Network design optimization device, network design optimization method, and program

Abstract

This network design optimization device comprises: a first acquisition unit that acquires network traffic information and an objective function relating to network design; a second acquisition unit that acquires network resource information and a cost model relating to the network; and a network design calculation unit that, on the basis of the traffic information acquired by the first acquisition unit and the resource information and cost model acquired by the second acquisition unit, calculates a network design that minimizes the objective function acquired by the first acquisition unit.

Description

Network Design Optimization Device, Network Design Optimization Method, and Program

[0001] The present invention relates to a network design optimization device, a network design optimization method, and a program.

[0002] With the increase in network traffic demand, conventionally, it has been considered to expand the transmission capacity by spatial multiplexing or multi-band. Along with this, an increase in the cost of optical cross-connects (OXCs, Optical Cross-Connects) serving as relay nodes in the network has become an issue. As technologies for optical cross-connects assuming spatial multiplexing, Non-Patent Document 1 and Non-Patent Document 2 are known. Also, as a technology for optical cross-connects assuming multi-band, Non-Patent Document 3 is known.

[0003] "Hierarchical Optical Path Network Design Algorithm Considering Waveband Add / Drop Ratio Constraint", Hai-Chau Le, Hiroshi Hasegawa, and Ken-ichi Sato, J. OPT. COMMUN. NETW. / VOL. 2, NO. 10 / OCTOBER 2010 "Technoeconomic analysis of spatial channel networks (SCNs): benefits from spatial bypass and spectral grooming [Invited]", Masahiko Jinno, Yu Asano, Yoshiki Azuma, Takahiro Kodama, and Riku Nakai, Vol. 13, No. 2 / February 2021 / Journal of Optical Communications and Networking "Double-decker CDC-ROADM node for multi-band network with wavelength band granularity", Kenya Suzuki, Masashi Ota, Yoshie Morimoto, Keita Yamaguchi, Fukutaro Hamaoka, Shuto Sugawara, Takeo Sasai, Takayuki Kobayashi, Masanori Nakamura, Satomi Katayose, Takeshi Umeki, Daisuke Ogawa, Yiran Ma, Stefano Camatel, Mitsunori Fukutoku, Yutaka Miyamoto, and Osamu Moriwaki, OFC 2024

[0004] Traditionally, spatial multiplexing has faced the challenge of increasing the number of wavelength-selective switches due to the increase in the number of paths. On the other hand, multiband applications have faced the challenge of increased device costs. To address these challenges, in spatial multiplexing studies, efforts have been made to reduce node costs by using hierarchical switching with switches of different granularities. However, for multiband optical cross-connects, node cost reduction through hierarchical switching has not been considered.

[0005] In other words, conventional technologies were not capable of reducing network node costs and designing low-cost spatially multiplexed and multiband networks.

[0006] The present invention aims to provide a technology that reduces network node costs and enables the design of low-cost spatially multiplexed and multiband networks.

[0007] One aspect of the present invention is a network design optimization device comprising: a first acquisition unit that acquires network traffic information and an objective function relating to network design; a second acquisition unit that acquires network resource information and a cost model relating to the network; and a network design calculation unit that calculates a network design that minimizes the objective function acquired by the first acquisition unit, based on the traffic information acquired by the first acquisition unit, the resource information acquired by the second acquisition unit, and the cost model.

[0008] Another aspect of the present invention is a network design optimization method comprising: a first acquisition process for acquiring network traffic information and an objective function relating to network design; a second acquisition process for acquiring network resource information and a cost model relating to the network; and a network design calculation process for calculating a network design that minimizes the objective function acquired in the first acquisition process, based on the traffic information acquired in the first acquisition process and the resource information and cost model acquired in the second acquisition process.

[0009] According to the present invention, it is possible to reduce the node cost of a network and design a low-cost spatially multiplexed and multiband network.

[0010] This is a schematic diagram of a network design system according to the first embodiment of the present invention (and the second embodiment described later). This is a diagram showing an example of a network topology optimized by the network design system. This is a diagram showing the specific configuration of the nodes shown in Figure 1. This is a diagram showing an example of the configuration of band granularity paths and wavelength granularity paths in the first embodiment of the present invention (and the second embodiment described later). This is a table showing an example of a cost model used in the first embodiment of the present invention (and the second embodiment described later). This is a diagram showing an example of a network topology. This is a graph showing the relationship between the number of connection requests and network design costs when all paths are considered as candidates and when only the shortest path is considered as a candidate when allocating band granularity paths. This is a diagram showing an example of a network topology. This is a graph showing the relationship between the number of connection requests and network design costs when network design is optimized by integer linear programming for hierarchical OXC (this application) and WXC (conventional). This is a graph showing the relationship between the number of connection requests and the number of WSS used when network design is optimized by integer linear programming for hierarchical OXC (this application) and WXC (conventional). This is a flowchart showing the processing of the approximation algorithm used for network design calculation by the NW design calculation unit of the network design optimization device according to the second embodiment of the present invention. This is a flowchart of the first pathfinding algorithm performed in step S105 of Figure 11. This is a diagram showing the auxiliary graph (initial state) used in step S202 of Figure 12. This is a flowchart of the second pathfinding algorithm performed in step S110 of Figure 11. This is a diagram showing the auxiliary graph (initial state) used in step S302 of Figure 14. This is a graph showing the relationship between the number of connection requests per node pair and the network design cost when using an approximation algorithm and integer linear programming. This is a graph showing the relationship between the number of connection requests per node pair and the network design cost when the network design is optimized using the approximation algorithm for hierarchical OXC (this application) and WXC (conventional). This is a graph showing the relationship between the number of connection requests per node pair and the number of WSS used when the network design is optimized using the approximation algorithm for hierarchical OXC (this application) and WXC (conventional).This figure shows the network configuration when network design is performed using hierarchical OXC (the present application). This figure shows the network configuration when network design is performed using WXC (conventional). This table shows the calculated path configurations in an embodiment using hierarchical OXC (the present application) and a comparative example using WXC (conventional).

[0011] Hereinafter, the first and second embodiments of the present invention will be described with reference to the drawings. First, the first embodiment of the present invention will be described.

[0012] [First Embodiment] First, a first embodiment of the present invention will be described. Figure 1 is a schematic diagram of a network design system 100 according to the first embodiment of the present invention. The network design system 100 comprises a network management device 10, a network design optimization device 20, and a node 30.

[0013] The network design optimization device 20 is connected to the network management device 10 and node 30 by a wired network. The network design optimization device 20 includes a database 21 and a network design optimization unit 22. The network design optimization unit 22 includes an initial setting unit 221 and a network design calculation unit 222 (also referred to as the network design calculation unit). The initial setting unit 221 includes a first acquisition unit 2211, a second acquisition unit 2212, and an output unit 2213. Node 30 includes a BXC (Band Cross-Connect) 31 and a WXC (Wavelength Cross-Connect) 32.

[0014] First, the network management device 10 transmits traffic information and the NW design objective function to the first acquisition unit 2211 of the initial setup unit 221 of the network design optimization device 20. In other words, the first acquisition unit 2211 of the initial setup unit 221 of the network design optimization device 20 acquires traffic information and the NW design objective function from the network management device 10.

[0015] Database 21 pre-stores resource information and cost models. The second acquisition unit 2212 of the initial setup unit 221 of the network design optimization device 20 acquires resource information and cost models by reading them from database 21. Next, the output unit 2213 of the initial setup unit 221 of the network design optimization device 20 generates weight setting information based on the traffic information and network design objective function acquired by the first acquisition unit 2211, and the resource information and cost models acquired by the second acquisition unit 2212.

[0016] Next, the output unit 2213 of the initial setup unit 221 of the network design optimization device 20 outputs traffic information, resource information, and weight setting information to the NW design calculation unit 222. Then, the NW design calculation unit 222 creates band granularity path setting information and wavelength granularity path setting information based on the traffic information, resource information, and weight setting information output from the initial setup unit 221, and outputs it to the database 21 for storage. The NW design calculation unit 222 also transmits the created band granularity path setting information and wavelength granularity path setting information to the network management device 10.

[0017] Next, the network management device 10 transmits the band granularity path setting information received from the NW design calculation unit 222 to the BXC 31 of node 30. The network management device 10 also transmits the wavelength granularity path setting information received from the NW design calculation unit 222 to the WXC 32 of node 30.

[0018] Although there are many nodes on the network of the network design system 100, Figure 1 shows only one node 30, representing all of those nodes.

[0019] Figure 2 shows an example of a network topology to be optimized by the network design system 100, and is part of the known topology model JPN25. Figure 2 shows the case where the network design system 100 performs network design on a network topology consisting of nodes N111, N112, N113, N114, N115, N116, N117, N118, N119, N120, and N121. These nodes N111 to N121 are connected by a wired network.

[0020] Figure 3 shows the specific configuration of node 30 shown in Figure 1. Node 30 is equipped with Hierarchical Optical Path Cross-Connects (HOXC) 30-1, 30-2, and 30-3 for each wavelength band. Hierarchical Optical Path Cross-Connect 30-1 is an element for the S-band (Short-wavelength-band) from 1460 nm to 1530 nm. Hierarchical Optical Path Cross-Connect 30-2 is an element for the C-band (Conventional-band) from 1530 nm to 1565 nm. Hierarchical Optical Path Cross-Connect 30-3 is an element for the L-band (Long-wavelength-band) from 1565 nm to 1625 nm.

[0021] Each of the multi-layer optical cross-connectors 30-1, 30-2, and 30-3 comprises a BXC 31, a WXC 32, band demultiplexers 33-1 to 33-M (where M is an integer of 2 or more), amplifiers 34-1-1, 34-1-2, 34-1-3, ..., 34-M-1, 34-M-2, 34-M-3, band multiplexers 35-1, ..., 35-M, an add-drop unit 36, and amplifiers 37-1 to 37-N (where N is an integer of 2 or more).

[0022] In Figure 3, band demultiplexers 33-1 to 33-M at node 30 receive signals from M paths. Each of the band demultiplexers 33-1 to 33-M splits the input signal into three wavelength bands and outputs them to amplifiers 34-1-1, 34-1-2, 34-1-3, ..., 34-M-1, 34-M-2, and 34-M-3. Amplifiers 34-1-1, 34-1-2, 34-1-3, ..., 34-M-1, 34-M-2, and 34-M-3 amplify the signals output from the band demultiplexers 33-1 to 33-M and output them to BXC 31.

[0023] Here, amplifiers 34-1-1 to 34-M-1 output the amplified signal to the BXC31 of the multi-layer optical cross-connect 30-1. Amplifiers 34-1-2 to 34-M-2 also output the amplified signal to the BXC31 of the multi-layer optical cross-connect 30-2. Amplifiers 34-1-3 to 34-M-3 also output the amplified signal to the BXC31 of the multi-layer optical cross-connect 30-3. The multi-layer optical cross-connects 30-1 to 30-3 output the signal processed by the band cross-connect to the band multiplexer 35-1 to 35-M or WXC32.

[0024] The band multiplexers 35-1 to 35-M combine and output the signals output from the multi-layer optical cross-connectors 30-1 to 30-3. The add-drop section 36 in the multi-layer optical cross-connectors 30-1 to 30-3 inserts or extracts signals of a predetermined wavelength from the signal input to the WXC 32. The WXC 32 processes the signal output from the BXC 31 using wavelength cross-connection and outputs it to the amplifiers 37-1 to 37-N. The amplifiers 37-1 to 37-N amplify the signal output from the WXC 32 and output it to the BXC 31 of the multi-layer optical cross-connector 30-1. Similarly, wavelength cross-connection processing and output to the BXC are performed in the multi-layer optical cross-connectors 30-2 and 30-3.

[0025] Furthermore, at node 30 in Figure 3, cross-connect processing of the wavelength band granularity signal D11 is performed at BXC31 of the multi-layer optical cross-connect 30-1. Also, at node 30 in Figure 3, the wavelength granularity signal (λ) is processed at WXC32 of the multi-layer optical cross-connect 30-1.１ , λ ２ , λ ３ Cross-connect processing is performed on signal D12 (...).

[0026] Next, the problem setting in the network design calculation method by the NW design calculation unit 222 (see Figure 1) of the network design optimization unit 22 of the network design optimization device 20 in the first embodiment (and second embodiment) of the present invention will be described. The NW design calculation unit 222 configures band granularity paths and wavelength granularity paths based on multiband hierarchical OXC (Optical Cross-Connect) and determines the network design.

[0027] As shown in Figure 3, the WXC 32 of node 30 is equipped with input-side WSS (Wavelength Selective Switch) 321-1 to 321-N and output-side WSS 322-1 to 322-N. In Figure 4, the symbol P11 indicates a wavelength granularity path that bypasses all WXCs of the relay nodes. The symbol P12 indicates a wavelength granularity path that bypasses one of the relay nodes and grooms with one. The symbol P13 indicates a wavelength granularity path that grooms with all relay nodes.

[0028] The band-granularity path has WSS on both the input and output sides, and at the relay node, it bypasses WXC (i.e., it does not pass through WXC, but only BXC) to reduce the number of required WSS. For example, in Figure 4, the wavelength-granularity path P11 bypasses the relay node, thereby reducing the WSS in the relay node's WXC.

[0029] According to the first embodiment (and the second embodiment) of the present invention, as shown in Figure 4, wavelength-grained paths P11, P12, and P13 are housed in a band-grained path, and by performing grooming (i.e., switching band-grained paths) at a relay node as needed, routing becomes possible via multiple band-grained paths.

[0030] If we denote the wavelength band used to transmit the signal as b, and we consider three wavelength bands, S, C, and L, then b is expressed by the following equation (1).

[0031]

[0032] That is, the wavelength band (band) for transmitting signals is any one of the S band, C band, and L band. Note that the band multiplexers 35-1 to 35-M shown in FIG. 3 have ω as a weight. （ＤＥ）ＭＵＸ Further, the optical fibers 42-1 to 42-M shown in FIG. 3 have ω as a weight. ｆｉｂｅｒ Further, the input port and output port of the BXC 31 shown in FIG. 3 have ω as a weight. ＳＷ　ｐｏｒｔ have.

[0033] Further, the WSSs 321-1 to 322-N shown in FIG. 3 are also referred to as b-band WSSs and have ω as a weight. ＷＳＳ（ｂ） Further, the amplifiers 34-1-1 to 34-M-3 shown in FIG. 3 are also referred to as b-band amplifiers and have ω as a weight. ａｍｐ（ｂ） Further, the transponders installed in the add-drop section 36 shown in FIG. 3 are also referred to as b-band transponders and have ω as a weight. ＴＲｘ（ｂ） have.

[0034] FIG. 5 is a table showing an example of a cost model used in the first embodiment (and the second embodiment) of the present invention. In the table shown in FIG. 5, values indicating the relationship between the S band, C band, L band, and each weight are determined. For example, the value of the weight ω ＷＳＳ（ｂ） is 7.5 when b is S (that is, the S band), 5 when b is C (that is, the C band), and 6 when b is L (that is, the L band).

[0035] Next, the integer linear programming (ILP, Integer Linear Programming) used when the NW design calculation unit 222 performs network design calculation according to the first embodiment of the present invention will be described. Here, the case where the following conditions are given as a premise that the NW design calculation unit 222 uses integer linear programming will be described.

[0036]

[0037] Here, for example, the cost model shown in FIG. 5 is used as the cost model.

[0038] Furthermore, we will explain the case where the following parameters are used when the NW design calculation unit 222 uses integer linear programming.

[0039]

[0040] Furthermore, we will explain the case where the NW design calculation unit 222 uses integer linear programming and the following variables are used.

[0041]

[0042] Furthermore, when the NW design calculation unit 222 uses integer linear programming, it uses, for example, the objective function (2) and constraint equations (3) to (10) shown below.

[0043] The objective function (2) below is used to minimize the equipment costs required for the network configuration.

[0044]

[0045] Furthermore, the following constraint equation (3) is used to assign one band and one wavelength to each connection request.

[0046]

[0047] Furthermore, the following constraint equation (4) shows the flow rate conservation law for wavelength particle size paths.

[0048]

[0049] Furthermore, the following constraint equation (5) is used to ensure that the wavelengths of multiple wavelength granularity paths do not overlap within the same band granularity path.

[0050]

[0051] Furthermore, the following constraint equation (6) is used to construct the band-granularity path to which the connection request is assigned.

[0052]

[0053] Furthermore, the following constraint equation (7) is used to assign one candidate path k to the band granularity path.

[0054]

[0055] Furthermore, the following constraint equation (8) is used to ensure that each band granularity path corresponds to only one virtual link.

[0056]

[0057] Furthermore, the following constraint equation (9) is used to count the number of band b required for each link.

[0058]

[0059] Furthermore, the following constraint equation (10) is used to count the number of fibers required for each link.

[0060]

[0061] The key points of the formulation of the integer linear programming method used by the NW design calculation unit 222 according to the first embodiment of the present invention are as follows: The wavelength particle size path is searched exhaustively using the flow rate conservation law, and the band particle size path is searched from pre-given path candidates.

[0062] The objective function (2) is the network design cost (sum of device costs such as fiber, band demultiplexer, amplifier, BXC switch port, WSS, transponder, etc.), and different cost weights can be considered for each band. In addition to network design cost, the objective function can be arbitrarily set depending on the target of minimization. For example, by changing the coefficients of the objective function (2) described above, it is possible to extend the optimization to minimize transponder cost, the number of WSSs, the number of fibers, etc.

[0063] In the integer linear programming method used by the NW design calculation unit 222 according to the first embodiment of the present invention when performing network design calculations, two optimizations are performed simultaneously: a band granularity path and a wavelength granularity path. Therefore, the calculation time is longer compared to conventional RWA (Routing and Wavelength Assignment), but by narrowing down the candidate paths K for the band granularity path and reducing the search range, the calculation time can be shortened without significantly compromising optimality.

[0064] Next, we will explain the optimality and computation time obtained by reducing the search range when the NW design calculation unit 222 according to the first embodiment of the present invention uses integer linear programming to perform network design calculations. Here, we will explain the case using the physical topology shown in Figure 6. In Figure 6, the network topology is composed of nodes N211, N212, N213, N214, and N215. The distance of each link is 50 km.

[0065] Figure 7 is a graph showing the relationship between the number of connection requests and network design costs when all paths are considered as candidates during band-granularity path allocation (all paths) and when only the shortest path is considered as a candidate (shortest paths). Note that |W| = 5, the number of connection requests |R| = 1 to 20, connection requests are generated uniformly and randomly, and the objective function and cost model are the aforementioned equation (2) and Figure 5. When only the shortest path is considered as a path candidate, as shown in Figure 7, the increase rate of the optimal value is sufficiently small, averaging about 0.1%, while the calculation time is reduced by an average of about 23%, and by more than 70% at its peak.

[0066] Next, we will explain the effects on network design obtained by using integer linear programming when the NW design calculation unit 222 according to the first embodiment of the present invention performs network design calculations. Here, we will explain the case using the physical topology shown in Figure 8. In Figure 8, the network is configured in a 2x3 grid with nodes N311, N312, N313, N314, N315, and N316. The distance of each link is 50 km.

[0067] Figure 9 is a graph showing the relationship between the number of connection requests and the network design cost when the network design is optimized using integer linear programming for hierarchical OXC (this application) and WXC (conventional). Figure 10 is a graph showing the relationship between the number of connection requests and the number of WSS used when the network design is optimized using integer linear programming for hierarchical OXC (this application) and WXC (conventional).

[0068] Figures 9 and 10 show the average values ​​calculated over 10 trials, where |W| = 5, the number of connection requests is |R| = 1 to 30, connection requests are generated uniformly and randomly, and the cost model and objective function are those shown in Figure 5 and Equation (2). Although the cost reduction effect is limited due to the small traffic scale, as shown in Figure 9, the network design cost of the hierarchical OXC (in this application) is reduced by approximately 0.5% on average compared to WXC (conventional). Furthermore, as shown in Figure 10, the number of WSSs used in the hierarchical OXC (in this application) is reduced by approximately 11.3% on average compared to WXC (conventional).

[0069] In the network design optimization device 20 according to the first embodiment described above, the first acquisition unit 2211 acquires network traffic information and an objective function related to network design. The second acquisition unit 2212 acquires network resource information and a cost model related to the network. The network design calculation unit 222 calculates a network design that minimizes the objective function acquired by the first acquisition unit 2211 by using integer linear programming based on the traffic information acquired by the first acquisition unit 2211 and the resource information and cost model acquired by the second acquisition unit 2212. This makes it possible to design a network that assumes spatial multiplexing and multiband operation at low cost.

[0070] [Second Embodiment] Next, a second embodiment of the present invention will be described. Note that the second embodiment is the same as the first embodiment, so the explanation will be omitted. In the first embodiment, the NW design calculation unit 222 of the network design optimization device 20 calculated a network design that minimizes the objective function by using integer linear programming, but in the second embodiment, an approximation algorithm is used instead of integer linear programming.

[0071] Figure 11 is a flowchart showing the processing of an approximation algorithm used by the NW design calculation unit 222 of the network design optimization device 20 according to the second embodiment of the present invention for network design calculation. First, the NW design calculation unit 222 processes G(V,E), R=R ０The cost model is input (step S101). In G(V,E), G represents a graph, V represents a set of points, and V represents a set of edges connecting two points. R represents a set of connection requests, and R ０ This represents the initial connection request.

[0072] Next, the NW design calculation unit 222 performs the following processing in step S102.

[0073]

[0074] In (s, d), s represents the starting node and d represents the ending node.

[0075] Next, the NW design calculation unit 222 performs the following processing in step S103.

[0076]

[0077] Next, the NW design calculation unit 222, n ｓ,ｄ However, it is determined whether or not it is above a predetermined threshold TH (step S104). ｓ,ｄ However, if it is not above a predetermined threshold TH, the NW design calculation unit 222 performs the process of step S107 described later. ｓ,ｄ However, if the threshold TH is greater than or equal to a predetermined threshold, the NW design calculation unit 222 constructs a band granularity path using a first path search algorithm (step S105). The first path search algorithm will be described later with reference to Figure 12.

[0078] Next, the NW design calculation unit 222 removes the connection request assigned to the band granularity path from R, and n ｓ，ｄ n ｓ，ｄ -min(n) ｓ，ｄ Update to |W|) (step S106).

[0079] Next, the NW design calculation unit 222 determines whether all (s,d) have been extracted (step S107). If not all (s,d) have been extracted, the NW design calculation unit 222 repeats the process in step S103. On the other hand, if all (s,d) have been extracted, the NW design calculation unit 222 performs the following process in step S108. Note that the band particle size path is configured by performing the processes in steps S102 to S107.

[0080] Next, the NW design calculation unit 222 constructs a wavelength granularity path using a second pathfinding algorithm (step S109). The second pathfinding algorithm will be described later with reference to Figure 14. Next, the NW design calculation unit 222 removes the assigned connection request r from R (step S110).

[0081] Next, the NW design calculation unit 222 determines whether the set of connection requests R is empty or not (step S111). If the set of connection requests R is not empty, the NW design calculation unit 222 performs the process in step S108 again. On the other hand, if the set of connection requests R is empty, the NW design calculation unit 222 terminates the process shown in the flowchart in Figure 11. Note that the wavelength granularity path is configured by performing the processes in steps S108 to S111.

[0082] Figure 12 is a flowchart showing the first pathfinding algorithm performed in step S105 of Figure 11. First, the NW design calculation unit 222 determines the node pair (s, d) and the number of connection requests n ｓ，ｄ The input is made (step S201). Next, in step S202, the NW design calculation unit 222 performs the following processing.

[0083]

[0084] In step S202 of Figure 12, the NW design calculation unit 222 constructs a weighted auxiliary graph consisting of multiple layers corresponding to network resources of different granularities. Also in step S202 of Figure 12, the NW design calculation unit 222 assigns weights to each side of the auxiliary graph based on device costs.

[0085] Figure 13 shows the auxiliary graph (initial state) used in step S202 of Figure 12. This auxiliary graph is used to construct the band particle size path. In Figure 13, point p １ , p ２ , p ３ , p ４ This belongs to the band-grained path layer. Also, in Figure 13, point b １ ｂ , b ２ ｂ , b ３ ｂ , b ４ ｂ It belongs to the band layer. Also, in Figure 13, point f １ ｂ , f ２ ｂ , f ３ ｂ , f ４ ｂ This belongs to the fiber layer (physical topology). Note that point p １ , point b １ ｂ , point f １ ｂ These belong to the same node in the physical topology.

[0086] Returning to the explanation of Figure 12, the NW design calculation unit 222 performs the following processing in step S203.

[0087]

[0088] The NW design calculation unit 222 determines the Dijkstra path using the Dijkstra method, a known method for finding the shortest path starting from a certain point on the graph. The NW design calculation unit 222 determines the weight of the Dijkstra path (edge) using the following formula.

[0089]

[0090] Next, the NW design calculation unit 222 performs the following processing in step S204.

[0091]

[0092] Furthermore, the NW design calculation unit 222 updates the auxiliary graph based on the following information.

[0093]

[0094] Figure 14 is a flowchart of the second route search algorithm performed in step S109 of Figure 11. First, the NW design calculation unit 222 receives the connection request r = (s, d) as input (step S301). Next, in step S302, the NW design calculation unit 222 performs the following processing.

[0095]

[0096] Figure 15 shows the auxiliary graph (initial state) used in step S302 of Figure 14. This auxiliary graph is used to construct the wavelength granularity path. In Figure 15, point p １ , p ２ , p ３ , p ４ This belongs to the wavelength-grain size path layer. Also, in Figure 15, point w １ ｂ，λｉ , point w ２ ｂ，λｉ , point w ３ ｂ，λｉ , point w ４ ｂ，λｉ It belongs to the wavelength layer. Also, in Figure 15, point b １ ｂ，λｉ , point b ２ ｂ，λｉ , point b ３ ｂ，λｉ , point b ４ ｂ，λｉ It belongs to the band layer. Also, in Figure 15, point f １ ｂ，λｉ , point f ２ ｂ，λｉ , point f ３ ｂ，λｉ , point f ４ ｂ，λｉ This belongs to the fiber layer (physical topology). Note that point p １ , point w １ ｂ，λｉ , point b １ ｂ，λｉ , point f １ ｂ，λｉThese belong to the same node in the physical topology.

[0097] Returning to the explanation of Figure 14, the NW design calculation unit 222 calculates the weights of the Dijkstra paths in the auxiliary graph AG, selects the path with the smallest weight, and constructs a wavelength-granularity path (step S303). The weights of the Dijkstra paths (edges) are determined based on the following formula.

[0098]

[0099] Next, the NW design calculation unit 222 updates the auxiliary graph AG according to the selected route (step S304). The update of the auxiliary graph AG is performed based on the following:

[0100]

[0101] Next, we will explain the key points when using the approximation algorithm. When using the approximation algorithm, for each connection request, it is determined whether to groom an existing band-granularity path or add new resources, and resources are allocated accordingly to minimize the additional cost. This algorithm can be used by changing the auxiliary graph, for example, the conventional method of finding Dijkstra paths on an auxiliary graph (the traffic grooming method described in "Dynamic Traffic Grooming in Elastic Optical Networks", Shuqiang Zhang, Charles Martel, and Biswanath Mukherjee, IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 31, NO. 1, JANUARY 2013).

[0102] Furthermore, the objective function can be arbitrarily set depending on the cost equation to be minimized. In this case, when using integer linear programming according to the first embodiment, the objective function is changed, while when using the approximation algorithm, the weight settings of the edges of the auxiliary graph are changed.

[0103] When constructing a band-granularity path, the node pair (s,d) with the smallest number of hops for the shortest path is selected in order. If the number of connection requests between (s,d) is equal to or greater than a predetermined threshold TH, a band-granularity path is constructed and the connection requests are allocated together. When constructing a band-granularity path, the calculation is performed on an auxiliary graph using Dijkstra's method. When constructing a wavelength-granularity path, the remaining connection requests are selected in order from those with the largest number of hops for the shortest path, and a wavelength-granularity path is constructed and the connection requests are allocated.

[0104] When constructing wavelength-granularity paths, the calculation is performed using Dijkstra's method on an auxiliary graph. To reduce costs by optimizing the search order, when constructing band-granularity paths in steps S102 to S107 of the flowchart in Figure 11, the node pair with the fewest hops for the shortest path is allocated first, so that low-cost bands are used for band-granularity paths with a small number of hops. Furthermore, when constructing wavelength-granularity paths in steps S108 to S111 of the flowchart in Figure 11, the connection request with the largest number of hops for the shortest path is allocated first, thereby improving the efficiency of path allocation.

[0105] Next, we will explain the effects obtained by using an approximation algorithm when the NW design calculation unit 222 according to the second embodiment of the present invention performs network design calculations. Here, we will explain the case using the physical topology shown in Figure 6.

[0106] Figure 16 is a graph showing the relationship between the number of connection requests per node pair and network design cost when using the approximation algorithm and integer linear programming (ILP). In Figure 16, to investigate the optimality of the approximation algorithm, the results calculated using the approximation algorithm are compared with the optimal solution obtained using integer linear programming. Here, |W| = 5, connection requests are generated uniformly and randomly with an average number of connection requests between each node ranging from 1 to 5 (|R| = 110 to 550), and the cost model and objective function are calculated using Figure 5 and equation (2) described above. The difference from the optimal value obtained using integer linear programming is approximately 4.2%, indicating that the approximation algorithm can obtain a near-optimal solution with sufficient performance.

[0107] Next, other effects obtained by using an approximation algorithm when the NW design calculation unit 222 according to the second embodiment of the present invention performs network design calculations will be explained. Here, the case using the physical topology shown in Figure 2 will be explained. Figure 2 is a part of the known network model JPN25, and the distance between each node is set according to this model (the average distance between nodes is 88.2 km).

[0108] Figure 17 is a graph showing the relationship between the number of connection requests per node pair and the network design cost when the network design is optimized using an approximation algorithm for hierarchical OXC (this application) and WXC (conventional). Figure 18 is a graph showing the relationship between the number of connection requests per node pair and the number of WSS used when the network design is optimized using an approximation algorithm for hierarchical OXC (this application) and WXC (conventional).

[0109] In Figures 17 and 18, W=16, connection requests are generated uniformly and randomly with a number of connection requests between each node ranging from 1 to 20 (|R| = 440 to 2200), and the cost model and objective function are those described in Figure 5 and Equation (2), with the average value of 10 calculations used. As shown in Figure 17, the network design cost of the hierarchical OXC (present application) is reduced by up to approximately 14% compared to WXC (conventional). Also, as shown in Figure 18, the number of WSSs used in the hierarchical OXC (present application) is reduced by an average of approximately 71.7% compared to WXC (conventional).

[0110] In the network design optimization device 20 according to the second embodiment described above, the first acquisition unit 2211 acquires network traffic information and an objective function related to network design. The second acquisition unit 2212 acquires network resource information and a cost model related to the network. The network design calculation unit 222 calculates a network design that minimizes the objective function acquired by the first acquisition unit 2211 by using an approximation algorithm based on the traffic information acquired by the first acquisition unit 2211 and the resource information and cost model acquired by the second acquisition unit 2212. This reduces the node cost of the network and makes it possible to design a low-cost spatially multiplexed and multiband network.

[0111] [Examples and Comparative Examples] Next, we will describe an example of a network design based on hierarchical OXC (the present application) and a comparative example of a network design based on WXC (conventional). Figure 19 shows the network configuration when a network design is performed using hierarchical OXC (the present application). Figure 20 shows the network configuration when a network design is performed using WXC (conventional).

[0112] Figure 21 is a table showing the calculation results of the network configurations in an embodiment using hierarchical OXC (present application) and a comparative example using WXC (conventional). In WXC, (v １ ,v ２ Things expressed in the form of v １ and v２ This indicates the physical link between them. In hierarchical OXC, (v １ ,v ２ , ..., v ｎ ) are represented by nodes v2, ..., v ｎ－１ Bypass v １ and v ｎ This indicates a virtual link between them.

[0113] In the case of the hierarchical OXC (present application) shown in Figure 21, |W| = 3, R = {(0,5),(4,5),(0,4),(3,5),(5,0),(1,4),(3,2)}, and the objective function was given by the following equation (11).

[0114]

[0115] As shown in Figures 19 and 21, in the hierarchical OXC (present application), the path for connection requests from node 0 to node 5 is R1, and its settings are b=C, w=3, p=[(0,3,4),(4,5)]. In other words, path R1 uses the third wavelength in the C band and passes through a band-granularity path (virtual link) connecting node 0, node 3, and node 4, and a band-granularity path (virtual link) connecting node 4 and node 5. Furthermore, WXC is bypassed at node 3, and grooming is performed by passing through WXC at node 4.

[0116] Furthermore, as shown in Figures 20 and 21, in WXC (conventional), the path for connection requests from node 0 to node 5 is R1, and its settings are b=L, w=3, p=[(0,3),(3,4),(4,5)]. In other words, path R1 uses wavelength 3 in the L band and passes through the physical links connecting node 0 and node 3, node 3 and node 4, and node 4 and node 5. Also, it passes through WXC at all the nodes it passes through.

[0117] In the case of the hierarchical OXC (present application) shown in Figure 19, the number of S-band WSSs used was 0, the number of C-band WSSs used was 8, and the number of L-band WSSs used was 4, resulting in a cost of 304. On the other hand, in the case of the WXC (conventional) shown in Figure 20, the number of S-band WSSs used was 0, the number of C-band WSSs used was 12, and the number of L-band WSSs used was 6, resulting in a cost of 336. In other words, in the case of the hierarchical OXC (present application), network design costs could be reduced compared to the case of the WXC (conventional).

[0118] Furthermore, at least some of the functions of the network management device 10, network design optimization device 20, and node 30 in the first and second embodiments described above may be implemented by a computer. In that case, the functions may be implemented by recording a program for implementing these functions on a computer-readable recording medium, loading the program recorded on this recording medium into a computer system, and executing it. Here, "computer system" includes hardware such as an OS (Operating System) and peripheral devices. Furthermore, "computer-readable recording medium" refers to portable media such as flexible disks, magneto-optical disks, ROMs (Read Only Memory), CD-ROMs, and storage devices such as hard disks built into a computer system. Moreover, "computer-readable recording medium" may also include those that dynamically hold programs for a short period of time, such as communication lines used when transmitting programs via networks such as the Internet or communication lines such as telephone lines, and those that hold programs for a certain period of time, such as volatile memory inside a computer system that acts as a server or client in such cases. Furthermore, the above-mentioned program may be for implementing some of the functions described above, or it may be a program that can implement the above-mentioned functions in combination with a program already recorded in the computer system, or it may be implemented using a programmable logic device such as an FPGA.

[0119] Although the first and second embodiments of this invention have been described in detail above with reference to the drawings, the specific configuration is not limited to these embodiments and includes designs and the like that do not depart from the spirit of this invention.

[0120] The present invention can be applied to network design optimization devices, network design optimization methods, and programs that require reducing network node costs and designing low-cost spatially multiplexed and multiband networks.

[0121] 10...Network management device, 20...Network design optimization device, 21...Database, 22...Network design optimization unit, 30...Node, 31...BXC, 32...WXC, 33-1 to 33-M...Band splitter, 34-1-1 to 34-M-3...Amplifier, 35-1 to 35-M...Band multiplexer, 36...Add / drop unit, 37-1 to 37-N...Amplifier, 321-1 to 321-N...Input WSS, 322-1 to 322-N...Output WSS, 42-1 to 42-M...Fiber, 100...Network design system, 221...Initial setup unit, 222...NW design calculation unit, 2211...First acquisition unit, 2212...Second acquisition unit, 2213...Output unit

Claims

1. A network design optimization device comprising: a first acquisition unit that acquires network traffic information and an objective function relating to network design; a second acquisition unit that acquires network resource information and a cost model relating to the network; and a network design calculation unit that calculates a network design that minimizes the objective function acquired by the first acquisition unit, based on the traffic information acquired by the first acquisition unit, the resource information acquired by the second acquisition unit, and the cost model.

2. The network design optimization device according to claim 1, wherein the network design calculation unit calculates a network design that minimizes the objective function acquired by the first acquisition unit by using integer linear programming or an approximation algorithm.

3. A network design optimization method comprising: a first acquisition process for acquiring network traffic information and an objective function relating to network design; a second acquisition process for acquiring network resource information and a cost model relating to the network; and a network design calculation process for calculating a network design that minimizes the objective function acquired in the first acquisition process, based on the traffic information acquired in the first acquisition process and the resource information and cost model acquired in the second acquisition process.

4. A program that causes the computer of a network design optimization device to execute the network design optimization method described in claim 3.

Images