Network design apparatus and program
The network design method optimizes three-stage Clos networks by dividing switches into IOCs and controlling blocking rates, achieving a larger network capacity and flexibility.
Patent Information
- Application Number
- JP2024112519
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2024-07-12
- Publication Date
- 2026-01-23
AI Technical Summary
Existing Clos network designs with two layers struggle to maximize network capacity while maintaining a specified allowable blocking rate, particularly when expanding to three layers.
A network design method that divides first and second stage switches into multiple IOCs and connects one or more third stage switches to each IOC, optimizing network capacity by controlling blocking rates in three-stage Clos networks.
Enables a larger network capacity and increased flexibility by optimizing blocking rates across three layers, allowing for more efficient utilization of switches and terminals.
Smart Images
Figure 2026011696000001_ABST
Abstract
Description
[Technical Field]
[0001] The present disclosure relates to a network design device, a network design method, and a network design program. [Background technology]
[0002] In Clos networks, there are technologies that theoretically guarantee the path blocking rate, increase the flexibility of the network configuration, and obtain larger network capacity (Non-Patent Documents 1-3). Non-Patent Document 1 discloses a technology that divides the network hierarchy into two layers, the input / output layer and the middle layer, divides the middle switches in the middle layer into two or more groups, and uses the middle switches in each group sequentially. [Prior art documents] [Non-patent literature]
[0003] [Non-Patent Document 1] H. Taka, T. Inoue, and E. Oki, "Design of twisted and folded Clos network with guaranteeing admissible blocking probability," IEEE Networking Letters, vol. 5, no. 4, pp. 265-269, 2023 [Non-patent document 2] H. Taka, T. Inoue, and E. Oki, "Twisted and folded Clos-network design model with two-step blocking probability guarantee," IEEE Networking Letters, vol. 6, no. 1,pp. 60-64, 2024, [Non-patent document 3] H. Taka, T. Inoue, and E. Oki, "Design model of a twisted and folded Clos network with multi-step grouped intermediate switches guaranteeing admissible blocking probability," Journal of Optical Communications and Networking, vol. 16, no. 3, pp. 328-341, 2024 Summary of the Invention [Problem to be solved by the invention]
[0004] Non-patent documents 1-3 disclose cases where the network hierarchy is divided into two layers: an input / output layer and an intermediate layer. However, if the network hierarchy is increased to three layers, it is expected that the degree of freedom will be further increased and a larger network capacity will be obtained.
[0005] The present disclosure has been made in consideration of the above circumstances, and an object of the present disclosure is to provide a technology that enables setting three network layers in a Clos network and achieving a larger network capacity. [Means for solving the problem]
[0006] A network design device according to one aspect of the present disclosure is a network design device for designing a network including a plurality of first stage switches belonging to a first stage and connected to a plurality of terminals, a plurality of second stage switches belonging to a second stage and connected to the plurality of first stage switches, and a plurality of third stage switches belonging to a third stage and connected to the plurality of second stage switches, the network design device including an acquisition unit for acquiring the number of ports of each switch, the total number of switches, terminal utilization rates of terminals connected to the plurality of first stage switches, and an allowable blocking rate that allows communication to be impossible between two terminals connected to the plurality of first stage switches, and a network design device for designing a network including a plurality of first stage switches belonging to a first stage and connected to the plurality of first stage switches, the network design device including an acquisition unit for acquiring the number of ports of each switch, the total number of switches, terminal utilization rates of terminals connected to the plurality of first stage switches, and an allowable blocking rate that allows communication to be impossible between two terminals connected to the plurality of first stage switches, the network design device being configured to acquire the number of ports of each switch, the total number of switches, and the terminal utilization rates, the network design device being configured to acquire the number of ports of each switch, the total number of switches, and the terminal utilization rates, the network design device being configured to acquire the allowable blocking rate that allows communication to be impossible between two terminals connected to the plurality of first stage switches, ... The network capacity is maximized without exceeding the allowable blocking rate. The calculation unit divides each of the first stage switches and the second stage switches into a plurality of IOCs, and one or more third stage switches are connected to each IOC. The network capacity is calculated by multiplying the number of terminals connected to one first stage switch, the number of IOCs, and the number of first stage switches in one IOC. Among the plurality of combinations of the variables of the number of terminals (n) connected to one first stage switch, the number of IOCs (k), and the number of first stage switches in one IOC (k'), the calculation unit calculates the combination with the largest network capacity, which satisfies the condition that the total probability of each blocking occurrence probability when the number of terminals is increased by one from the maximum number of terminals when the network is non-blocking does not exceed the allowable blocking rate.
[0007] A program according to one aspect of the present disclosure includes a computer for designing a network including a plurality of first stage switches belonging to a first stage and connected to a plurality of terminals, a plurality of second stage switches belonging to a second stage and connected to the plurality of first stage switches, and a plurality of third stage switches belonging to a third stage and connected to the plurality of second stage switches, the program including an acquisition unit for acquiring the number of ports of each switch, the total number of switches, a terminal utilization rate of terminals connected to the plurality of first stage switches, and an allowable blocking rate that allows communication to be impossible between two terminals connected to the plurality of first stage switches, and a program for calculating a total probability of each blocking occurrence probability when the number of terminals is increased by one from the maximum number of terminals when the network is non-blocking in a network that satisfies the number of ports, the total number, and the terminal utilization rate. The calculation unit functions as a calculation unit that calculates the number of terminals and the number of first-stage switches so that the network capacity is maximized without exceeding the allowable blocking rate, and the calculation unit divides each of the first-stage switches and the second-stage switches into a plurality of IOCs, and one or more third-stage switches are connected to each IOC. The network capacity is calculated by multiplying the number of terminals connected to one first-stage switch, the number of IOCs, and the number of first-stage switches in one IOC, and calculates the combination with the largest network capacity from among a plurality of combinations of the variables of the number of terminals (n) connected to one first-stage switch, the number of IOCs (k), and the number of first-stage switches in one IOC (k'), such that the total probability of each blocking occurrence probability when the number of terminals is increased by one from the maximum number of terminals when the network is non-blocking does not exceed the allowable blocking rate. [Effects of the Invention]
[0008] According to the present disclosure, it is possible to provide a technique that enables a Clos network to have three network layers and obtain a larger network capacity. [Brief explanation of the drawings]
[0009] [Figure 1] FIG. 1 is a diagram illustrating an example of path setting. [Figure 2] FIG. 2 is a diagram showing a path closure situation. [Figure 3] FIG. 3 is a diagram illustrating the blockage. [Figure 4] FIG. 4 is a diagram showing an example of a conventional Clos network. [Figure 5] FIG. 5 is a diagram illustrating an example of a Clos network configuration according to the present disclosure. [Figure 6] FIG. 6 is a diagram illustrating an example of the process of the network design method according to the present disclosure. [Figure 7] FIG. 7 is a flowchart illustrating an example of processing in a network design method according to the present disclosure. [Figure 8] FIG. 8 is a functional block diagram of a network design device according to the present disclosure. [Figure 9] FIG. 9 is a diagram illustrating the network capacity relative to the number of switches as an effect of the network design device according to the present disclosure. [Figure 10] FIG. 10 is a diagram illustrating the network capacity relative to the allowable blocking rate as an effect of the network design device according to the present disclosure. [Figure 11] FIG. 11 is a diagram illustrating the calculation time versus switch size as an effect of the network design device according to the present disclosure. [Figure 12] FIG. 12 is a diagram illustrating the hardware configuration of a computer used in a network design device. DETAILED DESCRIPTION OF THE INVENTION
[0010] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. In the description of the drawings, the same parts are designated by the same reference numerals and the description thereof will be omitted.
[0011] A network design device 1 according to the present disclosure designs an indoor network for a data center, a communication building, or the like.
[0012] Multiple racks installed in a data center are connected to a network, and a communication line (path) is established by connecting designated rack pairs so that they can communicate. The racks house switches, servers, etc.
[0013] As shown in Figure 1, in a telecommunications building, multiple transmission devices are housed in a network and connected to transmission devices in other buildings, other operators, etc. Current transmission devices use packet switching technologies such as IP and Ethernet, but in the future, circuit switching technologies such as optical communications will be used to improve transmission capacity and power efficiency.
[0014] A path closure situation will be explained with reference to Figure 2. In circuit switching, if there is no unused path between the pair for which a connection is requested, the path will not be established. This is called blocking. We want to establish a path between the second port P12 of the first switch S1 and the fourth port P34 of the third switch S3. However, because two previously established paths (dashed lines) mean that no free path exists, the desired path cannot be established.
[0015] Referring to FIG. 3, the occlusion will be further explained.
[0016] First, let us supplement the usage of the network that is the subject of this disclosure. In the subject network, the terminals and the input ports or output ports are connected from the beginning to the end. For example, terminal t1 1 is connected to switch S1, but this wiring does not change from beginning to end. The same is true for the other terminals. All terminals t and r are connected to one of the input / output switches from beginning to end.
[0017] The only things that change within the network are the connection configuration within the switches and the presence or absence of inter-terminal signals. Communication takes place between transmitting terminal t and receiving terminal r. A single terminal cannot communicate with multiple terminals simultaneously, but it can communicate with different terminals at different times. The percentage of time that each terminal is attempting to communicate is called the "terminal utilization rate."
[0018] For example, two terminals t1 1 , r1 kConsider a situation where two terminals t1 and t2 are trying to communicate using only idle links (links with no signal). 1 , r1 k Prepare a path in between.
[0019] In the example of FIG. 3, the input / output switch S1 and the intermediate switch S k+1 There is an empty link between the k+1 and in / out switch S k Therefore, we will introduce an intermediate switch S k+1 Change the internal configuration of the two in / out switches S1 and S k The internal structure of the same is two terminals t1 1 , r1 k As a result, terminal t1 1 From terminal r1 k A path consisting of only free links is created up to terminal t1. 1 The signal sent by k will be delivered to.
[0020] This disclosure describes a network design technique that maximizes network capacity (the number of switches and transmission devices that can be accommodated) while satisfying a given allowable blocking rate. This technology is particularly targeted at Clos networks with a Clos structure, which are often used in data centers. This disclosure also describes a case where switches are divided into three tiers.
[0021] A typical Clos network will be described with reference to Figure 4. Figure 4 shows the Clos network disclosed in Non-Patent Document 2. A Clos network consists of two layers: an input / output layer and an intermediate layer. In Figure 4, the input layer and output layer are combined into one input-output layer, which is folded back at the intermediate layer.
[0022] (Traditional method) An example of the construction of a Clos network according to Non-Patent Document 2 will be described with reference to Fig. 4. Terminals such as racks and transmission devices are connected to the input / output switches in the input / output layer. Terminal t is a transmitting terminal, and terminal r is a receiving terminal.
[0023] The Clos network shown in Figure 4 has a number of switches. Each switch is an NxN switch with N ports on each of the input and output sides. The number of input / output switches in the input / output layer is k. Each input / output switch is connected to n terminals as input and output. The number of intermediate switches is m. Each switch is connected to S1, ..., S2 in order from the input / output switch. k , S k+1 …S k+m Give each number.
[0024] Here, the middle switches are divided into two groups. There are m1 middle switches with v1 links to each input / output switch, and m2 middle switches with v2 links to each input / output switch. m = m1 + m2. Since the number of available switches is a, k + m ≤ a must be satisfied.
[0025] For such a Clos network, we seek a solution to design a network with as large a capacity as possible. The network capacity is defined as the number of terminals nk in the entire network. Here, inputs and outputs can be counted separately. However, the blocking rate must be less than or equal to a specified value ε. The utilization rate p of each terminal is assumed to be known. Note that the utilization rate p of a terminal is estimated from past service performance, etc.
[0026] Non-Patent Document 2 divides the intermediate switches into a first group and a second group. The first group has m1 intermediate switches. In the first group, each intermediate switch has v1 links to each input / output switch. The second group has m2 intermediate switches. In the second group, each intermediate switch has v2 links to each input / output switch.
[0027] Non-Patent Document 2 distinguishes between the two groups as follows:
[0028] The first step is to check whether the number of terminals n connected to the input / output switch on either the input or output side of the connection request is equal to or greater than n1. snb When the number of terminals is n1 or less, the connection is made using one of the m1 intermediate switches included in the first group. snb is obtained as part of the solution. The probability of being able to connect in the first step is less than or equal to the allowable blocking rate ε. The probability of not being able to connect in the first step is less than or equal to a predetermined blocking rate ε1 (ε≦ε1≦1). SNB stands for strictly nonblocking, which means that blocking does not occur.
[0029] The second step is to check whether the number of terminals n connected to the input / output switch on either the input side or output side of the connection request is equal to or greater than the number of terminals n1. snb Beyond the individual, "n1 snb +n2 snb When the number of nodes is less than or equal to m, one of the m2 intermediate switches included in the second group is further used to establish a connection. The probability that a connection cannot be established even in the second step is less than or equal to the allowable blocking rate ε.
[0030] If none of the above applies, the connection request is rejected with a rejection probability equal to or less than the allowable blocking rate ε.
[0031] The ILP (Integer Linear Programming) of Non-Patent Document 2 is expressed by the formulas (1a) to (1m).
[0032]
number
[0033] ε1 is a parameter required only for the solution. Equations (1j)-(1m) are the domain of the variables.
[0034] Equation (1a) is the objective function that maximizes the network capacity. Equation (1b) is a condition regarding the number of ports of the input and output switches. Equations (1c)-(1d) are conditions regarding the number of ports of the middle switches. Equation (1i) is a condition regarding the number of switches.
[0035] Equations (1e)-(1f) are constraints for the first step. From equation (1f), the number of ports used is n1 snb If the number of ports used is n1, then blocking is guaranteed. snb It is guaranteed that the sum of the blocking rates in the more cases does not exceed ε1.
[0036] Equations (1g)-(1h) are constraints on the second step. Equation (1g) ensures that the sum of the blocking rates does not exceed ε. Note that when m2 = 0, the second step is omitted, so ε is also written in equation (1e) as the upper limit of the blocking rate corresponding to m2 = 0.
[0037] We will now outline the algorithm for solving equation (1). The integer decision variables n and n1 snb , k, v1, v2, m1, n2 snb , m2, an exhaustive search is performed. Specifically, all possible combinations of values for these integer decision variables are examined. Since the value of the real variable ε1 is not necessary for operation, only the existence of a feasible ε1 is confirmed.
[0038] (Network design method of the present disclosure) The network design method according to the present disclosure also addresses the same problem as that in Non-Patent Document 1. As shown in Fig. 5, the network according to the present disclosure increases the network hierarchy from two to three, thereby increasing the flexibility of the network configuration and achieving a larger network capacity.
[0039] Each layer of the switch is called a "stage." The first and second stages are collectively called the input-output layer. The third stage is called the intermediate layer.
[0040] The input-output layer consists of k 2-stage Clos networks, each of which is called an input-output-layer switching-network component (IOC).
[0041] The network shown in Figure 5 has three layers. The three layers are called the first stage, second stage, and third stage from the one closest to the terminal. The network includes a plurality of first stage switches, a plurality of second stage switches, and a plurality of third stage switches. The first stage switch belongs to the first stage and is connected to a plurality of terminals. The second stage switch belongs to the second stage and is connected to a plurality of first stage switches. The third stage switch belongs to the third stage and is connected to a plurality of second stage switches.
[0042] In this disclosure, each of the first stage switch and the second stage switch is divided into k IOCs, and one or more third stage switches are connected to each IOC.
[0043] IOC is similar to the two-stage Clos network of Non-Patent Document 2 shown in Figure 4, but the second-stage switch of Non-Patent Document 2 does not have wiring to a higher level (to the right in the figure), so the solution disclosed in Non-Patent Document 2 cannot be applied directly.
[0044] If each IOC is considered as one switch, the overall network shown in FIG. 5 becomes a two-stage Clos network similar to that of Non-Patent Document 2, but there is no technology that can handle the blocking properties of each IOC and the blocking properties of the entire network together.
[0045] Therefore, the network design method according to the present disclosure re-formulates the three-stage Clos network shown in Fig. 5. Table 1 shows a list of variables that appear in Fig. 5.
[0046] [Table 1]
[0047] In this disclosure, the j-th l-th stage switch of the i-th IOC is denoted by S ij l Also, the i-th third stage switch is represented as S i 3 Each first stage switch has n input and n output terminals connected to it.
[0048] In this disclosure, k is the number of IOCs. m is the number of switches in the intermediate layer. The number of links from each third-stage switch to each IOC is v. The intermediate layer as a whole has vm links for each IOC.
[0049] Let k' be the number of first-stage switches in each IOC. Let m' be the number of second-stage switches in each IOC, where the primes represent variables within the IOC. Each pair of first-stage and second-stage switches is connected by v' links.
[0050] Each second-stage switch connects to the intermediate layer with at most z links. Since each IOC has m' second-stage switches, there are at most zm' links from the IOC to the intermediate layer.
[0051] Let us consider the following network design problem. Suppose there are a NxN switches, each with N input and output ports. Using these a NxN switches, we design a network with as large a capacity as possible. The network capacity is defined as the number of terminals nk'k that can be accommodated in the network. The network capacity is nk'k for both input and output. The blocking rate must be less than or equal to a specified value ε. The utilization rate p of each terminal is assumed to be known. The utilization rate is estimated from past service performance, etc.
[0052] Here, the N×N switch is used as a first stage switch, a second stage switch, or a third stage switch.
[0053] This disclosure uses the idea of "considering the blocking process in two steps" in Non-Patent Document 2. Non-Patent Document 2 describes the process when the number of ports used in the first stage switch is n1 snb The non-blocking state is guaranteed when the following condition is met, and the blocking rate is controlled so that it is equal to or less than ε when the condition is exceeded.
[0054] In this disclosure, the number of ports used by the first stage switch is n1 snb Guarantees non-blocking when: However, n1 snb By treating the blocking rate when ε is exceeded in IOC units rather than in switch units, the blocking rate of the 3-stage Clos network is controlled to be less than ε.
[0055] The calculation formulas for the network design method according to the present disclosure are shown in formulas (2a) to (2m).
[0056]
number
[0057] The following is given as a problem. Number of switch ports N Total number of switches a Terminal utilization rate · Allowable blockage rate ε
[0058] Below are the decision variables for the optimization problem. The number of terminals connected to one first stage switch is n. In one IOC in a non-blocking state, the maximum number of ports available on each first stage switch is n1. snb In a non-blocking network, the maximum number of ports available for each IOC is n2. snb ·The number of first stage switches in one IOC, k' Number of IOCs k The number of links v' between one first-stage switch and one second-stage switch Number of links between one IOC and one third-stage switch · Number of second stage switches in one IOC, m' Number of third stage switches m
[0059] Equation (2a) is the objective function for maximizing the network capacity of a three-stage Clos network. Equations (2b)-(2d) represent constraints on the number of available ports on the first, second, and third stage switches, respectively. Equation (2e) indicates that the total number of ports from the third stage switches to the IOC is greater than or equal to vm. Note that equations (2c) and (2e) can be handled by a single constraint: equation (3).
[0060]
number
[0061] Here, the value of the ceiling function in equation (3) is determined by v, m, and m'. The ceiling function refers to the smallest integer that is not less than the value in the function.
[0062] Equation (2f) indicates that the blocking rate in the first stage does not exceed η. η is a parameter required only for the solution, similar to ε1 in Non-Patent Document 2. Equation (2g) corresponds to the SNB condition in the first stage.
[0063] Equation (2h) shows that the blocking rate in the third stage does not exceed ε. The probability that a request will not be blocked in the third stage must be greater than 1-ε, which is expressed as in equation (4). Equation (2h) is obtained from equation (4).
[0064]
number
[0065] Equation (2i) corresponds to the SNB condition in the third stage. Equation (2j) indicates that the total number of switches must not exceed a. Equation (2k) indicates the range of η, which must not exceed ε. Equations (2l) and (2m) indicate the range of each decision variable.
[0066] n1 snb n2 is the maximum number of ports available on each switch in the first stage. snb n2 is the maximum number of ports available to each IOC in a non-blocking network. snb is defined per IOC, not per switch, so the number of available terminals on a switch for a given IOC may vary from switch to switch.
[0067] Since the left side of equation (2h) is a monotonically increasing function, the optimal solution can be obtained by substituting the smallest η that satisfies equation (2f) into equations (2h) and (2k). Therefore, equation (2f) can be regarded as equation (5).
[0068]
number
[0069] (Net design method) The optimization problem shown in equations (2a)-(2m) is solved using the algorithm shown in Figure 6. The algorithm in Figure 6 examines the feasibility of solutions by starting with the decision variables with the largest network capacity. Therefore, the first solution found is the optimal solution, i.e., the solution with the largest network capacity. By appropriately ordering the decision variables to be searched, the search time can be reduced.
[0070] In the third line of Figure 6, the decision variables n, k', and k are searched in descending order of nk'k, and the search ends when a feasible solution is found. The maximum value of nk'k is N(N-1)(N-1) if N≦a, and Na(a-1) otherwise.
[0071] Similarly, n1 snb The maximum values of v, v' and n2 are all N. snb The maximum value of m is N(N-1) if N≦a, otherwise it is N(a-1). The maximum value of m and m' is a-1.
[0072] The algorithm shown in Figure 6 can avoid unnecessary searches by determining whether a condition is satisfied at each point in the for loop. Specifically, line 5 determines whether equation (2d) is satisfied; line 7 determines whether equations (5) and (2k) are satisfied; line 9 determines whether equation (2h) is satisfied; line 11 determines whether equation (2i) is satisfied; line 13 determines whether equation (2j) is satisfied; and line 15 determines whether equations (2b), (2g), and (3) are satisfied.
[0073] The complexity of the algorithm is O(N 8 a 2 ), otherwise O(N 5 a 5 )
[0074] Referring to FIG. 7, the algorithm shown in FIG. 6 will be explained in a flow chart.
[0075] In step S101, the network design device 1 identifies conditions under which the blocking rate is equal to or less than ε in a three-stage network. Specifically, it identifies equations (2a) to (2m).
[0076] In step S102, the network design device 1 identifies each combination of n, k', and k. In step S103, the network design device 1 rearranges the combinations identified in step S102 so that the multiplication of n, k', and k is in descending order.
[0077] In step S104, the network design device 1 sets the combination with the largest multiplication as the target. Here, "target" is the target for judgment in step S105. In step S105, it is judged whether the target is a solution that satisfies the condition specified in step S101. If it does, the processing ends immediately. If it does not, in step S106, the combination with the next largest multiplication after the current target is set as the target, and the processing of step S105 is performed.
[0078] (Net design device) A network design device 1 according to the present disclosure will be described with reference to Figure 8. The network design device 1 designs a network that satisfies a predetermined network capacity, the network including a plurality of first-stage switches that belong to a first stage and are connected to a plurality of terminals, a plurality of second-stage switches that belong to a second stage and are connected to the plurality of first-stage switches, and a plurality of third-stage switches that belong to a third stage and are connected to the plurality of second-stage switches. Here, the predetermined network capacity is maximized under the condition that, when specifications such as the number of switch ports N, the total number of switches a, and the terminal utilization rate p are given, an allowable blocking rate ε that allows communication between two terminals connected to a first-stage switch is satisfied.
[0079] The network design device 1 includes an acquisition unit 11, a calculation unit 12, and a storage unit 13.
[0080] The storage unit 13 stores data to be processed by the network design device 1.
[0081] The acquisition unit 11 acquires the number of ports N of each switch, the total number a of the multiple switches, the terminal utilization rate p of the terminals connected to the multiple first stage switches, and the allowable blocking rate ε that allows communication to be disabled between two terminals connected to the multiple first stage switches.
[0082] The calculation unit 12 calculates the number of terminals n and the number of first stage switches k'k so that in a network that satisfies the number of ports N, the total number a, and the terminal utilization rate p, the total probability of each blocking occurrence probability when the number of terminals is increased by one from the maximum number of terminals when the network is non-blocking does not exceed the allowable blocking rate ε and the network capacity is maximized.
[0083] Here, the first stage switch and the second stage switch are each divided into k IOCs, and one or more third stage switches are connected to one IOC.
[0084] The network capacity is the product of the number of terminals n connected to one first stage switch, the number k of IOCs, and the number k' of first stage switches in one IOC.
[0085] The calculation unit 12 calculates the combination with the largest network capacity from among multiple combinations of the variables of the number of terminals n connected to one first stage switch, the number of IOCs k, and the number of first stage switches k' in one IOC, which satisfies the condition that the total probability of each blocking occurrence probability when the number of terminals is increased by one from the maximum number of terminals when the network is non-blocking does not exceed the allowable blocking rate ε.
[0086] The calculation unit 12 rearranges the multiple combinations so that the product of the number of terminals n connected to one first stage switch, the number of IOCs k, and the number of first stage switches k' in one IOC is in descending order, and calculates whether each combination satisfies the condition that the allowable blocking rate ε is not exceeded. The calculation unit 12 calculates the combination that first satisfies this condition as the combination with the largest network capacity.
[0087] Specifically, the calculation unit 12 calculates combinations that satisfy the formulas (2a) to (2m).
[0088] 9 to 11, the effects of the network design device 1 according to the present disclosure will be described. In Fig. 9 to 11, 3TF relates to the three-stage Clos network according to the present disclosure, and 2TF relates to the two-stage Clos network according to Non-Patent Document 2.
[0089] 9 is a diagram illustrating the network capacity versus the number of switches. In a two-stage Clos network, even if the number of switches is increased, the network capacity does not increase because the switches are not fully utilized. On the other hand, the three-stage Clos network according to the present disclosure increases its network capacity until a = 2400. At a = 2400, the three-stage Clos network can form a network approximately 20 times larger than the two-stage Clos network.
[0090] Figure 10 is a diagram illustrating the network capacity versus the allowable blocking rate. As shown in Figure 10, a three-stage Clos network can be configured as a network approximately 20 times larger than a two-stage Clos network.
[0091] Figure 11 is a diagram explaining the calculation time for switch size. The calculation cost depends heavily on N, so it takes about two weeks when N = 50. In addition, in order to solve a formula containing many decision variables, the network design device 1 can obtain a solution in a realistic time by using the techniques described in Figures 6 to 7.
[0092] The network design device 1 according to the present disclosure can increase the flexibility of the network configuration and obtain a larger network capacity by increasing the network layer from 2 to 3. To solve equation (2) which includes many decision variables, a solution can be obtained within a realistic time by searching from decision variables with large network capacity, simplifying equation (3), and substituting equation (5).
[0093] The network design device 1 according to the present disclosure can set three network layers in a Clos network, thereby achieving a larger network capacity.
[0094] The network design device 1 according to the present disclosure described above uses, for example, a general-purpose computer system including a CPU (Central Processing Unit, processor) 901, a memory 902, a storage 903 (HDD: Hard Disk Drive, SSD: Solid State Drive), a communication device 904, an input device 905, and an output device 906. In this computer system, the CPU 901 executes a program loaded on the memory 902, thereby realizing each function of the network design device 1.
[0095] The network design device 1 may be implemented in one computer or in multiple computers, or may be a virtual machine implemented in a computer.
[0096] The program of the network design device 1 can be stored in a computer-readable recording medium such as a HDD, SSD, USB (Universal Serial Bus) memory, CD (Compact Disc), DVD (Digital Versatile Disc), or can be distributed via a network. The computer-readable recording medium is, for example, a non-transitory recording medium.
[0097] The present disclosure is not limited to the above-described embodiments, and various modifications are possible within the scope of the present disclosure. [Explanation of symbols]
[0098] 1 Network design device 11 Acquisition Department 12 Calculation section 13 Storage section 901 CPU 902 memory 903 Storage 904 Communication equipment 905 Input Device 906 Output Device
Claims
1. a plurality of first stage switches belonging to the first stage and connected to a plurality of terminals; a plurality of second stage switches belonging to a second stage and connected to the plurality of first stage switches; A network design device for designing a network including a plurality of third stage switches that belong to a third stage and are connected to the plurality of second stage switches, an acquisition unit that acquires the number of ports of each switch, the total number of the plurality of switches, a terminal utilization rate of terminals connected to the plurality of first stage switches, and an allowable blocking rate that allows communication between two terminals connected to the plurality of first stage switches to be disabled; a calculation unit that calculates the number of terminals and the number of first stage switches so that in a network that satisfies the number of ports, the total number, and the terminal utilization rate, the total probability of each blocking occurrence probability when the number of terminals is increased by one from the maximum number of terminals when the network is non-blocking does not exceed the allowable blocking rate and the network capacity is maximized; The calculation unit The first stage switch and the second stage switch are each divided into a plurality of IOCs, and one or more third stage switches are connected to each IOC; The network capacity is a product of the number of terminals connected to one first stage switch, the number of IOCs, and the number of first stage switches in one IOC; Among multiple combinations of the variables of the number of terminals (n) connected to one first stage switch, the number of IOCs (k), and the number of first stage switches in one IOC (k'), a combination that satisfies the condition that the total probability of each blocking occurrence probability when the number of terminals is increased by one from the maximum number of terminals when the network is non-blocking does not exceed the allowable blocking rate and has the largest network capacity is calculated. Net design device.
2. The calculation unit The plurality of combinations are rearranged in a descending order of the multiplications, Calculating whether each combination satisfies the condition; The combination that first satisfies the above condition is calculated as the combination with the largest network capacity.
2. The network design device according to claim 1.
3. The calculation unit calculates a combination that satisfies formulas (2a) to (2m).
2. The network design device according to claim 1. [Equation 1]
4. a plurality of first stage switches belonging to the first stage and connected to a plurality of terminals; a plurality of second stage switches belonging to a second stage and connected to the plurality of first stage switches; a computer for designing a network including a plurality of third stage switches that belong to a third stage and are connected to the plurality of second stage switches; an acquisition unit that acquires the number of ports of each switch, the total number of the plurality of switches, a terminal utilization rate of terminals connected to the plurality of first stage switches, and an allowable blocking rate that allows communication between two terminals connected to the plurality of first stage switches to be disabled; a calculation unit that calculates the number of terminals and the number of first stage switches so that, in a network that satisfies the number of ports, the total number, and the terminal utilization rate, the total probability of each blocking occurrence probability when the number of terminals is increased by one from the maximum number of terminals when the network is non-blocking does not exceed the allowable blocking rate and the network capacity is maximized; The calculation unit The first stage switch and the second stage switch are each divided into a plurality of IOCs, and one or more third stage switches are connected to each IOC; The network capacity is a product of the number of terminals connected to one first stage switch, the number of IOCs, and the number of first stage switches in one IOC; Among multiple combinations of the variables of the number of terminals (n) connected to one first stage switch, the number of IOCs (k), and the number of first stage switches in one IOC (k'), a combination that satisfies the condition that the total probability of each blocking occurrence probability when the number of terminals is increased by one from the maximum number of terminals when the network is non-blocking does not exceed the allowable blocking rate and has the largest network capacity is calculated. program.