Control device, control method, and program

Abstract

A control device according to an aspect of the present disclosure is for embedding a virtual network in a physical network, the control device including: an acquisition unit that acquires samples of the energy production at physical nodes included in the physical network where there are power sources with uncertainty in the probability distribution of energy production; and a calculation unit that uses a prescribed distance function to define an uncertainty set representing the uncertainty regarding the probability distribution on the basis of the samples, and calculates an assignment of the virtual network so as to optimize the expected value of a prescribed objective function within the range of the uncertainty set.

Description

Control device, control method, and program

[0001] The present disclosure relates to a control device, a control method, and a program.

[0002] In recent years, with the development of Network Function Visualization (NFV), a virtualization technology, it has become possible to provide a huge variety of network services by flexibly combining a wide variety of virtual network resources (VRs) and virtual network functions (VNFs). To provide such services, it is necessary to appropriately allocate VRs and VNFs to the physical network for each service. In addition, it is necessary to appropriately control the end-to-end route for transmitting the service and guarantee the desired communication performance (e.g., the desired communication delay time, etc.). Furthermore, to process a huge number of services, it is necessary to be able to respond to sudden fluctuations in traffic volume, so it is necessary to suppress link congestion in the network and realize a durable network.

[0003] In recent years, there has been a global movement to introduce renewable energy (hereinafter referred to as "renewable energy") in order to reduce the environmental impact. This trend is no exception in the operation of communication networks for running virtual network environments, and it is expected that renewable energy will account for a large proportion of the electricity supplied to communication networks.

[0004] On the other hand, because the amount of power supplied by renewable energy depends on the natural environment, etc., it is assumed that the amount of power available to the communication network will be obtained as uncertain information that includes prediction errors. Also, to prevent supply shortages in the event of sudden fluctuations in the amount of renewable energy power generation, it is necessary to realize a durable network by suppressing congestion related to the amount of power consumed by each node in the network.

[0005] For example, Non-Patent Document 1 discloses a method for determining VNF allocation and routing that minimizes some objective function (e.g., an objective function representing total power consumption) of the entire communication network robustly with respect to both the amount of renewable energy power generation and the amount of traffic related to the service, under the assumption that there is uncertainty in both the amount of renewable energy power generation and the amount of traffic related to the service. Also, for example, Non-Patent Document 2 discloses a method for determining robust virtual network allocation in consideration of the probability of renewable energy power generation and in a situation where there is uncertainty in the probability of the generation.

[0006] Kengo Urata, Ryota Nakamura, and Shigeaki Harada, "Robust VNF ​​Allocation for Uncertainties of Traffic Volume and Renewable Energy," IEICE Technical Report, vol. 120, no. 414, IN2020-57, pp. 19-24, March 2021. Kengo Urata, Tsubasa Harada, Ryota Nakamura, and Tsubasa Harada, "Virtual Network Allocation Based on Distributed Robust Optimization for Utilizing Uncertain Renewable Energy," IEICE General Conference, B-7, March 10, 2023.

[0007] However, conventional techniques may not be robust enough to withstand the uncertainty in forecasting the amount of power generated by renewable energy.

[0008] The present disclosure has been made in consideration of the above points, and aims to realize a virtual network control that is robust against the uncertainty of power generation amount predictions.

[0009] A control device according to one aspect of the present disclosure is a control device for embedding a virtual network in a physical network, and includes: an acquisition unit that acquires samples of power generation at physical nodes included in the physical network, where a power source with uncertainty in the probability distribution of power generation exists; and a calculation unit that defines an uncertainty set representing uncertainty regarding the probability distribution based on the samples using a predetermined distance function, and calculates the allocation of the virtual network so as to optimize the expected value of a predetermined objective function within the range of the uncertainty set.

[0010] This makes it possible to realize robust virtual network control against the uncertainty of power generation forecasts.

[0011] It is a figure for explaining an example of allocation of a virtual network. It is a figure which shows an example of a hardware configuration of a virtual network control device based on this embodiment. It is a figure which shows an example of a functional configuration of a virtual network control device based on this embodiment. It is a flowchart which shows an example of a virtual network control process based on this embodiment.

[0012] An embodiment of the present invention will be described below. In the following embodiment, renewable energy is assumed as a power source with uncertainty in the power generation amount prediction, and a case will be described in which robust virtual network control (VNF allocation and route determination) is realized with respect to the uncertainty in the power generation amount prediction of renewable energy. However, renewable energy is only one example of a power source with uncertainty in the power generation amount prediction, and the following embodiment can be similarly applied to any power source with uncertainty in the power generation amount prediction.

[0013] <Details of the Problems of the Prior Art> In the method disclosed in Non-Patent Document 1, the uncertainty of predicted values ​​for traffic volume and renewable energy power generation volume is treated as a polyhedron set, but the distribution of generated uncertainty is not taken into consideration. As a result, it is not possible to distinguish between events with extremely low probability of occurrence (e.g., incorrect predictions due to sudden changes in weather conditions caused by natural disasters) and events that are empirically determined to have a high probability of occurrence (e.g., prediction errors due to technical limitations in weather forecasting), which can result in inefficient VNF ​​allocation and route determination under normal circumstances.

[0014] Furthermore, in the method disclosed in the above-mentioned Non-Patent Document 1, a set of polyhedra representing uncertainty is assumed to be given in advance, but because the accuracy of renewable energy power generation forecasts varies greatly depending on weather conditions, the information regarding uncertainty given in advance is not necessarily precise and accurate. That is, for example, in order to increase the precision, it is necessary to define a small set of polyhedra representing uncertainty, but in that case, the precision may be impaired. On the other hand, for example, in order to accurately capture the uncertainty, it is necessary to define a large set of polyhedra, but in that case, the precision may be impaired.

[0015] On the other hand, in the method disclosed in the above-mentioned Non-Patent Document 2, the uncertainty regarding the probability distribution of the occurrence probability of renewable energy power generation is defined as a polyhedron set using information on average values ​​calculated from accumulated data on renewable energy power generation. However, within the polyhedron set defined in this way, if the average value is the same, the probability distributions are treated as the same distribution even if their shapes are different, which may reduce the robustness of the actual virtual network allocation results.

[0016] <Outline of the Proposed Method> To solve the above-mentioned problems associated with the conventional methods disclosed in Non-Patent Documents 1 and 2, we propose a method for realizing virtual network control by treating the uncertainty in renewable energy power generation forecasts as a probability distribution and assuming that the predetermined probability distribution itself has uncertainty. By treating the uncertainty in renewable energy power generation forecasts as a probability distribution, the proposed method can realize robust virtual network control that captures the probability of incorrect predictions that must be taken into account during operation. Furthermore, the proposed method can realize more robust virtual network control by assuming that the predetermined probability distribution itself has uncertainty and defining the set of uncertainties based on a distance measure between the probability distributions.

[0017] To achieve the above-mentioned virtual network control, the proposed method introduces a distributional robust optimization method based on the Wasserstein metric (Reference 1). This method is a robust optimization method that takes into account the uncertainty of probability distributions. In the proposed method, we first define a set that represents uncertainty based on the Wasserstein metric, which measures the degree of discrepancy between the true probability distribution and the actual sample distribution. We then use this set to formulate a distributional robust optimization problem that solves VNF allocation and routing in virtual network control. The proposed method then reduces the distributional robust optimization problem to a solvable class of optimization problems and solves them using existing solvers to determine VNF allocation and routing. This enables real-time control that is robust to the uncertainty of renewable energy power generation forecasts.

[0018] <Details of the Proposed Method> As an example, consider a physical network 1000 as shown in Fig. 1. The physical network 1000 shown in Fig. 1 includes physical nodes n １ ~n ４ , physical link l １ ~l ５ In addition, a part of the physical nodes in the physical network 1000 (physical node n １ and n ３ ) will be supplied with electricity from renewable energy power generation facilities.

[0019] In this case, under the circumstances where there is uncertainty in the forecast of renewable energy generation, Ｓ Consider the problem of embedding virtual networks into a physical network. Hereinafter, let S be the set of virtual networks, N be the set of physical nodes, and L be the set of physical links. Note that the service provided by the sth virtual network will also be referred to as the sth service.

[0020] Here, it is assumed that there are two types of physical nodes. One is a physical node that acts as a router or a user accommodating station (physical node n in the example shown in FIG. 1). ２ , n ４ ), and the other is a physical node (in the example shown in Figure 1, physical node n) that, in addition to its role, also serves as a data center that processes virtual machines (VMs) that realize VNFs, etc. １ , n ３ Hereinafter, the physical node that serves as the data center is called a DC node, and the set of DC nodes is called N. ＤＣ Let ⊆N.

[0021] Assuming that power is supplied to a DC node from a renewable energy power generation facility attached to the DC node or from a renewable energy power generation facility via a power transmission and distribution network, the amount of renewable energy power generation supplied to the nth DC node is defined as ξ ｎ That is, the DC node is a physical node that demands renewable energy power generation.

[0022] A virtual network is composed of three elements: a user, a virtual machine (VM), and a virtual link. The user's location (i.e., one of the user accommodation nodes present on the physical node) is assumed to be given in advance. Furthermore, when a virtual network is assigned and the user's location differs from the VM's placement location (one of the DC nodes), traffic occurs between the user and the physical node where the VM is placed. In this case, the amount of traffic generated in the sth virtual network is defined as λ ｓ For example, in the example shown in FIG. 1, the location of user 1 in the first virtual network is physical node n １ , the VM is placed on physical node n ３ and the physical link l ３ The traffic volume λ １ Similarly, the location of user 2 in the second virtual network is physical node n ２ , the VM is placed on physical node n ３ and the physical link l ２ The traffic volume λ ２ Similarly, the location of user 3 in the third virtual network is physical node n ４ , the VM is placed on physical node n １ and the physical link l ４ The traffic volume λ ３ In the following description, the VM in the s-th virtual network will also be referred to as the s-th VM.

[0023] The vector representing the amount of renewable energy generated and supplied to the DC node is denoted as ξ and expressed as follows:

[0024] That is, the renewable energy power generation vector ξ is the renewable energy power generation amount ξ supplied to the nth DC node ｎ Let be the nth element |N ＤＣ |-dimensional vertical vector.

[0025] First, to define the set of uncertainties for the probability distribution Q of the renewable energy generation vector ξ, we define two probability distributions Q on the measurable space K(Z) constrained by a certain constraint Z. １ , Q ２ The Wasserstein distance dＷ : K(Z) × K(Z) → R is defined by the following equation (1), where R is the set of all real numbers.

[0026] where Π is the marginal distribution Q １ , Q ２ The random variable ζ １ , ζ ２ ∈R ｋ (where k is the random variable ζ １ , ζ ２ represents the joint probability distribution of the number of dimensions of R ｋ is any norm on

[0027] Also, n∈N ＤＣ Renewable energy generation amount ξ ｎ The upper limit of ξ ｎ Ｕ , the lower limit is ξ ｎ Ｌ The region Ξ that the renewable energy power generation vector ξ can take is given by the following equation (2).

[0028] In this case, the uncertainty set D regarding the probability distribution Q of the renewable energy power generation amount vector ξ is defined by the following equation (3).

[0029] Here, P Ｍ is the sample distribution when M renewable energy generation vectors ξ are given, and ε is the true probability distribution Q and the sample distribution P Ｍ The uncertainty set D given by the above formula (3) is a parameter that defines the Wasserstein distance between the true probability distribution Q and the sample distribution P Ｍ This means that the object lies within a Wasserstein sphere defined by the radius ε.

[0030] Based on the above preparations, we formulate the robust VM allocation problem under the uncertainty set D defined by the above equation (3) (i.e., the uncertainty set related to the probability distribution of renewable energy power generation) as a distributional robust optimization problem shown in the following equation (4).

[0031] Here, x ｎｓis a binary variable that takes the value of 1 when the sth VM is placed on the nth DC node, and takes the value of 0 otherwise. ｌｓ is a binary variable that takes the value of 1 when the traffic of the sth virtual network is transmitted on the lth physical link, and takes the value of 0 otherwise. １ (x ｎｓ , ξ ｎ ) is the sum of the surplus power on all DC nodes, and the objective function J ２ (y ｌｓ ) represents the sum of delay times occurring in all virtual networks, and are defined by the following equations (5) and (6), respectively.

[0032] Here, c ｎｓ represents the amount of power consumption generated when the sth VM is placed on the nth DC node, and p ｌｓ represents the delay time required for the traffic of the sth virtual network to be transmitted on the lth physical link. Also, γ ≧ 0 in the above equation (4) is the sum of the two objective functions J １ , J ２ The weighting coefficient between the objective function J １ (x ｎｓ , ξ ｎ ) leads to maximizing the utilization of renewable energy power generation on the physical network, which means that optimizing the above equation (4) is a VM allocation problem for efficient use of renewable energy.

[0033] Furthermore, X satisfies the constraints shown in the following formulas (7a) to (7h) (x ｎｓ , y ｌｓ ) is a set of

[0034] Here, the constraint condition shown in the above formula (7a) is a constraint on the CPU (Central Processing Unit) in each DC node, and θ ｎｓ ＣＰＵ is the CPU processing amount generated when the sth VM is placed on the nth DC node, and θ ＣＰＵThe constraints shown in the above formula (7b) are constraints on the memory in each DC node, and θ ｎｓ ｍｅｍ is the memory usage generated when the sth VM is placed on the nth DC node, and θ ｍｅｍ The constraints shown in the above formula (7c) are constraints on storage in each DC node, and θ ｎｓ ｓｔ is the storage usage generated when the sth VM is placed on the nth DC node, and θ ｓｔ The constraints shown in the above formula (7d) are the bandwidth φ ｌ is the capacity constraint of the l-th physical link, where I is the capacity constraint of the l-th physical link. The constraint shown in the above formula (7e) is a constraint that the delay time of the service provided by each virtual network is equal to or less than the latency upper limit value T. The constraint shown in the above formula (7f) is a constraint that a VM in each virtual network can only be placed on one physical node (DC node). The constraint shown in the above formula (7g) is a constraint that represents the conservation law of the outflow amount of traffic volume at the n-th physical node. However, I ｎ ⊆L is the set of links that allow traffic to flow into the nth physical node, O ｎ ⊆L is the set of links through which the traffic volume flowing from the nth physical node flows, and x ｎｓ （０） is a binary value that takes a value of 1 when the user in the sth virtual network is accommodated in the nth physical node, and takes a value of 0 otherwise. The constraint shown in the above equation (7h) is the decision variable x ｎｓ , y ｌｓ is a constraint that indicates that is a binary variable.

[0035] In order to solve the distributional robust optimization problem shown in the above formula (4), the problem is reduced to an integer linear programming problem that is equivalent to formula (4) and is a solvable class. Let F:={1,...,M}, and let m (where m∈F) sampled renewable energy power generation be ξ ｎ （ｍ）In this case, the distributional robust optimization problem shown in the above equation (4) is reduced to an integer linear programming problem shown in the following equations (8a) to (8f).

[0036] where a∈R ２ , b ｎ ∈R ２ are vectors defined as follows:

[0037] In addition, π in the above formula (8f) ｍｎ１ , π ｍｎ２ ≧0 are two-dimensional vectors, ｍｎ１ All elements of are 0 or more, and π is a two-dimensional vector. ｍｎ２ are constraints that all elements of are greater than or equal to 0.

[0038] The integer linear programming problems shown in the above equations (8a) to (8f) can be easily solved using existing solvers, etc. Therefore, the VNF allocation (i.e., VM placement) and routing decision that are robust against the uncertainty of renewable energy power generation forecasts can be calculated using (x ｎｓ , y ｌｓ ) (n∈N, s∈S, l∈L) is obtained as a solution.

[0039] Below, we show the VNF allocation and routing decision that is robust against the uncertainty of renewable energy power generation forecasts using the proposed method described above (x ｎｓ , y ｌｓ ) (nεN, sεS, lεL) and then controls the virtual network so as to realize the VNF allocation and route determination.

[0040] <Example of Hardware Configuration of Virtual Network Control Device 10> An example of the hardware configuration of the virtual network control device 10 according to this embodiment will be described with reference to Fig. 2. Fig. 2 is a diagram showing an example of the hardware configuration of the virtual network control device 10 according to this embodiment.

[0041] 2, the virtual network control device 10 according to this embodiment includes an input device 101, a display device 102, an external I / F 103, a communication I / F 104, a RAM (Random Access Memory) 105, a ROM (Read Only Memory) 106, an auxiliary storage device 107, and a processor 108. Each of these pieces of hardware is connected to each other via a bus 109 so as to be able to communicate with each other.

[0042] The input device 101 is, for example, a keyboard, a mouse, a touch panel, a physical button, etc. The display device 102 is, for example, a display, a display panel, etc. Note that the virtual network control device 10 does not necessarily have to include at least one of the input device 101 and the display device 102, for example.

[0043] The external I / F 103 is an interface with an external device such as a recording medium 103a. Examples of the recording medium 103a include a CD (Compact Disc), a DVD (Digital Versatile Disk), an SD memory card (Secure Digital memory card), and a USB (Universal Serial Bus) memory card.

[0044] The communication I / F 104 is an interface for connecting the virtual network control device 10 to a physical network. The RAM 105 is a volatile semiconductor memory (storage device) that temporarily stores programs and data. The ROM 106 is a non-volatile semiconductor memory (storage device) that can store programs and data even when the power is turned off. The auxiliary storage device 107 is a non-volatile storage device (storage device) such as an HDD (Hard Disk Drive), an SSD (Solid State Drive), or a flash memory. The processor 108 is an arithmetic device such as a CPU.

[0045] 2 is an example, and the hardware configuration of the virtual network control device 10 is not limited to this. For example, the virtual network control device 10 may have multiple auxiliary storage devices 107 or multiple processors 108, may not have some of the hardware shown in the figure, or may have various hardware other than the hardware shown in the figure.

[0046] <Example of Functional Configuration of Virtual Network Control Device 10> An example of the functional configuration of the virtual network control device 10 according to this embodiment will be described with reference to Fig. 3. Fig. 3 is a diagram showing an example of the functional configuration of the virtual network control device 10 according to this embodiment.

[0047] 3, the virtual network control device 10 according to this embodiment includes a renewable energy power generation amount acquisition unit 201, a physical network information acquisition unit 202, a virtual network information acquisition unit 203, a solution-finding unit 204, and a control unit 205. Each of these units is realized by, for example, processing in which one or more programs installed in the virtual network control device 10 are executed by the processor 108 or the like.

[0048] The renewable energy power generation amount acquisition unit 201 acquires the renewable energy power generation amount ξ supplied to each DC node. ｎ (n∈N ＤＣ ) for M samples. That is, the renewable energy power generation amount acquisition unit 201 acquires M samples of each n∈N ＤＣ Regarding renewable energy generation amount ξ ｎ （１） , ..., ξ ｎ （Ｍ） Here, the renewable energy power generation amount acquisition unit 201 only needs to acquire the renewable energy power generation amount that can be used to construct a sample distribution in the time period that is the control target of the virtual network. For example, the renewable energy power generation amount ξ is calculated by calculating the value of the power demand of each DC node for the past M days. ｎ （ｍ） Alternatively, the renewable energy power generation amount acquisition unit 201 may predict the amount of renewable energy power generation supplied to each DC node in the time period by using some prediction method, and obtain these predicted values ​​as the renewable energy power generation amount ξ ｎ （ｍ）(m∈F).

[0049] Furthermore, the renewable energy power generation amount acquisition unit 201 acquires the renewable energy power generation amount n∈N ＤＣ Renewable energy generation amount ξ ｎ The upper limit of ξ ｎ Ｕ and the lower limit ξ ｎ Ｌ The renewable energy power generation amount acquisition unit 201 sets the upper limit value ξ given by the user of the virtual network control device 10, for example. ｎ Ｕ and the lower limit ξ ｎ Ｌ Alternatively, the renewable energy generation amount ξ ｎ （ｍ） (n∈N ＤＣ , m∈F) to the upper limit ξ ｎ Ｕ and the lower limit ξ ｎ Ｌ may be set.

[0050] The physical network information acquisition unit 202 acquires physical network information. Here, the physical network information includes topology information of the physical network (a set of physical nodes N, a set of DC nodes N ＤＣ ⊆N, a set of physical links L), and the bandwidth of each physical link φ ｌ The physical network information also includes the delay time p ｌｓ , the power consumption generated in each DC node when each VM is placed, c ｎｓ The physical network information also includes the CPU processing amount θ generated in each DC node when each VM is placed. ｎｓ ＣＰＵ and its upper limit θ ＣＰＵ , the memory usage θ generated in each DC node when each VM is placed. ｎｓ ｍｅｍ and its upper limit θ ｍｅｍ ,The storage usage θ generated in each DC node when each VM is placed ｎｓ ｓｔ and its upper limit θ ｓｔThe physical network information acquisition unit 202 may acquire the above physical network information from, for example, a management device that manages the physical network.

[0051] The virtual network information acquisition unit 203 acquires virtual network information. Here, the virtual network information includes the location x of the user of each virtual network. ｎｓ （０） , the traffic volume generated in each virtual network, λ ｓ The virtual network information acquisition unit 203 may acquire the virtual network information from, for example, a management device that manages the virtual network, or may acquire the virtual network information provided by a user of the virtual network control device 10.

[0052] The solution-finding unit 204 solves two objective functions J １ , J ２ After setting a parameter γ representing a weighting coefficient between the two and a parameter ε for defining the uncertainty set D, the distributional robust optimization problem shown in the above equation (4) is formulated using these parameters γ and ε, the physical network information, and the virtual network information. Then, the solution finding unit 204 reduces the distributional robust optimization problem shown in the above equation (4) to an integer linear programming problem shown in the above equations (8a) to (8f), and then finds an optimal solution using an existing solver or the like.

[0053] The control unit 205 determines whether the optimal solution is included (x ｎｓ , y ｌｓ ) (n∈N, s∈S, l∈L) controls the VM placement and routes (service routes from users to VMs) on the physical network. As a result, a virtual network that realizes the VM placement and routes is embedded on the physical network.

[0054] <Virtual Network Control Processing> The virtual network control processing according to this embodiment will be described with reference to FIG. 4. FIG. 4 is a flowchart showing an example of the virtual network control processing according to this embodiment. Note that the virtual network control processing is repeatedly executed, for example, at predetermined control intervals. As a typical example, the virtual network control processing is repeatedly executed at predetermined time intervals of several hours to one day, and the virtual network is controlled during that time interval. Below, a case where a virtual network is controlled during a certain time interval will be described.

[0055] The renewable energy power generation amount acquisition unit 201 acquires the renewable energy power generation amount ξ ｎ （１） , ..., ξ ｎ （Ｍ） (n∈N ＤＣ ) is acquired (step S101).

[0056] The renewable energy power generation amount acquisition unit 201 acquires the renewable energy power generation amount ξ ｎ The upper limit of ξ ｎ Ｕ and the lower limit ξ ｎ Ｌ is set (step S102).

[0057] The physical network information acquisition unit 202 acquires the physical network information (step S103). Note that, for example, if the physical network information was acquired in a control period prior to the current control period and there is no change in the physical network information, the processing of this step may not be executed.

[0058] The virtual network information acquisition unit 203 acquires the virtual network information (step S104).

[0059] The solution-finding unit 204 sets the values ​​of the parameters γ and ε (step S105). Note that the parameters γ and ε can be set to any non-negative value.

[0060] The solution-finding unit 204 formulates the distributional robust optimization problem shown in the above equation (4) using the parameters γ and ε, the physical network information, and the virtual network information, and then reduces the problem to an integer linear programming problem shown in the above equations (8a) to (8f) (step S106).

[0061] The solution-finding unit 204 solves the integer linear programming problem shown in the above equations (8a) to (8f) using an existing solver or the like (step S107), thereby obtaining an optimal solution to this integer linear programming problem.

[0062] The control unit 205 determines whether the optimal solution is included (x ｎｓ , y ｌｓ ) (n∈N, s∈S, l∈L) to control VM placement and routes on the physical network (step S108). As a result, a virtual network that realizes the VM placement and routes is embedded on the physical network, and a virtual network for the time period is constructed.

[0063] <Summary> As described above, the virtual network control device 10 according to this embodiment defines an uncertainty set D that represents uncertainty regarding the probability distribution of renewable energy power generation, and then formulates a distributional robust optimization problem using this uncertainty set D. Furthermore, the virtual network control device 10 according to this embodiment reduces the distributional robust optimization problem to a solvable class and then uses an existing solver or the like to find an optimal solution. This makes it possible to realize a virtual network allocation that can utilize renewable energy power generation robustly despite the uncertainty, even if there is uncertainty in the probability distribution of renewable energy power generation. Therefore, for example, even if the probability distribution of actual renewable energy power generation differs from the probability distribution of renewable energy power generation modeled during control execution, it is possible to realize a virtual network allocation that ensures robustness in terms of utilizing renewable energy power generation.

[0064] Here, the degree to which the uncertainty of the probability distribution of renewable energy power generation is taken into account is adjusted by the parameter ε. Increasing the value of the parameter ε is expected to result in a more robust allocation result. ｎ （ｍ）When the number of samples M is small, it is expected that the allocation result will clearly show the guarantee of robustness by taking into full consideration the uncertainty of the probability distribution due to the number of samples. On the other hand, when the number of samples M is large, the conservatism for guaranteeing robustness will be reduced, and the sampled renewable energy power generation amount ξ ｎ （ｍ） Therefore, the distributional robust optimization of the proposed method is expected to obtain results close to those of stochastic programming, which fully utilizes the characteristics of ｎ （ｍ） By making the conservatism associated with the robustness guarantee flexible according to the size of the sample number M, it can be said that this algorithm can realize VM placement and route determination that can effectively utilize renewable energy power generation.

[0065] In the above embodiment, the virtual network control device 10 has the control unit 205. However, for example, a device or system other than the virtual network control device 10 may have the control unit 205. In this case, the optimal solution calculated by the virtual network control device 10 may include (x ｎｓ , y ｌｓ ) (n∈N, s∈S, l∈L) is transmitted to a device or system having a control unit 205, and a virtual network is constructed in the device or system.

[0066] The present invention is not limited to the above-described specifically disclosed embodiments, and various modifications, changes, and combinations with known technologies are possible without departing from the scope of the claims.

[0067] [References] Reference 1: Mohajerin Esfahani, Peyman, and Daniel Kuhn. "Data-driven distributionally robust optimization using the Wasserstein metric: Performance guarantees and tractable reformulations." Mathematical Programming 171(1-2), 115-166, 2018.

[0068] REFERENCE SIGNS LIST 10 Virtual network control device 101 Input device 102 Display device 103 External I / F 103a Recording medium 104 Communication I / F 105 RAM 106 ROM 107 Auxiliary storage device 108 Processor 109 Bus 201 Renewable energy power generation amount acquisition unit 202 Physical network information acquisition unit 203 Virtual network information acquisition unit 204 Solution unit 205 Control unit

Claims

1. A control device for embedding a virtual network in a physical network, comprising: an acquisition unit that acquires samples of power generation at physical nodes included in the physical network, the physical nodes having power sources with uncertainty in the probability distribution of the power generation; and a calculation unit that defines an uncertainty set representing uncertainty regarding the probability distribution based on the samples using a predetermined distance function, and calculates the allocation of the virtual network so as to optimize the expected value of a predetermined objective function within the range of the uncertainty set.

2. The control device according to claim 1, wherein the distance function is the Wasserstein distance, and the uncertainty set is defined as a set in which the Wasserstein distance between the true probability distribution of the power generation amount and a sample distribution based on the samples is less than or equal to a predetermined parameter ε≧0.

3. The control device according to claim 1, wherein the objective function is a function that calculates the sum of the total surplus power in the physical node where the power source is located and the total delay time of the service provided by the virtual network.

4. A control device as claimed in any one of claims 1 to 3, wherein the calculation unit calculates the allocation of the virtual network by solving an integer linear programming problem equivalent to an optimization problem for optimizing an expected value of a predetermined objective function within the uncertainty set.

5. The control device described in claim 4, wherein the calculation unit calculates the allocation of the virtual network using constraints including at least the conditions that the traffic of the virtual network is equal to or less than the maximum capacity of each physical link and that the delay time in the virtual network is equal to or less than a predetermined upper limit value.

6. The control device according to claim 5, wherein the calculation unit calculates whether or not a virtual machine is placed on each physical node and whether or not a path of the virtual network is placed on each physical link as the allocation of the virtual network.

7. A control method in which a control device for embedding a virtual network in a physical network executes: an acquisition procedure for acquiring samples of power generation at physical nodes included in the physical network, the physical nodes including a power source having uncertainty in the probability distribution of the power generation; and a calculation procedure for defining an uncertainty set representing uncertainty regarding the probability distribution based on the samples using a predetermined distance function, and calculating an allocation of the virtual network so as to optimize the expected value of a predetermined objective function within the range of the uncertainty set.

8. A program that causes a control device for embedding a virtual network in a physical network to execute: an acquisition procedure for acquiring samples of power generation at physical nodes included in the physical network, where a power source having uncertainty in the probability distribution of the power generation is present; and a calculation procedure for defining an uncertainty set representing uncertainty regarding the probability distribution based on the samples using a predetermined distance function, and calculating the allocation of the virtual network so as to optimize the expected value of a predetermined objective function within the range of the uncertainty set.

Images