2, the NFV service is virtualized at nodes 222, 224, and 226 of the corresponding network, including different source nodes 210 and destination nodes 230. Therefore, routing of NFV service is defined as determining the path connecting the source node 210 and the destination node (@ 30) and performing the required network functions. When applying them to one embodiment a multi-resource infrastructure management apparatus 400 according to the present _invention, ρ _{k, p} ≠ 0 in all p ∈ P _w while the k ∈ K _p for a s _w ≠ d _w condition k Means if present. Here, although the network paths are the same, the networking resources and the processing resources may be different according to the functions allocated to the nodes 222, 224, and 226 included in the paths.

The CO service in the CO service routing environment shown in FIG. 3 includes the same source node 310 and destination node 310 and provides a computation function. Typically, nodes for executing computation functions include, but are not limited to, end hosts, edge clouds, and remote clouds. Thus, the routing of a CO service can be defined as determining a computation offloading policy among multiple processing resources. When computational functions are performed locally, i.e. on the end host, the service does not require any networking resources and is limited by the processing resources of the end host. As computation functions are offloaded to the edge cloud 340 or remote cloud 350, more powerful processing resources may be used, but additional networking resources will be consumed. Of this, when applied to a multi-resource infrastructure management apparatus 400 according to an embodiment of the present _invention, ρ _{k, p} ≠ 0 in all p ∈ P _w while s _w = d _w for the condition k ∈ K _p k Means if present. Computation functions of either service may be performed through multiple processing resources. For example, intermediate stage processing may be performed at the edge cloud 340 before proceeding with the main processing at the remote cloud 350.

On the other hand, the transmission rate determination unit 430 of the multi-resource infrastructure management apparatus 400 according to an embodiment of the present invention minimizes the total cost of networking resources and processing resources, and maximizes the service satisfaction, that is, the total service utilization To do this, a dual variable is defined for each of the processing resource and the networking resource. The transmission rate determiner 430 updates a defined dual variable to determine a transmission rate at which data is transmitted through various paths formed by the multipath forming unit 420.

5 is a flowchart illustrating a method of managing a multi-resource infrastructure according to an embodiment of the present invention.

The multi-resource infrastructure management method according to an embodiment of the present invention includes a candidate path set definition process (S510) and a multipath formation process (S520). The candidate path set definition process S510 and the multipath formation process S520 are the same as those described with reference to the multi-resource infrastructure management apparatus 400 of FIG. 4.

The multi-resource infrastructure management method according to an embodiment of the present invention may further include an object function calculation process (S530).

In the objective function calculation process (S530), a problem for maximizing the service utilization, that is, the sum of satisfaction of services, that is, reducing the consumption of costs incurred in the network environment and performing calculations. The multi-resource infrastructure management method according to an embodiment of the present invention is based on a networking cost to be generated using networking resources, a processing cost to be generated using processing resources, and a service satisfaction level which is a satisfaction level for at least one service. Calculate the objective function. Here, the problem, that is, the objective function G (x) is the same as the equation (6 ) .

를 원 변수(primal variable)로 정의한다. 상수 V는 전체 비용 최소화와 서비스 활용도 최대화 사이에서 발생하는 상충 관계 파라미터(trade-off parameter)이다. 이 목적함수 G(x)는 x에 대해 "엄격하게(strictly)" 볼록하지 않으며, 그에 따라 동일한 목적값을 얻을 수 있는 여러 최적해가 존재할 수 있다.here,

Is defined as the primary variable. The constant V is a trade-off parameter that occurs between minimizing overall costs and maximizing service utilization. This objective function G (x) is not "strictly" convex with respect to x , and thus there may be several optimal solutions that yield the same objective value.

The constraint in the objective function calculation process (S530) is the available capacity for each of the usage of the networking resource l and the processing resource p . In the multi-resource infrastructure management method according to an embodiment of the present invention, first, an available capacity of each of at least one networking resource and an available capacity of each of at least two processing resources are defined. Determining the input transmission rate R _w for each service in order to optimize this behavior, and proceeds to share the multi-path routing them through the x _p number of candidate paths for the given service.

인 x ^* 를 목적함수 G(x)에 대한 최적 원 해(optimal primal solution) 집합 X ^* 에 속하는 특정 속도 할당 벡터로 정의한다.here,

Which defines the x ^* as the objective function G (x) to the optimum source (optimal primal solution) specific rate assigned belongs to the set X ^* vector for.

Meanwhile, the Lagrangian dual problem for removing the constraint may be represented by Equation 7.

를 쌍대 변수로 정의한다.

은 수학식 8로 정의되는 라그랑주 함수이다.here,

Is defined as a dual variable.

Is a Lagrange function defined by Equation 8.

인 λ ^* 를 수학식 7에 대한 최적 쌍대 해(optimal dual solution) 집합 Λ ^* 에 속하는 특정 벡터로 정의함으로써, 커플링 문제를 일으키는 가용 용량 제약사항을 제거할 수 있다. 그러나 그래디언트 프로젝션 알고리즘(gradient projection algorithm)은 원 목적함수가 엄격히 볼록인 경우에만 적용할 수 있기 때문에, 안장점 정리(saddle point theorem)를 이용하여 목적함수를 변동부등식(variational inequality)으로 변경할 수 있다. 수학식 7에 대한 최적 원 쌍대 해(optimal primal and dual solution)를 찾는 것은 수학식 9를 만족시키는

를 구하는 안장점 정리 문제를 푸는 것과 등가하다. 이때, 수학식 9는 아래와 같이 나타낼 수 있다.here,

By defining λ ^* as a specific vector belonging to the optimal dual solution set Λ ^* for Equation 7, the usable capacity constraint causing the coupling problem can be removed. However, since the gradient projection algorithm can be applied only when the original objective function is strictly convex, the objective function can be changed into variational inequality by using saddle point theorem. Finding the optimal primal and dual solution for Eq. (7) satisfies Equation (9).

Equivalent to solving the saddle theorem problem. In this case, Equation 9 may be expressed as follows.

를 구하는 것은

와 수학식 10을 만족시키는 변동부등식과 등가하다.By substituting Equation 7 into Equation 9, a dual variable λ that does not depend on the original variable x can be considered. Saddle to satisfy Equation 9

Finding

Equivalent to the variable inequality that satisfies (10).

와 수학식 11을 만족시킨다.Where f is

And Equation 11 are satisfied.

In the multi-resource infrastructure management method according to an embodiment of the present invention, an extra gradient based on the above-described variation inequality in order to allow the above-described solution of the variation inequality to converge to an optimal solution in the objective function calculation process (S530). Transform into an algorithm.

The multi-resource infrastructure management method according to an embodiment of the present invention defines at least one networking resource and at least two processing resources to calculate each of the networking cost and the processing cost, the usage of each of the at least one networking resource and Compute the usage of each of the at least two processing resources.

In addition, the multi-resource infrastructure management method according to an embodiment of the present invention calculates a set of candidate paths using each of at least one networking resource, and calculates a set of candidate paths using each of the at least two processing resources. do.

The objective function calculation process S530 includes identifying a current usage of each of the at least two processing resources in order to update each of the at least two processing resources. In addition, the objective function calculation process S530 may include checking a current usage of each of the at least one networking resource in order to update each of the at least one networking resource.

,

및 x _p 를 출력하며, 그 과정을 정리하면 표 1과 같다. 여기서, τ 및 γ는 각각 반복되는 업데이트 회차(1, 2, 3, ...) 및 업데이트 간격(step size)를 의미한다.The multi-resource infrastructure management method according to an embodiment of the present invention described above is to update x _p

,

And x _p are output. Table 1 shows the process. Here, τ and γ denote update cycles 1, 2, 3, ... that are repeated, and step size.

number process Equation One

를 업데이트After each processing resource k checks its current usage, it is a dual variable

Update

2 After each networking resource l checks its current usage, it is a dual variable.

Update

3 The end node of each service w updates x _p for each candidate path p ∈ P given to it. Equation 12

Here, Equation 12 in Table 1 is as follows.

FIG. 6 illustrates a topology of two cases for explaining the result of applying the multi-resource infrastructure management method according to an embodiment of the present invention.

In case 1, the network includes three nodes 611, 613, 622 and four links 632, 634, 636, 638. The first SC1 node 611 and the second SC1 node 613 are nodes that contain processing resources, and the third SC1 node 622 is a node that does not contain processing resources. In Case 1, two services with different source and destination nodes, for example NFV services, require virtual functions in a triangular topology. Each service may use a one-hop path using two resources and a two-hop path using three resources.

In case 2, the network includes three nodes 612, 614, 616 and four links 631, 633, 635, 637. The first SC2 node 612, the second SC node 614, and the third SC node 616 all contain processing resources. In this case, one service where the source node and the destination node are the same, eg a CO service, requires virtual functions in a serially connected topology. The path using only processing resource k ₁ may be the case of local computing, the path using processing resource k ₂ may be computation offloading using the edge cloud, and the path using processing resource k ₃ may be compute off using remote cloud It may be loading. In this case, the cloud node has the largest processing available capacity. That is, C _k1 < C _k2 < C _k3 . However, it consumes the least cost for the same processing load. That is, D _k1 (F) > D _k2 (F) > D _k3 (F) .

Table 2 shows the details of the two cases described above, namely, Case 1 and Case 2.

Case 1 Case 2 Relationship between source node and destination node Different equivalence Number of services | W | = 1 | W | = 1 Candidate path p ₁₁ : k ₁ → l ₃
p ₁₂ : l ₂ → k ₂ → l ₄
p ₂₁ : k ₂ → l ₄
p ₂₂ : l ₁ → k ₁ → l ₃ p ₁₁ : k ₁
p ₁₂ : l ₁ → k ₂ → l ₂
p ₁₃ : l ₁ → l ₃ → k ₃ → l ₄ → l ₂ Resource usable capacity (Mbps or GHz) C _l = 2, ∀ l ∈ L
C _k1 ∈ {1.3, 1.4,... , 2.3}, C _k2 = 2 C _l = 2, ∀ l ∈ L
C _k1 = 2 , C _k2 = 20 , C _k3 = 40 Resource cost D _l (F _l ) = F _l , ∀ l ∈ L
D _k ( F _k ) = F _k , ∀ k ∈ K D _l (F _l ) = F _l , ∀ l ∈ L
D _k1 ( F _k1 ) = F _k1 , D _k2 ( F _k2 ) = ½ F _k2 , D _k3 ( F _k3 ) = ⅓ F _k3 , Service utilization U _w ( R _w ) = log (0.1 + R _w )-log (0.1) U _w ( R _w ) = log (0.1 + R _w )-log (0.1) Processing Density (Kcycles / bit) ρ _{k, p} = 1, ∀ k ∈ Kp , p ∈ P ρ ∈ {1, 2,... , 20}
ρ _{k, p} = ρ , ∀ k ∈ Kp , p ∈ P Bit conversion ratio (bits / bit) σ _{l, p} = 1, ∀ l ∈ Lp , p ∈ P sigma ∈ {0.05, 0.1,... ,One}
σ _{l1, p12} = 1, σ _{l2, p12} = σ
σ _{l1, p13} = 1, σ _{l3, p13} = 1,
σ _{l4, p13} = σ , σ _{l2, p13} = σ

In order to examine the effect of resource management considering cost, utilization of cost-utility-minimal algorithm to which the multi-resource infrastructure management method according to an embodiment of the present invention is applied when there is no cost item- Compared with the utility-maximal algorithm. That is, D _l (F) = D _k ( F _k ) = 0 for all l ∈ L and k ∈ K.

FIG. 7 is a graph illustrating service utilization and system cost when the multi-resource infrastructure management method according to an embodiment of the present invention is applied to Case 1 of FIG. 6.

7 (a) shows the service utilization according to the number of repetitions when C _k1 = 2 GHz, V = 20 and γ = 0.005, Figure 7 (b) and Figure 7 (b) is C _k1 = 2 GHz , System cost according to the number of iterations when V = 20 and γ = 0.005. Initial baud rate assigned value x ⁰ to x ⁰ 시뮬레이션 The simulation is considered by considering four things: {(0, 0, 0, 0), (0, 1, 0, 1), (1, 2, 1, 2), (2, 2, 2, 2)}. Proceeded.

We can see that both algorithms converge to maximum utilization regardless of the initial rate assignment. Maximum utilization was achieved when the transmission rate of each service was 2 Mbps, that is, when R ₁ = R ₂ = 2 Mbps. Referring to Figure 7 (b), when the multi-resource infrastructure management method according to an embodiment of the present invention, two services do not share any resources, that is, x _p11 When = x _p21 = 2 Mbps, and, x _p12 = x _p22 = 0 can be confirmed that the convergence to the minimum system cost.

On the other hand, it can be seen that the utilization-maximum algorithm converges to a system cost value other than the minimum value according to the change of the initial transmission rate allocation value.

FIG. 8 is a graph showing the effect on resource available capacity when the multi-resource infrastructure management method according to an embodiment of the present invention is applied to Case 1 of FIG. 6.

To this end, the optimal transmission rate allocation value was calculated while changing the available capacity of the processing resource k ₁ , that is, C _k1 from 1.3 GHz to 2.3 GHz. If C _k1 exists between 1.3 GHz and 1.8 GHz, it can be seen that Service 1 used both p ₁₁ and p ₁₂ candidate paths, and that processing resource k ₂ and networking resource l ₄ were shared by both services. . When each service used only the 1-hop path of the service, that is, p ₁₁ and p ₂₁ , there was a big difference in the transmission speed between the two services. The two-hop path p ₁₂ causes a higher cost than the one-hop path p ₂₁ , and the rate assignment for service 2 is greater than the rate assignment for service 1.

In order to examine the effect of different service parameters on service throughput, the behavior of service parameters was examined while changing the processing density ( ρ ) and the bit conversion ratio ( σ ).

FIG. 9 is a graph illustrating an effect on service characteristics when the multi-resource infrastructure management method according to an embodiment of the present invention is applied to Case 2 of FIG. 6.

As the processing density ρ and the bit conversion ratio σ increased, the total throughput was found to decrease. The reason is that large processing density ρ and bit conversion ratio σ require higher system cost to achieve the same utilization. With low processing density values, it can be seen that each service is provided by local computing. With low processing densities, local processing resources can fully handle compute functions, achieving high throughput without relying on edge and remote clouds that require excessive networking costs.

In order to apply the multi-resource infrastructure management method according to an embodiment of the present invention to a large-scale, realistic case, a simulation was performed using a real Internet topology of the US provided by MCI in 2011.

10 illustrates a real Internet topology in the United States.

This data set contains 19 nodes and 45 wired links. Since this data set does not specify processing resources, processing resources are randomly calculated.

FIG. 11 is a graph showing simulation results obtained by applying a multi-resource infrastructure management method according to an embodiment of the present invention to an actual large-scale example.

As the number of candidate paths increases, the transmission speed may increase, but the cost may increase. Therefore, transmission rate allocation for each candidate path must be carefully determined. This result means that the multi-resource infrastructure management method according to an embodiment of the present invention can utilize multiple paths to increase service utilization while keeping the system cost low.

In FIG. 5, each process is described as being sequentially executed, but is not necessarily limited thereto. In other words, since the process described in FIG. 5 may be applied by changing or executing one or more processes in parallel, FIG. 5 is not limited to the time series order.

Meanwhile, each step of the flowchart shown in FIG. 5 may be embodied as computer readable codes on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. That is, the computer-readable recording medium may be a magnetic storage medium (for example, ROM, floppy disk, hard disk, etc.), an optical reading medium (for example, CD-ROM, DVD, etc.) and a carrier wave (for example, the Internet Storage medium). The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

The above description is merely illustrative of the technical spirit of the embodiments according to the present invention, and those skilled in the art to which the embodiments of the present invention belong will not depart from the essential characteristics of the embodiments. Various modifications and variations will be possible. Therefore, the exemplary embodiments according to the present invention are not intended to limit the technical spirit of the present embodiments, but to describe the present invention, and the scope of the technical spirit of the present embodiments is not limited by these embodiments. The protection scope of the embodiments according to the present invention should be interpreted by the following claims, and all technical ideas falling within the equivalent scope should be interpreted as being included in the scope of the embodiments according to the present invention.

110, 210, 310: source nodes 122, 124, 126: hardware routers
130, 230, 310: destination node 222, 224, 226: NFV support node
340: edge cloud 350: remote cloud
400: multi-resource infrastructure management device
410: candidate path set definition unit 420: multi-path forming unit
430: transmission rate determining unit

Claims (11)

A first SC1 node and a second SC1 node forming two links (l ₁ , l ₂ ) as source nodes on a network to process processing resources k ₁ ;
A third SC1 node which forms two links (l ₃ , l ₄ ) with each of the first SC1 node and the second SC1 node as a destination node on a network, thereby processing the processing resource but processing a networking resource. ;
The first SC1 node, the second SC1 node, and the third SC1 node are a one-hop path using two resources in a triangular topology and a two-hop path using three resources. network function virtualization (NFV) service using a path,
Define a set of candidate paths for the NFV service among the paths according to the four links (l ₁ , l _2, l ₃ , l ₄ ), and divide the input transmission rate for the NFV service by the set of the candidate paths. And a dual variable for each of the processing resource and the networking resource, and updating the dual variable to determine the rate of transmission over the multipath.
Multi-resource infrastructure management system comprising a. The method of claim 1,
The multi-resource infrastructure management device,
For managing the multi-resource infrastructure, the objective function is calculated based on the networking cost incurred using the networking resource, the processing cost incurred using the processing resource, and the service satisfaction, which is the satisfaction with the NFV service. Multi-resource infrastructure management system. The method of claim 2,
The multi-resource infrastructure management device,
Define at least one networking resource and at least two processing resources to calculate each of the networking cost and the processing cost according to the objective function, and each usage of the at least one networking resource and each of the at least two processing resources A multi-resource infrastructure management system, characterized in that for calculating the usage of. The method of claim 3, wherein
The multi-resource infrastructure management device,
And calculating a set of candidate paths using each of the at least one networking resource, and calculating a set of candidate paths using each of the at least two processing resources. The method of claim 4, wherein
The multi-resource infrastructure management device,
And define available capacity of each of the at least one networking resource and available capacity of each of the at least two processing resources according to the objective function calculation. The method of claim 5,
The multi-resource infrastructure management device,
Calculating a set of networking resources used by each path included in the set of candidate paths using each of the at least one networking resource according to the objective function calculation, and a set of candidate paths using each of the at least two processing resources A multi-resource infrastructure management system, characterized in that to calculate a set of processing resources used by each path included in. delete The method of claim 2,
The multi-resource infrastructure management device,
Identify a current usage of each of the at least two processing resources to update each of the at least two processing resources according to the objective function calculation, and determine the current usage of each of the at least one networking resources to update each of the at least one networking resources. Multi-resource infrastructure management system, characterized in that to check the current usage. A first SC2 node for processing a processing resource k ₁ on a network;
A second SC2 node forming a link (l ₁ , l ₂ ) with the first SC2 node on the network to process a processing resource k ₂ ;
A third SC2 node forming a link (l ₃ , l ₄ ) with the second SC2 node on the network to process a processing resource k3;
The first SC2 node, the second SC2 node, and the third SC2 node each have a processing resource k ₁ processed by the first SC2 node in a topology connected in series so that a source node and a destination node are identical to each other. And provide a processing offload (CO) service through the processing resource k ₂ processed by the second SC2 node to a path of an edge cloud, and use the processing resource k ₃ processed by the third SC2 node as a path to a remote cloud. Provide services,
Define a set of candidate paths for the CO service among the paths according to the four links (l ₁ , l _2, l ₃ , l ₄ ), and divide the input transmission rate for the CO service by the set of candidate paths. A multi-resource infrastructure management device defining a dual variable for the processing resource and defining the speed of transmitting the multi-path by transmitting the multivariate to form a path.
Multi-resource infrastructure management system comprising a.
delete delete

Images