Communication Network Modeling Support Device, Communication Network Modeling Support Method, and Program

Abstract

An NW model unit (21) of a communication network modeling support apparatus (10) constructs, on an NW simulator, NW models (1P1 through 1Pn) corresponding to a real NW (31) in which real nodes (32a through 32d) are connected to each other. A real node information collection unit (11) collects routing configuration information and node internal information from the real nodes (32a through 32d). A generation unit (12) generates, on the basis of the routing configuration information and the node internal information, multiple NW model patterns for configuring node models (21a through 21d) corresponding to the real nodes (32a through 32d). The NW model unit 21 constructs, on the NW simulator, the NW models 1P1 through 1Pn that correspond to the multiple NW model patterns and in each of which the node models 21a through 21d are connected to each other.

Description

Technical Field

[0001] The present invention relates to a communication network modeling support device, a communication network modeling support method, and a program for assisting in constructing a communication network model in which a plurality of node models are connected, corresponding to a communication network in which a plurality of actual nodes are connected.

Background Art

[0002] An actual communication network is constructed by connecting nodes (also referred to as actual nodes) such as servers, routers, and switches through a wired or wireless network {also referred to as NW (Network)}. The actual communication network is also referred to as an actual NW.

[0003] In the prior art, there are techniques (Non-Patent Documents 1 and 2) for generating an NW model regarding the connection between actual nodes constituting an actual NW. Further, there is a technique (Non-Patent Document 3) for estimating NW performance by deep learning (graph neural NW) that inputs the NW topology obtained by the techniques of Non-Patent Documents 1 and 2 as graph data. By these techniques, an NW model corresponding to the actual NW is constructed.

Prior Art Documents

Non-Patent Documents

[0004]

Non-Patent Document 1

Non-Patent Document 2

Non-Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0005] By the way, in order to configure a better actual NW, a NW model that accurately reflects the performance characteristics of the actual NW is required. However, conventional NW models could not accurately reflect the performance characteristics of the actual NW. For example, performance characteristics such as the number of queues and queue lengths as the internal configuration of nodes in the actual NW, and the processing speed and number of parallel executions of the processor could not be accurately reflected in the node model to construct a NW model. That is, there was a problem that a NW model that accurately reflects the performance characteristics of the actual NW could not be constructed.

[0006] The present invention has been made in view of such circumstances, and an object thereof is to construct a NW model that accurately reflects the performance characteristics of an actual NW.

Means for Solving the Problems

[0007] To solve the above problems, a communication network modeling support apparatus according to the present invention Through corresponds to an actual NW in which a plurality of actual nodes as communication devices are NW (Network) - connected and deployed, and constructs a NW model in which a plurality of node models are connected on a simulator, Before an actual node information collection unit that collects routing setting information and node internal information from each actual node of the actual NW, Before estimates connections between each actual node based on the routing setting information, and generates a plurality of NW model patterns that configure node models corresponding to the actual nodes based on node internal information in which the node internal information is changed into a plurality of types, Before a first DB (Data Base) that stores a plurality of generated NW model patterns And A second database that stores multiple types of traffic patterns that change the load of nodes A model load test unit that inputs a plurality of traffic patterns acquired from the second database into each NW model of the NW model unit A node model information collection unit that collects statistical information of each node model for each NW model that operates with the input of the traffic pattern A model performance measurement unit that measures the performance of each NW model that operates with the input of the traffic pattern to obtain performance measurement information An inference unit that generates an inference device that estimates, from among a plurality of NW models constructed in the NW model unit, an NW model having characteristics approximated to the traffic pattern, statistical information, and performance measurement information by performing learning using learning data combining the traffic pattern, the statistical information, and the performance measurement information And and includes Before the NW model unit constructs a plurality of NW models in which each node model corresponding to a plurality of NW model patterns acquired from the first DB is connected on a simulator This characterized by the above.

Effects of the Invention

[0008] According to the present invention, it is possible to construct a NW model that accurately reflects the performance characteristics of an actual NW.

Brief Description of the Drawings

[0009]

Figure 1

Figure 2

Figure 3

Figure 4

Mode for Carrying Out the Invention

[0010] Hereinafter, embodiments of the present invention will be described with reference to the drawings. <Configuration of Embodiment> FIG. 1 is a block diagram showing the configuration of a communication network modeling support apparatus according to an embodiment of the present invention.

[0011] A communication network modeling support apparatus (also referred to as a support apparatus) 10 shown in FIG. 1 supports selecting an NW model corresponding to an actual NW (Network) 31 in which a plurality of nodes (actual nodes) 32a to 32d as communication devices are NW-connected from a plurality of NW models 1P1 to 1Pn. The actual NW is an actual communication network.

[0012] However, each of the NW models 1P1 to 1Pn is configured by virtually NW-connecting a plurality of node models 21a to 21d. The actual NW 31 is constructed by NW-connecting a plurality of actual nodes 32a to 32d. The actual nodes 32a to 32d are communication devices such as servers, routers, and switches.

[0013] This support device 10 is configured to include a physical node information collection unit 11, a NW model pattern generation unit 12, a NW model pattern DB (Data Base) unit 13, a model load test unit 14, a traffic pattern DB unit 15, a learning data input processing unit 16, a node model information collection unit 17, a model performance measurement unit 18, a learning data DB unit 19, a NW model inference unit 20, a NW model unit 21, a real environment load test unit 22, an inference unit input processing unit 23, and a real environment performance measurement unit 24. The NW model pattern DB unit 13 is also referred to as the DB unit 13, the traffic pattern DB unit 15 is also referred to as the DB unit 15, and the learning data DB unit 19 is also referred to as the DB unit 19.

[0014] Note that the NW model pattern DB unit 13 constitutes the first DB described in the claims. The traffic pattern DB 15 constitutes the second DB described in the claims. The NW model pattern generation unit 12 constitutes the generation unit described in the claims. The model performance measurement unit constitutes the measurement unit described in the claims.

[0015] The NW model unit 21 constructs a plurality of NW models 1P1 to 1Pn on the NW simulator, in which a plurality of node models 21a, 21b, 21c, 21d are virtually NW-connected. The NW simulator constitutes the simulator described in the claims.

[0016] The support device 10 has a first communication NW modeling support function (also referred to as the first support function), a second communication NW modeling support function (also referred to as the second support function), and a third communication NW modeling support function (also referred to as the third support function), which will be described later.

[0017] <First communication NW modeling support function> The first support function generates a plurality of NW models 1P1 to 1Pn corresponding to the real NW 31 based on the information that can be obtained from each physical node 32a to 32d of the real NW 31, and registers them in the NW model unit 21. This first support function is realized by using the physical node information collection unit 11, the NW model pattern generation unit 12, the NW model pattern DB unit 13, and the NW model unit 21.

[0018] The actual node information collection unit 11 (also referred to as the collection unit 11) acquires information such as routing setting information indicating the connections among the actual nodes 32a to 32d of the actual NW 31, the operating status of the ports of the actual nodes 32a to 32d, and node internal information related to the processor, memory, queue, etc.

[0019] The NW model pattern generation unit 12 (also referred to as the generation unit 12) estimates (or identifies) the connections among the actual nodes 32a to 32d based on the routing setting information acquired by the collection unit 11. Further, the generation unit 12 can estimate (or identify) all the paths through which traffic can pass within the actual NW 31 by repeating the estimation over all the actual nodes 32a to 32d based on the routing setting information.

[0020] Furthermore, the generation unit 12 generates, for example, N NW model patterns including node models 21a to 21d corresponding to information such as the number of ports of the node, the number of connected ports, the name of the processor, the number of queues, queue length, processing speed of the processor, and the number of parallel executions related to the internal configuration of the node based on the node internal information of the actual nodes 32a to 32d.

[0021] More specifically, the generation unit 12 changes the node internal information of the group of actual nodes 32a to 32d constituting the actual NW 31 into N types, and generates N NW model patterns for constituting the group of node models 21a to 21d corresponding to each of these N types of node internal information.

[0022] Furthermore, when generating N NW model patterns, the generation unit 12 can, for example, prepare as many queues as the number of ports associated with the ports of the actual nodes 32a to 32d, and the length of the queues can correspond to the minimum required number of processor cores. Also, for example, when the node internal information acquired by the collection unit 11 does not clarify the processing speed and the number of parallel executions of the processor, NW model patterns having the operating frequency and the number of cores of the processor may be generated with reference to specifications (not shown) of commercially available models of the same purpose or similar performance. Furthermore, starting from the NW model patterns generated in this way, NW models 1P1 to 1Pn with changed queue numbers, queue lengths, processor operating frequencies, and numbers of cores may be prepared.

[0023] For each of the node models 21a to 21d corresponding to the N NW model patterns generated by the generation unit 12 in this way, the NW models 1P1 to 1Pn are virtually connected so as to correspond to the connection information between the actual nodes 32a to 32d estimated by the generation unit 12 as described above.

[0024] The NW model pattern DB unit 13 stores the N NW model patterns generated and registered by the generation unit 12.

[0025] The NW model unit 21 acquires the N NW model patterns registered in the DB unit 13, and constructs the NW models 1P1 to 1Pn corresponding to the respective NW model patterns on the NW simulator by storing them in a storage unit (not shown) of the NW simulator.

[0026] <Operation of the First Communication NW Modeling Support Function> Next, the communication NW modeling support process by the first support function of the support device 10 will be described with reference to the flowchart shown in FIG. 2.

[0027] In step S1 shown in FIG. 2, the collection unit 11 collects information such as routing setting information and node internal information from the actual nodes 32a to 32d of the actual NW 31, and outputs it to the generation unit 12.

[0028] In step S2, the generation unit 12 estimates the connections between the actual nodes 32a to 32d based on the routing setting information.

[0029] Furthermore, in step S3, the generation unit 12 generates N NW model patterns for constructing the node models 21a to 21d corresponding to the actual nodes 32a to 32d based on the node internal information obtained by changing the node internal information of the group of actual nodes 32a to 32d into N types. Each of the node models 21a to 21d for each of these NW model patterns is virtually connected corresponding to the connections between the actual nodes 32a to 32d estimated in step S2 above.

[0030] In step S4, the generation unit 12 registers the N generated NW model patterns in the DB unit 13.

[0031] In step S5, the NW model unit 21 acquires the N NW model patterns registered in the DB unit 13, and constructs them on the NW simulator by storing the NW models 1P1 to 1Pn corresponding to the NW model patterns in the storage unit (not shown) of the NW simulator.

[0032] <Second communication NW modeling support function> The second support function sequentially inputs the M traffic pattern information (also referred to as traffic patterns) stored in the traffic pattern DB unit 15 into the NW models 1P1 to 1Pn of the NW model unit 21. It measures the performance of each of the NW models 1P1 to 1Pn that operate in response to this input, and uses machine learning or the like with the input traffic pattern information, the performance measurement results for each of these NW models 1P1 to 1Pn, the statistical information (described later) of the node models 21a to 21d for each of the NW models 1P1 to 1Pn, and the above-mentioned N NW model patterns as learning data to generate an inference device 20a. The inference device 20a estimates the NW model that most closely approximates the actual NW 31 among the N NW models 1P1 to 1Pn. However, the statistical information is information on the setting information, log information, statistical results such as the arrival number of packets for each node model 21a to 21d or for each queue within the node model, and the usage rate of the processor.

[0033] This second support function is realized by using, in addition to the NW model unit 21 described above, a load test unit 14 for models, a traffic pattern DB unit 15, a learning data input processing unit 16, a node model information collection unit 17, a performance measurement unit 18 for models, a learning data DB unit 19, and an NW model inference unit 20. Note that the learning data input processing unit 16 and the NW model inference unit 20, and an inference unit input processing unit 23 described later constitute the inference unit described in the claims. The traffic pattern DB unit 15 stores a plurality of types (assumed to be M types) of traffic patterns. It is assumed that the traffic patterns are numbered from the first to the Mth. The traffic pattern is a pattern such as an input packet in which the load on the node changes according to the input of this traffic pattern.

[0034] For example, even if the total traffic volume is the same, the load on the router is different when a large number of short packets are input during a certain time interval and when a small number of long packets are input. In this way, patterns in which the load on the node differs according to differences such as the amount and size of the input packets, the number of input packets per unit time, the protocol followed by the packets, and flow information such as the destination address included in the input packets are defined as M types of traffic patterns.

[0035] The load test unit 14 for models (also referred to as the test unit 14) reads the M traffic patterns from the DB unit 15 one by one and inputs them to each NW model 1P1 to 1Pn for a predetermined time. Further, the test unit 14 inputs the traffic patterns input to the NW models 1P1 to 1Pn to the learning data input processing unit 16 (also referred to as the input processing unit 16).

[0036] The node model information collection unit 17 collects statistical information on each of the node models 21a to 21d that operate with the input of the M types of traffic patterns.

[0037] The performance measurement unit 18 for models (also referred to as the measurement unit 18) measures the performance during the operation of each of the NW models 1P1 to 1Pn when M types of traffic patterns are sequentially input, and obtains information on the performance measurement results of the measurement results (performance measurement information). Specifically, the measurement unit 18 measures performance such as what the throughput is, what the delay is, and what percentage the discard rate is in the NW models 1P1 to 1Pn.

[0038] The traffic patterns input to the NW models 1P1 to 1Pn from the above-described test unit 14, the statistical information of each node model 21a to 21d aggregated by the collection unit 17, and the performance measurement information that is the performance measurement result of the measurement unit 18 are grouped together by the learning data input processing unit 16 (also referred to as the input processing unit 16) to form learning data. This learning data is stored in the learning data DB unit 19 (also referred to as the DB unit 19).

[0039] That is, the input processing unit 16 associates the input traffic pattern information when a certain traffic pattern is input, the performance measurement information for each of the NW models 1P1 to 1Pn that operate in response to this input, the statistical information of the node models 21a to 21d for each of the NW models 1P1 to 1Pn, and the above-described N NW model patterns, and records them in the learning data DB unit 19 as N pieces of learning data. By repeating the above for a total of M traffic pattern inputs, M×N pieces of learning data are generated, and these are recorded in the learning data DB unit 19.

[0040] The NW model inference unit 20 (inference unit 20) generates an inference device 20a by machine learning or the like based on the learning data stored in the DB unit 19. This inference device 20a, based on the information of the input traffic patterns when several traffic patterns are input to a certain NW, the performance measurement information such as throughput, delay, and discard rate of this NW for this input, and the statistical information obtained from each node constituting this NW, estimates and selects the NW model (for example, NW model 1P1) that is considered to be the closest (approximate) to this NW among the N NW models 1P1 to 1Pn.

[0041] <Operation of the Second Communication NW Modeling Support Function> Next, the communication NW modeling support process by the second support function of the support device 10 will be described with reference to the flowchart shown in FIG. 3. However, it is assumed that the traffic pattern DB unit 15 stores M types of traffic patterns numbered from the first to the Mth.

[0042] In step S11 shown in FIG. 3, initialize to i = 1 and j = 1. However, i corresponds to M traffic patterns, and i is variably changed one by one from i = 1 to M. j corresponds to N NW models 1P1 to 1Pn, and j is variably changed one by one from j = 1 to N.

[0043] In step S12, the model load test unit 14 reads one of the M traffic patterns from the DB unit 15, and in step S13, inputs it to each of the NW models 1P1 to 1Pn in the NW model unit 21 for a predetermined time.

[0044] In step S14, the test unit 14 outputs the traffic pattern input to the NW model to the learning data input processing unit 16.

[0045] In step S15, the node model information collection unit 17 collects statistical information of the node models 21a to 21d of the NW model.

[0046] In step S16, the model performance measurement unit 18 measures the performance during the operation of each of the NW models 1P1 to 1Pn to which the traffic pattern is input in step S12, and outputs this performance measurement information to the input processing unit 16.

[0047] In step S17, the input processing unit 16 generates learning data by combining the traffic pattern, statistical information, and performance measurement information into a set. This learning data is stored in the DB unit 19.

[0048] Next, in step S18, it is determined whether j = N?. If this determination result is not j = N, then j = j + 1, and in the above step S13, the traffic pattern is input to the next NW model.

[0049] If the determination result of the above step S18 is j = N, then in step S19, it is determined whether i = M?. If this determination result is not i = M, then i = i + 1, and in the above step S12, the next traffic pattern is read from the DB unit 15.

[0050] If the determination result of the above step S19 is i = M, it means that a total of M × N pieces of learning data have been recorded in the DB unit 19. In step S20, the inference unit 20 generates the above inference device 20a based on the learning data stored in the DB unit 19.

[0051] <Third communication NW modeling support function> The third support function sequentially inputs a plurality of types (assumed to be L types) of traffic patterns stored in the traffic pattern DB unit 15 to the actual NW 31. At this time, the model load test unit 14 stops. The actual node information collection unit 11 collects statistical information (described later) of the actual nodes 32a to 32d that operate according to the above L types of traffic patterns. The measurement unit 24 measures the performance of the actual NW 31 that operates according to the L types of traffic patterns. A set of the information of the traffic pattern input to the actual NW 31, this performance measurement result, and the statistical information of the actual nodes 32a to 32d is input to the inference device 20a of the NW model inference unit 20 described in the above second support function, and the inference device 20a estimates the NW model that is most approximate to the actual NW 31 among the N NW models 1P1 to 1Pn. However, the statistical information is, for example, the setting information of the actual nodes 32a to 32d, log information, statistical result information such as the arrival number of packets for each actual node or each queue within the actual node, and the usage rate of the processor.

[0052] This third support function is realized by using the actual node information collection unit 11, the traffic pattern DB unit 15, the NW model inference unit 20, the actual environment load test unit 22, the inference unit input processing unit 23, and the actual environment performance measurement unit 24.

[0053] The actual environment load test unit 22 sequentially reads L types of traffic patterns from the DB unit 15 and inputs them to the actual NW 31 for a predetermined time. Also, the input traffic pattern information is input to the inference unit input processing unit 23.

[0054] The actual node information collection unit 11 collects the statistical information of the actual nodes 32a to 32d that operate according to L types of traffic patterns and inputs it to the inference unit input processing unit 23.

[0055] The actual environment performance measurement unit 24 measures the performance of the actual NW 31 that operates according to L types of traffic patterns, and inputs this performance measurement information to the inference unit input processing unit 23.

[0056] For each input of a certain traffic pattern to the actual NW 31, the inference unit input processing unit 23 associates the information of the input traffic pattern with the performance measurement information of the actual NW 31 and the statistical information of the actual nodes 32a to 32d at the time of the traffic pattern input, converts it into a data format corresponding to the above learning data, and inputs it to the NW model inference unit 20.

[0057] The NW model inference unit 20 inputs the input data to the inference device 20a described in the above second support function, and estimates and selects the NW model (for example, NW model 1P1) that is considered to be closest to the actual NW 31 from among N NW models 1P1 to 1Pn.

[0058] <Operation of the Third Communication NW Modeling Support Function> Next, the communication NW modeling support process by the third support function of the support device 10 will be described with reference to the flowchart shown in FIG. 4.

[0059] However, it is assumed that the traffic pattern DB unit 15 stores L types of traffic patterns numbered from the first to the L-th.

[0060] In step S21 shown in FIG. 4, it is initially set to k = 1. However, k corresponds to L traffic patterns, and k is variably changed one by one from k = 1 to L.

[0061] In step S22, the actual environment load test unit 22 reads one of the L types of traffic patterns from the DB unit 15 and inputs it to the actual NW 31 for a predetermined time.

[0062] In step S23, the actual node information collection unit 11 collects statistical information of the actual nodes 32a to 32d that operate according to the input traffic pattern, and inputs it to the inference unit input processing unit 23.

[0063] In step S24, the actual environment performance measurement unit 24 measures the performance of the actual NW 31 that operates according to the input traffic pattern, obtains this performance measurement information, and inputs it to the inference unit input processing unit 23.

[0064] In step S25, for each input of a certain traffic pattern to the actual NW 31, the inference unit input processing unit 23 associates the information of the input traffic pattern with the performance measurement information of the actual NW 31 and the statistical information of the actual nodes 32a to 32d at the time of input of the traffic pattern, converts it into a data format corresponding to the above learning data, and inputs it to the inference unit 20.

[0065] In step S26, the inference unit 20 inputs the input data to the inference device 20a described in the above second support function, and estimates and selects the NW model (for example, NW model 1P1) that is considered to be the closest to the actual NW 31 from the N NW models 1P1 to 1Pn.

[0066] In step S27, the inference unit 20 determines whether or not the estimation error when the information of the traffic pattern, the performance measurement information of the actual NW 31, and the statistical information of the actual nodes 32a to 32d are used as inputs to the inference device 20a is equal to or less than a predetermined threshold value (for example, 5%). As a result, if it is equal to or less than the threshold value (Yes), this support process is terminated.

[0067] On the other hand, if it is not below the threshold value (No), in step S28, it is determined whether k = L? If this determination result is not k = L, then k = k + 1, and returning to the above step S22, the next traffic pattern is read from the DB unit 15.

[0068] On the other hand, if the determination result is k = L, then this support process is terminated.

[0069] <Effect of the Embodiment> The effect of the communication network modeling support device 10 according to the embodiment of the present invention will be described.

[0070] (1a) This support device 10 (first support function) includes an actual node information collection unit 11, a generation unit 12, a NW model pattern DB unit 13 (first DB), and a NW model unit 21.

[0071] The NW model unit 21 corresponds to the actual NW 31 in which a plurality of actual nodes 32a to 32d as communication devices are NW-connected and deployed, and constructs NW models 1P1 to 1Pn to which a plurality of node models 21a to 21d are connected on a simulator.

[0072] The actual node information collection unit 11 collects routing setting information and node internal information from each of the actual nodes 32a to 32d of the actual NW 31. The generation unit 12 estimates the connections between the actual nodes 32a to 32d based on the routing setting information, and generates a plurality of NW model patterns that constitute the node models 21a to 21d corresponding to the actual nodes 32a to 32d based on the node internal information in which the node internal information is changed into a plurality of types. The DB unit 13 stores the plurality of generated NW model patterns.

[0073] The NW model unit 21 is configured to construct a plurality of NW models 1P1 to 1Pn to which the respective node models 21a to 21d corresponding to the plurality of NW model patterns acquired from the DB unit 13 are connected on a simulator.

[0074] According to this configuration, the support device 10 is provided with NW models 1P1 to 1Pn each having node models 21a to 21d corresponding to the respective actual nodes 32a to 32d of the actual NW 31, based on the routing setting information and node internal information acquired from the respective actual nodes 32a to 32d of the actual NW 31. For this reason, the NW model unit 21 including the NW models 1P1 to 1Pn that accurately reflect the performance characteristics of the actual NW 31 can be constructed. Since the internal structures of the actual nodes 32a to 32d of the actual NW 31 are basically unknown, such as a black box, it is assumed that if various NW model patterns in which the number of queues, length, clock number of the processor, etc. in the NW models 1P1 to 1Pn are changed are prepared, there is a configuration close to the actual NW 31 among the respective NW model patterns. That is, if a sufficient number of N NW model patterns are prepared, the NW models 1P1 to 1Pn that accurately reflect the performance characteristics of the actual NW 31 will be included therein.

[0075] (2a) The support device 10 (second support function) includes a traffic pattern DB unit (second DB) 15 that stores a plurality of types of traffic patterns for changing the load of nodes, and a model load test unit 14 that inputs the plurality of traffic patterns acquired from the DB unit 15 to each of the NW models 1P1 to 1Pn of the NW model unit 21.

[0076] Further, the support device 10 includes a node model information collection unit 17 that collects statistical information of each of the node models 21a to 21d for each of the NW models 1P1 to 1Pn that operate with the input of the traffic pattern, and a model performance measurement unit 18 that measures the performance for each of the NW models 1P1 to 1Pn that operate with the input of the traffic pattern to obtain performance measurement information.

[0077] Furthermore, the support device 10 performs learning based on learning data that combines the input traffic pattern, statistical information, and performance measurement information for each NW model for which the traffic pattern has been input, and from among the plurality of NW models 1P1 to 1Pn, when a traffic pattern is applied to a certain NW, it selects the NW model (for example, NW model 1P1) having the characteristics closest to (approximating) the traffic pattern, statistical information, and performance measurement information. It is configured with an inference unit 20 that generates an inference device 20a for estimating from among the plurality of NW models 1P1 to 1Pn.

[0078] According to this configuration, an inference device 20a for estimating the NW model that most closely approximates the actual NW 31 can be constructed from among the plurality of NW models 1P1 to 1Pn in the NW model unit 21.

[0079] (3a) The support device 10 (third support function) includes a DB unit 15, a model load test unit 14, a node model information collection unit 17, and a model performance measurement unit 18. Furthermore, it includes a traffic pattern DB unit (second DB) 15 that stores a plurality of types of traffic patterns for changing the load of the node, an actual environment load test unit 22 that inputs a plurality of NW model patterns acquired from the DB unit 15 to each actual node 32a to 32d of the actual NW 31, and an actual environment performance measurement unit 24 that measures the performance of the actual NW 31 activated by the input of the traffic pattern to obtain performance measurement information.

[0080] The actual node information collection unit 11 collects statistical information on the actual nodes 32a to 32d activated by the input of the traffic pattern.

[0081] The inference unit 20 is configured to estimate, from among the plurality of NW models 1P1 to 1Pn, the NW model (for example, NW model 1P1) that is closest to (approximates) the actual NW 31 by inputting the input data in which the traffic pattern input to the actual NW 31, the actual NW 31 performance measurement information, and the statistical information of the actual nodes 32a to 32d are combined into the inference device 20a.

[0082] According to this configuration, based on the performance measurement information of the actual NW 31 for which the traffic pattern is input and the statistical information related to each of the actual nodes 32a to 32d, it is possible to estimate and select the NW model that most closely approximates the actual NW 31 from among the plurality of NW models 1P1 to 1Pn in the NW model unit 21.

[0083] <Program> Next, the program executed by the computer of the present embodiment will be described. It is assumed that the computer is a communication network modeling support device 10 including an NW model unit 21 having a plurality of NW models 1P1 to 1Pn to which a plurality of node models 21a to 21d are connected, corresponding to the actual NW 31 in which a plurality of actual nodes 32a to 32d as communication devices are NW-connected and deployed.

[0084] This program causes the above computer to function as means for collecting routing setting information and node internal information from each of the actual nodes 32a to 32d of the actual NW 31, estimating the connections between the actual nodes 32a to 32d based on the routing setting information, and generating a plurality of NW model patterns constituting the node models 21a to 21d corresponding to the actual nodes 32a to 32d based on the node internal information in which the node internal information is changed into a plurality of types, means for storing the plurality of generated NW model patterns, and means for causing the NW model unit 21 to function as having a plurality of NW models 1P1 to 1Pn to which the respective node models 21a to 21d corresponding to the plurality of NW model patterns acquired from the DB unit 13 are connected.

[0085] According to this program, similar to the effect of the communication network modeling support device 10 described above, it is possible to construct NW models 1P1 to 1Pn that accurately reflect the performance characteristics of the actual NW 31. However, the program is stored in a storage medium, and the CPU (Central Processing Unit) reads the program from the storage medium and executes it.

[0086] <Effect> (1) A NW model unit that constructs a NW model with a plurality of node models connected on a simulator, corresponding to an actual NW in which a plurality of actual nodes as communication devices are deployed with NW connection; an actual node information collection unit that collects routing setting information and node internal information from each actual node of the actual NW; a generation unit that estimates connections between each actual node based on the routing setting information and generates a plurality of NW model patterns that configure node models corresponding to the actual nodes based on the node internal information changed into a plurality of types; and a first DB (Data Base) that stores the plurality of generated NW model patterns, wherein the NW model unit constructs a plurality of NW models with each node model corresponding to the plurality of NW model patterns acquired from the first DB connected on the simulator. A communication network modeling support device characterized by this is provided.

[0087] According to this configuration, based on the routing setting information and node internal information acquired from each actual node of the actual NW, a NW model having each node model corresponding to each actual node is provided. Therefore, a NW model that accurately reflects the performance characteristics of the actual NW can be constructed.

[0088] (2) A second DB that stores a plurality of types of traffic patterns for changing the load of nodes; a model load test unit that inputs the plurality of traffic patterns acquired from the second DB into each NW model of the NW model unit; a node model information collection unit that collects statistical information of each node model for each NW model that operates with the input of the traffic pattern; a model performance measurement unit that measures the performance of each NW model that operates with the input of the traffic pattern to obtain performance measurement information; and an inference unit that generates an inference device that estimates, from among the plurality of NW models constructed in the NW model unit, a NW model having characteristics approximated to the traffic pattern, the statistical information, and the performance measurement information by learning with the learning data combining the traffic pattern, the statistical information, and the performance measurement information. The communication network modeling support device according to (1) above is characterized by including this.

[0089] According to this configuration, an inference device can be generated that estimates an NW model having characteristics approximate to the traffic pattern input to the NW model, the statistical information related to the NW model, and the performance measurement information, from among a plurality of NW models in the NW model section.

[0090] (3) The communication network modeling support device according to (2) above, comprising: a load test unit for the actual environment that inputs a plurality of traffic patterns acquired from the second DB to the actual NW; and a performance measurement unit for the actual environment that measures the performance of the actual NW operating with the input of the traffic pattern to obtain performance measurement information, wherein the actual node information collection unit collects statistical information of actual nodes operating with the input of the traffic pattern by the load test unit for the actual environment, and the inference unit inputs input data in which the traffic pattern input to the actual NW, the performance measurement information of the actual NW, and the statistical information of the actual nodes are combined, to the inference device, so as to estimate, from among a plurality of NW models constructed in the NW model section, an NW model approximate to the actual NW.

[0091] According to this configuration, based on the performance measurement information related to the actual NW with the input of the traffic pattern and the statistical information related to each actual node, and the input traffic pattern, it is possible to estimate and select an NW model that is most approximate to the actual NW from among a plurality of NW models in the NW model section.

[0092] In addition, regarding specific configurations, appropriate changes can be made without departing from the gist of the present invention.

Explanation of Signs

[0093] 10 Communication network modeling support device 11 Actual node information collection unit 12 NW model pattern generation unit (generation unit) 13 NW model pattern DB unit (first DB) 14 Load test unit for the model 15 Traffic pattern DB unit (second DB) 16 Learning data input processing unit 17 Node Model Information Collection Unit 18 Model Performance Measurement Unit (Measurement Unit) 19 Learning Data DB Unit 20 NW Model Inference Unit (Inference Unit) 20a Inference Engine 21 NW Model Unit 21a~21d Node Models 22 Real Environment Load Test Unit 23 Inference Unit Input Processing Unit 24 Real Environment Performance Measurement Unit 31 Real NW 32a~32d Real Nodes 1P1~1Pn NW Models

Claims

1. An NW model unit that constructs an NW model in which a plurality of node models are connected on a simulator, corresponding to an actual NW in which a plurality of actual nodes as communication devices are NW-connected and deployed; An actual node information collection unit that collects routing setting information and node internal information from each actual node of the actual NW; A generation unit that estimates connections between each actual node based on the routing setting information and generates a plurality of NW model patterns that constitute a node model corresponding to an actual node based on the node internal information in which the node internal information is changed into a plurality of types; A first DB (Data Base) that stores the plurality of generated NW model patterns; A second DB that stores a plurality of types of traffic patterns for changing the load of nodes; A model load test unit that inputs a plurality of traffic patterns acquired from the second DB into each NW model of the NW model unit; A node model information collection unit that collects statistical information of each node model for each NW model that operates with the input of the traffic pattern; A model performance measurement unit that measures the performance of each NW model that operates with the input of the traffic pattern to obtain performance measurement information; An inference unit that generates an inference device that estimates, from among the plurality of NW models constructed in the NW model unit, an NW model having characteristics approximated to the traffic pattern, the statistical information, and the performance measurement information by performing learning with learning data combining the traffic pattern, the statistical information, and the performance measurement information; and comprising The NW model unit constructs a plurality of NW models on a simulator in which each node model corresponding to the plurality of NW model patterns acquired from the first DB is connected. A communication network modeling support device characterized by the above.

2. An actual environment load test unit that inputs a plurality of traffic patterns acquired from the second DB into the actual NW; A performance measurement unit for a real environment that measures the performance of the real NW operating with the input of the traffic pattern to obtain performance measurement information; further comprising; The real node information collection unit collects statistical information of real nodes operating with the input of the traffic pattern by the real environment load test unit. The inference unit inputs input data combined with the traffic pattern input to the real NW, the performance measurement information of the real NW, and the statistical information of the real nodes to the inference device, and estimates an NW model approximating the real NW from among a plurality of NW models constructed in the NW model unit. The communication network modeling support device according to claim 1, characterized in that.

3. A communication network modeling support method by a communication network modeling support device, comprising: The communication network modeling support device corresponds to a real NW in which a plurality of real nodes as communication devices are NW-connected and deployed, and includes an NW model unit that constructs an NW model in which a plurality of node models are connected on a simulator, and a second DB that stores a plurality of types of traffic patterns that change the load of nodes; collecting routing setting information and node internal information from each real node of the real NW; estimating connections between each real node based on the routing setting information, and generating a plurality of NW model patterns corresponding to the real NW based on the node internal information in which the node internal information is changed into a plurality of types; storing the plurality of generated NW model patterns in a first DB; constructing a plurality of NW models in which each node model corresponding to the plurality of NW model patterns acquired from the first DB is connected on the simulator by the NW model unit; inputting a plurality of traffic patterns acquired from the second DB to each NW model of the NW model unit; The step of collecting statistical information of each node model for each NW model operating with the input of the traffic pattern; The step of measuring the performance of each NW model operating with the input of the traffic pattern to obtain performance measurement information; By performing learning using the learning data combining the traffic pattern, the statistical information, and the performance measurement information, generating an inference device that estimates, from among a plurality of NW models constructed in the NW model unit, an NW model having characteristics approximated to the traffic pattern, the statistical information, and the performance measurement information; A communication network modeling support method characterized by executing the above.

4. The communication network modeling support device The step of inputting a plurality of traffic patterns acquired from the second DB into the actual NW; The step of measuring the performance of the actual NW operating with the input of the traffic pattern to obtain performance measurement information; The step of collecting statistical information of actual nodes operating with the input of the traffic pattern; The step of inputting the input data combining the traffic pattern input to the actual NW, the performance measurement information of the actual NW, and the statistical information of the actual nodes into the inference device, and estimating, from among a plurality of NW models constructed in the NW model unit, an NW model approximated to the actual NW; The communication network modeling support method according to claim 3, characterized by executing the above.

5. A program for causing a computer to function as the communication network modeling support device according to claim 1 or claim 2.

Images