Processing apparatus, processing method, and processing program

Abstract

This technology contributes to calculating the optimal network capacity for communication networks. [Solution] The processing device 1 for calculating the network capacity of a communication network includes an acquisition unit 11 that acquires available network information for each of a plurality of communication networks having the same configuration, and a selection unit 13 that uses the available network information for each of the plurality of communication networks to select a communication network with available capacity from among the plurality of communication networks.

Description

Technical Field

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

Background Art

[0002] In a data center, communication devices (for example, switches, routers, line devices, transmission devices) within the data center are housed in a network, and setting processing of communication lines for connecting a plurality of designated communication devices to each other is performed. Similar setting processing is also performed between a plurality of data centers.

[0003] In circuit switching, when there is no unused path between two communication devices for which a connection request has been made, a line cannot be set between those two communication devices. This state is called a blocking state. Therefore, when a certain allowable blocking rate is given, a network design method has been proposed that maximizes the network capacity (the number of communication devices that can be accommodated) of a Clos network while satisfying the allowable blocking rate (see Non-Patent Documents 1-4).

Prior Art Documents

Non-Patent Documents

[0004]

Non-Patent Document 1

Non-Patent Document 2

Non-Patent Document 3

[0005] However, in the network design methods of Non-Patent Documents 1 - 4, there was a possibility of missing the optimal solution of the network capacity of the Clos network.

[0006] The present disclosure has been made in view of the above circumstances, and an object thereof is to provide a technique that contributes to the calculation of the optimal solution of the network capacity of a communication network. [Means for Solving the Problems]

[0007] A processing device according to an aspect of the present disclosure is a processing device related to the calculation of the network capacity of a communication network, and includes an acquisition unit that acquires the free information of each of a plurality of communication networks having the same configuration, and a selection unit that selects a communication network with free space from among the plurality of communication networks using the free information of each of the plurality of communication networks.

[0008] One aspect of the present disclosure is a processing method performed by a processing device for calculating the network capacity of a communication network, wherein the processing method involves acquiring availability information for each of a plurality of communication networks having the same configuration, and using the availability information for each of the plurality of communication networks to select a communication network with availability from among the plurality of communication networks.

[0009] A processing program in one aspect of the present disclosure causes a computer to perform the following processes: acquiring availability information for each of a plurality of communication networks having the same configuration; and using the availability information for each of the plurality of communication networks, to select an available communication network from among the plurality of communication networks. [Effects of the Invention]

[0010] This disclosure provides technology that contributes to calculating the optimal network capacity for a communication network. [Brief explanation of the drawing]

[0011] [Figure 1] Figure 1 shows an example of a Clos network. [Figure 2] Figure 2 shows a conventional connection request processing flow. [Figure 3] Figure 3 shows a conventional algorithm. [Figure 4] Figure 4 shows the configuration of the system according to this embodiment. [Figure 5] Figure 5 shows the configuration of the processing apparatus according to this embodiment. [Figure 6] Figure 6 shows the connection request processing flow of this embodiment. [Figure 7] Figure 7 shows the usage of each surface at each step. [Figure 8] Figure 8 shows the network capacity by number of faces relative to the number of switches. [Figure 9] Figure 9 shows the network capacity by terminal utilization rate relative to the number of terminals. [Figure 10] Figure 10 shows an example of the hardware configuration of the processing unit. [Modes for carrying out the invention]

[0012] Embodiments of this disclosure will be described below with reference to the drawings.

[0013] In this embodiment, the network design targets a Clos network, which is commonly used in data centers. A Clos network is an example of a circuit-switched network. This embodiment is applicable to any communication network, including circuit-switched networks and optical transmission networks.

[0014] [Conventional method] As a comparative example of the method according to this embodiment, a conventional method will be described first.

[0015] Non-patent document 3 describes a network design method that theoretically guarantees a certain allowable occlusion rate for the Clos network illustrated in Figure 1.

[0016] A Clos network consists of multiple input / output switches located in the input-output layer and multiple intermediate switches located in the intermediate layer. Figure 1(a) illustrates a link connected to the first input / output switch. Figure 1(b) illustrates a link connected to the last input / output switch.

[0017] Terminals (transmitter, receiver) are connected to the input / output switches. There are k input / output switches, and each switch has n terminals connected to its input and n terminals to its output. There are m intermediate switches, and the input / output switches are numbered S1, ..., S1 in order from the first switch. k+m Assign numbers like this.

[0018] The intermediate switch is divided into S groups, and for each group s ∈ {1..S}, the link with the input / output switch is v S Each intermediate switch is m s Let's assume there are m s For m, m = Σ i∈{1..S} m iIt is related to this. When the number of available switches (the total number of input / output switches and intermediate switches) is a, it must satisfy k + m ≤ a.

[0019] Here, consider the following network design problem.

[0020] Suppose there are a N×N switches, each with N ports on both the input and output sides. Design a network with the largest possible capacity using a N×N switches. Let the number of terminals nk of the entire network be the network capacity, and the inputs and outputs are counted separately. However, the allowable blocking rate must be no more than the specified value ε. The utilization rate p of each terminal is known, for example, an estimated value from past service performance values.

[0021] Using FIG. 2, the conventional connection request processing flow will be described.

[0022] This flow is executed by the input / output switch or the management device of the Clos network.

[0023] If the connection request is a connection between ports under the same input / output switch (that is, a local connection that does not need to pass through an intermediate switch), it can always be connected, so accept the connection request (steps S101, S102).

[0024] If the connection request is not a local connection, set it as step s = 1, and perform the following processing for the intermediate switches in group s (step S103).

[0025] In the input / output switch where the connection request is made, let X s be the number of existing connections passing through the intermediate switches belonging to m s in group s. Also, assume that the allowable connection number n snb s for each group s is determined by the following set of equations.

[0026] At this time, X s ≤ n snb sIf -1, m s Connect using one of the intermediate switches (steps S104-S106). If there are multiple intermediate switches that can be configured for connection, any of them can be selected.

[0027] X s ≤n snb s If it is not -1, take s=s+1 steps and return to step S104 until s≦S is satisfied (steps S107, S108). If a connection cannot be made even if s>S is finally found, the connection request is rejected (step S109).

[0028] The conventional set of formulas is shown below.

[0029]

number

[0030] N, a, p, ε are given as part of the problem. S is determined by the person attempting to solve the optimization problem. Simply put, S should be increased as much as time allows, starting from 1, 2, 3, ... The larger S is, the longer the computation time. N, k, m1, ..., m S ,v1, ...,v S , n snb 1, ..., n snb S These are the decision variables of the optimization problem: ε1, ..., ε S-1 These are parameters necessary only for solving optimization problems. Equations (1h) to (1m) are the domains of each variable.

[0031] Equation (1a) is the objective function for maximizing network capacity. Equation (1b) is a condition regarding the number of ports on the input / output switches. The number of links to intermediate switch i is totaled. Equation (1c) is a condition regarding the number of ports on the intermediate switches. It is set to ∀i∈[1,S] to define it for each group i. Equation (1e) is a condition regarding the number of switches. The number of intermediate switches i is totaled.

[0032] Equations (1d) to (1g) are constraints for controlling blockage. Equation (1d) ensures that blockage does not occur if the number of ports in use is n snb i or less. For the s-th <S step, Equation (1f) ensures that the sum of the blockage rates when the number of ports in use is more than n snb s does not exceed ε s . For the S-th step, Equation (1g) ensures that the blockage rate does not exceed ε.

[0033] Using Figure 3, a conventional algorithm will be described.

[0034] Since there are many decision variables, the following two techniques are used to suppress the computational complexity.

[0035] (1) The number k of input / output switches and the number n of terminals are efficiently determined by binary search. For the other integer variables (n snb i , m i , v i ), exhaustive search is performed.

[0036] (2) For the real variable ε i , assuming the monotonicity of the constraint conditions, the feasibility (whether the constraints are satisfied) is checked only at the endpoints.

[0037] In the third line of Figure 3, an unblocked network (a network with a blockage rate of zero) is obtained using the existing method of Non-Patent Document 4. Since blockage below ε is tolerated, the network capacity will not be smaller than that of the unblocked network. Therefore, the unblocked network is used as the starting point of the search, and the capacity and structure of the unblocked network are substituted in the fourth and fifth lines.

[0038] Lines 6 to 14 loop through all combinations of m i , v i , n snb i .

[0039] Lines 15 to 17 are the range of k (lower limit K L and upper limit K UWe have determined the range, and then perform a binary search by repeatedly halving the range. The lower bound of k is derived from equation (1b), and the upper bound of k is derived from equations (1c) and (1e).

[0040] Next, a binary search is performed on n (lines 19-38). The range of n in the binary search is defined by its lower bound (left), which is the value obtained by dividing the network capacity SW at that point by k (line 19), and its upper bound (right), which is the maximum value of n, i.e., N (line 20). If the product of the maximum value of n and k is smaller than the network capacity calculated earlier, the capacity cannot be improved, so the binary search is terminated (lines 22-23).

[0041] In line 25, the constraints are checked using the median value of n. If all constraints are met, the network capacity and network structure are updated if the network capacity has increased (lines 25-28). Furthermore, the lower bound of n is updated to the median value of n, and the binary search continues (line 29).

[0042] If the constraints other than those in equation (1k) are satisfied at the midpoint of n, the lower bound of n is updated to the midpoint of n (making the range of n the upper half), and the binary search continues without updating the maximum network capacity (lines 30-31). The reason for re-searching the upper half of n is that there may be feasible solutions for n greater than the midpoint of n.

[0043] If the intermediate value of n does not satisfy any constraints other than those in equation (1k), the upper limit of n is updated to a value 1 less than the intermediate value of n (lines 32-33).

[0044] Once the binary search for n is complete, the value of k for the next search is determined (lines 34-38). If the maximum network capacity is updated during the search for n and the lower bound of k becomes greater than the value of k during the search, the lower bound of k is updated and the next loop starts with the new value of k (lines 34-36). Otherwise, the value of k is incremented by 1 (lines 37-38).

[0045] Finally, the algorithm terminates by returning the optimal network capacity and network structure (line 39).

[0046] In the algorithm shown in Figure 3, when checking the constraints, the following checks are performed: (1f) to (1g) (ε that satisfy them) i Determining whether or not it exists is not straightforward. Therefore, based on the following recurrence relation, ε i The minimum value of ε min i We will update ε min i We will use this to examine equations (1f) to (1g).

[0047]

number

[0048] [Method of this embodiment] ε min i By using this, if equations (1f) to (1g) are satisfied, then the constraints are met for ε. i Such a structure exists. However, just because a condition is not met does not mean that it does not exist. Therefore, conventional methods using this determination method may miss the optimal solution. In other words, there remains a possibility that a network structure with a larger network capacity exists.

[0049] In this regard, it is thought that increasing the number of network layers will increase the degree of freedom in network design, and that an even better solution may be found. Therefore, in this embodiment, multiple Clos networks are prepared as shown in Figure 4.

[0050] Each Clos network is the same as a conventional Clos network, and each is called a "plane." Let M be the number of planes, and all planes have exactly the same network configuration. That is, the hardware configuration, such as the number of switches and the number of links between switches, is exactly the same.

[0051] Each terminal is provided with either a 1xM selector or an Mx1 selector; in other words, a selector is provided for each terminal for multiple faces, and one face to be used for connection is selected from among the multiple faces. The aforementioned "increasing the network hierarchy..." means adding a hierarchy of selectors.

[0052] Figure 5 shows the configuration of the processing unit 1 that operates with a selector. The processing unit 1 comprises an acquisition unit 11, a storage unit 12, and a selection unit 13.

[0053] The acquisition unit 11 has the function of acquiring availability information for each of the multiple surfaces from the aforementioned management device.

[0054] The memory unit 12 has a function to store the available space information for each of the acquired surfaces in a readable format.

[0055] The selection unit 13 reads the available space information for each of the multiple surfaces from the storage unit 12, and has the function of using the read available space information for each of the multiple surfaces to select an available surface from among the multiple surfaces for the terminal.

[0056] The processing unit 1 may be located within the aforementioned management device. In this case, the selection unit 13 selects an available surface for each selector and notifies each selector of the selected available surface. As a result, when each selector receives a connection request, it selects the notified surface.

[0057] The connection request processing flow of this embodiment will be explained using Figure 6.

[0058] It is largely the same as before, but the inequality at the beginning of the loop is different. Specifically, X s ≤ Mn snb s -1, where n is on the right side snb s It is multiplied by M. In other words, in this embodiment, n snb s Since this value is defined for each face, it is multiplied by M to obtain the total value for all faces.

[0059] When accepting a connection, the selector (processing device 1) selects an available face from the M faces based on the availability information of each face (steps S205-S206).

[0060] This point will be explained using Figure 7.

[0061] Since each face is the same as the existing Clos network, in the sth step, n faces snb s It can accommodate connections up to the book. Therefore, the whole thing is X s ≤ Mn snb s Even if the condition is met, the number of connections may already be n on some surfaces. snb s It can reach that point.

[0062] Figure 7 illustrates this situation using s=2 steps as an example. The number of connections between face 2 and face 4 is n. snb s The target has been reached, but faces 1 and 3 have not yet been reached. In this case, faces 1 and 3 are considered available, and a connection is established to either face.

[0063] The set of mathematical formulas according to this embodiment is shown below.

[0064]

number

[0065] Equations (3c) to (3e) are replacements for the conventional equations (1e) to (1g). Equations (3b) and (3f) are the same as before. Equation (3c) shows that the number of switches on each face is less than or equal to a / M. Equations (3d) and (3e) take into account that there are M faces and n snb s This will be determined. Specifically, MΣ s-1 j=1 n snb j This term subtracts the number of connections up to step s-1 on all sides.

[0066] Equations (3a) to (3f) can be solved using the same algorithm as before, so the explanation of the algorithm is omitted. The only difference is the addition of M; otherwise, it is essentially the same as before.

[0067] Figure 8 shows the switching capacity per number of faces (M) relative to the number of switches a, with a switch size N=50, terminal utilization rate p=0.6, and allowable occlusion rate ε=0.01. SNB in ​​the figure represents reference results for zero occlusion rate. M=1 corresponds to the conventional method, and it can be seen that the network capacity increases as the number of faces (M) increases.

[0068] Figure 9 shows the change in network capacity when the horizontal axis is the number of faces (M). The network capacity peaks around M=5, suggesting that a value of 4 to 8 for the number of faces (M) is appropriate. Since 1×M selectors are created by combining 1×2 selectors, it is thought that powers of 2 are often chosen for M.

[0069] Based on the above, according to this embodiment, the selector (processing device 1) acquires availability information for each of multiple faces with the same configuration, and uses the acquired availability information for each of the multiple faces to select a face with availability from among the multiple faces. Therefore, this provides a technology that contributes to calculating the optimal network capacity of the Clos network.

[0070] [others] This disclosure is not limited to the embodiments described above. Numerous modifications are possible within the scope of the gist of this disclosure. The processing device 1 may be located within a control device.

[0071] In this case, the processing unit 1 can be implemented using a general-purpose computer system that includes, for example, a CPU 901, memory 902, storage 903, communication device 904, input device 905, and output device 906, as shown in Figure 10.

[0072] Memory 902 and storage 903 are storage devices. In this computer system, each function of the processing unit 1 is realized when the CPU 901 executes a predetermined program loaded onto memory 902.

[0073] Processing unit 1 may be implemented on a single computer. Processing unit 1 may be implemented on multiple computers. Processing unit 1 may be a virtual machine implemented on a computer.

[0074] The program for the processing unit 1 can be stored on a computer-readable recording medium such as an HDD, SSD, USB memory, CD, or DVD. A computer-readable recording medium is, for example, a non-transitory recording medium. The program for the processing unit 1 can also be distributed via a communication network. [Explanation of Symbols]

[0075] 1 Processing Unit 11 Acquisition Department 12 Storage section 13 Selection Section 901 CPU 902 memory 903 Storage 904 Communication equipment 905 Input device 906 Output device

Claims

1. In a processing device for calculating the network capacity of a communication network, An acquisition unit that acquires availability information for each of multiple communication networks with the same configuration, A selection unit that uses the availability information of each of the aforementioned multiple communication networks to select an available communication network from among the aforementioned multiple communication networks, A processing device equipped with the following features.

2. The processing device operates with a selector connected between the multiple communication networks and the terminal, The aforementioned selection unit is The apparatus according to claim 1, which selects the available communication network for the selector.

3. The aforementioned communication network is The processing apparatus according to claim 1, which is a circuit-switched network.

4. In a processing method performed by a processing device for calculating the network capacity of a communication network, Obtain availability information for each of multiple communication networks with the same configuration. Using the availability information of each of the aforementioned multiple communication networks, select an available communication network from among the aforementioned multiple communication networks. Processing method.

5. On the computer, The process of obtaining availability information for each of multiple communication networks with the same configuration, A process of selecting an available communication network from among the multiple communication networks using the availability information of each of the multiple communication networks, A processing program that executes [something].

Images