Network failure location specification device, network failure location specification method, and program

Abstract

To surely specify a failure location as possible as measuring the number of times to be limited under a situation where a network construction element indicates a random behavior.SOLUTION: A network failure location specification device for specifying a failure location of an objective network having a network construction element in which a canalization state is modulated in random, comprises: a state number reduction part that reduces previously a candidate of a network state expressed by a probability value; a test optimization part that calculates a test of an optimized pass measurement which has to be executed against the object network on the basis of a candidate of the network state where the reduction is performed by the state number reduction part; a test execution part that executes the test against the object network; and a test result analysis part that restricts the network state in accordance with the result of the test by the test execution part.SELECTED DRAWING: Figure 4

Description

[Technical field]

[0001] The present invention relates to a technique for identifying a fault location in a network. [Background technology]

[0002] In the following description, references are indicated by numbers, such as [1], and the names of the references corresponding to the numbers are listed at the end of the specification.

[0003] In recent years, "network tomography" has been attracting attention as a means of understanding the internal state of a communication network. Network tomography can estimate characteristics such as delay, packet loss, congestion, and failures at the level of links and nodes, which are network components, based on limited end-to-end traffic measurements.

[0004] In today's complex ICT systems, there are many failures that are difficult to identify and localize, such as silent failures that do not trigger alarms even when monitored on a device-by-device basis, and software-related failures that do not appear in device-layer logs. In such situations, network tomography based on end-to-end actual measurement results can accurately detect and identify abnormalities, greatly improving the quality and stability of network operations.

[0005] In particular, methods that estimate the binary state (faulty / not faulty) of each network component (node, link) are called binary network tomography and have been actively researched [1].

[0006] In network tomography, it is important to assume that the environment exhibits stochastic behavior. In fact, in protocols such as load balancing mechanisms and ECMP, which distribute traffic to paths with equal weights, situations arise where routing behaves stochastically. In addition to routing, each network component may exhibit stochastic behavior, such as stochastic failure of communication even in normal conditions, or intermittent failures that cause intermittent outages.

[0007] There are a few existing methods for network tomography in a stochastic environment. In particular, Non-Patent Document 1 (Reference [2]) discloses a method for estimating the binary state of network components according to the Bayesian estimation framework by prioritizing path measurements with a large amount of mutual information, i.e., the path measurements that are expected to narrow down the fault location the most, in order to pinpoint the fault location as accurately as possible with a limited number of path measurements under stochastic routing. [Prior art documents] [Non-patent literature]

[0008] [Non-Patent Document 1] Ikeuchi, Hiroki, Hiroshi Saito, and Kotaro Matsuda. "Network Tomography based on Adaptive Measurements in Probabilistic Routing." IEEE INFOCOM 2022-IEEE Conference on Computer Communications. IEEE, 2022. Summary of the Invention [Problem to be solved by the invention]

[0009] In a stochastic environment, due to the uncertainty of routing and the state of network components, a large number of path measurements are required to correctly identify the fault location. In general, a large number of path measurements should be avoided as much as possible because they may prolong the time to identify the fault location or place a heavy load on the network. Therefore, the existing network tomography method (Non-Patent Document 1), which identifies faults with a small number of path measurements, is useful.

[0010] However, although this technique can be applied to probabilistic routing, it cannot be applied to cases where the network components behave probabilistically, that is, where the connectivity of links varies probabilistically.

[0011] The present invention has been made in consideration of the above points, and aims to provide a fault location identification technology for identifying the fault location as accurately as possible with a limited number of measurements in a situation where network components exhibit probabilistic behavior. [Means for solving the problem]

[0012] According to the disclosed technology, there is provided a network fault location identification device for identifying a fault location in a target network having network elements whose communication state changes probabilistically, the network fault location identification device comprising: a state number reduction unit that reduces in advance candidates of network states represented by probability values; a test optimization unit that calculates an optimal path measurement test to be performed on the target network based on the network state candidates reduced by the state number reduction unit; a test execution unit that executes a test on the target network; a test result analysis unit that narrows down the network state according to the results of the test performed by the test execution unit; There is provided a network fault location identification device comprising: Effect of the Invention

[0013] According to the disclosed technique, a fault location technique is provided for identifying a fault location as accurately as possible with a limited number of measurements in a situation where network components exhibit stochastic behavior. [Brief description of the drawings]

[0014] [Figure 1] 1 is a system configuration diagram according to an embodiment of the present invention. [Diagram 2] FIG. 1 is a diagram showing Algorithm 1. [Diagram 3] This is a diagram showing Procedure 2. [Figure 4] FIG. 1 illustrates an example of the configuration of a network failure point identification device 100. [Diagram 5] 4 is a flowchart showing a processing procedure of the network failure point identification device 100. [Figure 6] FIG. 1 is a diagram showing the topology used in the evaluation. [Figure 7] FIG. [Figure 8] FIG. [Figure 9] FIG. [Figure 10] FIG. [Figure 11] FIG. [Figure 12] FIG. [Figure 13] FIG. 2 is a diagram illustrating an example of a hardware configuration of a network failure point identification device 100. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] Hereinafter, an embodiment of the present invention (the present embodiment) will be described with reference to the drawings. The embodiment described below is merely an example, and the embodiment to which the present invention is applied is not limited to the following embodiment.

[0016] (Overall system configuration, operation overview) FIG. 1 shows an example of the overall configuration of a system according to this embodiment. As shown in FIG. 1, this system has a configuration in which a network fault location identification device 100 is connected to a target network 200, which is a network for which a network fault location is to be identified. The network fault location identification device 100 identifies a fault location by repeating one or more times a process in which the path measurement device performs a test related to path measurement and obtains the results based on parameter information of network components whose communication state changes probabilistically in addition to topology information of routing in the target network 200, routing information that can change dynamically, and constraint information of the path measurement device. Note that in FIG. 1, the network fault location identification device 100 may have a path measurement device, or the target network 200 may be provided with a path measurement device, and the network fault location identification device 100 may instruct the path measurement device to perform path measurement.

[0017] The target network 200 has multiple nodes and multiple links, and data is sent and received by probabilistic routing. The communication state of each of the links and nodes varies probabilistically. The links may also be called branches or edges.

[0018] In this embodiment, the network fault location identification device 100 identifies the fault location as accurately as possible with a limited number of path measurements in a situation where the network components exhibit probabilistic behavior in addition to probabilistic routing. Even in a probabilistic environment, the network fault location identification device 100 prioritizes path measurements that are effective for fault identification, so that the number of path measurements required to identify the fault is reduced, and it is expected that the time required for fault identification will be shortened and the network load will be reduced.

[0019] In order to identify the fault location as described above, the network fault location identification device 100 performs modeling using a vector that lists the communication probability of each link as the network state. This makes it possible to express not only the probabilistic behavior of routing, but also the inherent probability of network components. By implementing this modeling, it becomes possible to identify the fault by using the method of selecting a preferential path based on mutual information, which is found in the conventional method [2].

[0020] The process executed by the network fault location identification device 100 will now be described in detail.

[0021] (Problem setting) First, the formulation of the problem in this embodiment will be described. However, the technology according to this embodiment can be applied even in situations that do not strictly follow the following definition. For example, although the following formulation is performed regarding the identification of a link failure, the technology according to this embodiment can be similarly applied to the identification of a node failure.

[0022] For convenience of description, in the text of this specification (text that is not an image), vector and matrix characters that are generally written in bold are written in normal font. Even if they are written in normal font, it is clear from the context that they are characters that indicate vectors or matrices. Other characters that indicate sets, etc., are also written in normal font in the text of this specification.

[0023] Let the network of interest be an undirected graph G(V,E). V={v i} i is the set of vertices, E={e j} j is the edge set. A "network state" or simply a "state" is one of the candidate set S = {1, , |S|}. Here, we assume that the network can potentially take a discrete number of states, and S is the set of all such states, with its elements, the natural numbers 1, , |S|, representing (arbitrarily numbered) states. Each state s∈S is represented by a vector q s =(q s1 , ,qs|E| )∈[0,1] |E| ), where q sj is a branch j represents the probability that the branch e is incommunicable under state s. j Whether each packet passing through is discarded or not is determined by the Bernoulli distribution Ber(q sj ) shall be determined independently in accordance with the

[0024] In the above example, we defined the state s∈S a priori, but we also defined the vector {q s} s We can think of it as first coming into being, and then assigning an index s∈S to the potential network state that defines them.

[0025] The monitoring path set A={a1, ,a |A|} is a binary vector a k =(a k1 ,···,a k|E| )∈{0,1} |E| A monitoring path is a path consisting of several connected branches along which measurement packets are sent. kj If =1, monitor path a k E j Since the monitoring path can be regarded as simply a set of edges, paths with loops are also allowed.

[0026] Monitoring path a k When the test is performed, a k If even one branch in is abnormal, the result is 1, and if all branches are normal, the result is 0. In other words, a result of 1 indicates that the monitoring path is not connected, and a result of 0 indicates that the monitoring path is connected. In this embodiment, probabilistic routing is assumed, so k It is not possible to directly specify a path, but only to specify a "group" of multiple monitoring paths that have the same source vertex (S node) and destination vertex (D node).

[0027] When a group c∈C={1, ,|C|} is specified, the monitoring path a k The test is independent with probability p ck Assume that the algorithm is executed with (c=1,...,|C|,k=1,...,|A|), where the following equation holds:

[0028]

number

[0029]

number

[0030] In actual network operations, monitoring systems are sometimes designed to send probe packets in multiple directions at the same time, or to send multiple packets in the same direction at once to obtain statistics. In light of this situation, the term "probe test" or simply "test" is often used. x ∈X={ξ1, ,ξ |X| In this embodiment, ξ x =(ξ x1 ,···,ξ x|C| )∈Z + |C| When you run this, c becomes ξ xc times (c=1, , |C|), where Z + represents the set of all non-negative integers. That is, ξ x is a package of simultaneous measurements between different node pairs or multiple measurements as one test. To ensure that each test contains the same number of measurements, |ξ x |=Σ c ξ xc It is natural to treat as a constant independent of x.

[0031] The problem with this embodiment is as follows.

[0032] Now, assume that the state s∈S is unknown. Also assume that the prior distribution of s and the number of tests ξ that can be performed N are given. In this case, what is the most efficient strategy for performing the tests (when and how many times to perform each test) in order to identify the true state as accurately as possible by performing N tests? Also, how should the true state be estimated based on the results of the tests?

[0033] In the above problem setting, in this embodiment, a vector q s The important point is that we have introduced the vector . In the existing technology [2], this vector was a binary vector (each component can be either 0 or 1), so each branch could only take a deterministic state. On the other hand, in the formulation in this embodiment, each component can take any real number between 0 and 1, which corresponds to the probability of failure, so probabilistic behavior can be expressed.

[0034] The network tomography method described later is a method for finding a finite set of vectors of arbitrary probability values, {q s} s However, in reality, the following two assumptions (i) and (ii) can often be made about the network state.

[0035] (i) Each branch can be considered to be in either a "good state" (a state in which it operates almost normally) or a "bad state" (a state in which it is difficult to operate normally due to congestion or failures), and (ii) it is rare for multiple branches to be in a "bad state" at the same time.

[0036] In view of the above, the network state {q s} s It is also useful to restrict and define as follows:

[0037] First, we define a set of vectors (the following formula) that represent the binary states of the edges.

[0038]

number

[0039] each ~ q s =( ~ q s1 , , ~ q s|E| ), network state q s =(q s1 , ,q s|E| )of, ~ q sj =1 then q sj = ε, ~ q sj =0 then q sj = 1 - ε, where ε<< 1 is a known small real number. ε can be interpreted as the degree of fluctuation in the state of the branch (if ε = 0, this leads to the discussion of deterministic states in reference [2]). In the "Effects of the embodiment" section described later, the network state will be defined using the method described here.

[0040] (Details of Processing by Network Fault Identification Device 100) The network tomography technique executed by the network fault location locating apparatus 100 will now be described in detail.

[0041] The method itself can be used as is as disclosed in reference [2], except for the preprocessing. The approximation guarantee disclosed in reference [2] also holds true as is. Below, only the main points of the parts that overlap with the conventional technology disclosed in reference [2] are explained, and the preprocessing is explained in detail because it is an original part of the technology according to the present embodiment.

[0042] In this embodiment, an "adaptive measurement" approach is taken. This means that the network fault location identification device 100 sequentially determines the next test to be performed according to the results of the tests obtained currently. Specifically, it is formulated as the following batch processing.

[0043] Divide N tests into B batches of size N B That is, N B ×B = N holds. In the b-th batch ((b ∈ [1, B] ∩ Z + ), determine the following test design.

[0044]

Number

[0045]

Number

[0046]

Number

[0047]

Number

[0048] In designing m (the subscript b of m is omitted as appropriate), the "goodness" of m must be quantitatively expressed. In this embodiment, y m As an indicator of the effectiveness of m The mutual information between and S is I(S;Y m ) will be used (y m When considering s as random variables, Y m , S are used in capital letters). This allows the problem considered in this embodiment to be written as follows:

[0049] [Problem]: Given the prior distribution Pr(s) and the following measurements obtained up to the (b-1)th batch,

[0050]

number

[0051]

number

[0052]

number

[0053]

number

[0054]

number

[0055] The above "problem" is a combinatorial optimization problem, and can be shown to be NP-hard. To solve this problem, any approximate optimization method or optimization method can be used in this embodiment. As an example, the method disclosed in reference [2] can be applied.

[0056] The procedure (algorithm) of all the processes executed by the network fault location identification device 100 is shown in FIG.

[0057] First, in the second line, there is a preprocessing step (StateSpaceReduction) that reduces the number of state candidates. However, since the method in reference [2] cannot be applied to this preprocessing step, we will explain it in detail later.

[0058] In the batch processing starting from line 5, m is created based on the greedy method in lines 6 to 10 (this corresponds to the approximate solution of the above "problem"). In line 8, there is a calculation of mutual information CalcMI, which can be calculated using the method disclosed in reference [2]. In the while loop, I(S;Y m |D b-1 ) with the largest increment x_max Sequentially select the xth maxAfter creating m, we increment the component y m We obtain the posterior distribution Pr(s|D b ) is updated. The above process is repeated for the number of batches (corresponding to the for statement in line 4). The state that maximizes the final posterior distribution is output as the estimated state (line 14).

[0059] Algorithm 1 is an approximation algorithm based on the greedy method, but it can be shown that it has a constant approximation in the sense of maximizing mutual information as follows. In other words, even in the worst case, it is guaranteed that the mutual information will be equal to or greater than a certain value.

[0060] [Theorem]:N y When is sufficiently large, m obtained by Algorithm 1 b of ~ m b Then, the optimal m b is as follows:

[0061]

number

[0062]

number

[0063] A similar process was also described in reference [2], but the method used the fact that the U matrix contains many components whose values ​​are 0 or 1. This method uses the fact that the state of the network components is deterministic (ε = 0), and cannot be applied to the problem setting of this embodiment. Below, we will explain a preprocessing method that can be generally applied without depending on the details of the U matrix.

[0064] The pre-processing procedure executed by the network fault location identification device 100 is shown in Procedure 2 of FIG. 3 (this process is incorporated inside Algorithm 1, so it is called Procedure instead of Algorithm to distinguish it).

[0065] The basic idea of ​​Procedure 2 is to estimate the random variable Y c (which takes the value 0 or 1) and the mutual information I(S;Y c ) to find a group c that is considered useful for identifying the state.

[0066] To do this, first, we select N states {s (l)} l Each sample s (l) For each c, generate a measurement result (1 or 0) by random number simulation. This is used as a data set, and the approximate value of the mutual information ^I(S;Y c ) can be calculated. For example, the method described in reference [4] can be used to calculate this approximation.

[0067] This process allows us to quantitatively evaluate how effective each group c is in identifying the state. Then, we select the group c that maximizes this mutual information. max ξ that includes x We arbitrarily select one of them and execute it (lines 4-5). Using the result, we update the state distribution Pr(s) based on Bayesian estimation. We repeat the above process a certain number of times N iter After iterations, we delete the states whose probability is smaller than the threshold δ. We also delete the corresponding columns of the U matrix (line 8).

[0068] This preprocessing can greatly reduce the number of state candidates, which speeds up the processing of the algorithm in Figure 2. If the prior distribution Pr(s) is uncertain, it can be designed using prior knowledge (e.g., multiple failures are less likely to occur than single failures) or a uniform distribution can be used.

[0069] (Example) As an embodiment of the above-mentioned processing, a configuration example of the network fault location identification device 100 and a processing procedure example using this configuration will be described.

[0070] Fig. 4 shows a configuration example of network fault location identification device 100. As shown in Fig. 4, network fault location identification device 100 has an input UI 110, a number of states reduction unit 120, a test execution unit 130, a test optimization unit 140, a test result analysis unit 150, and an output UI 160. The number of states reduction unit 120 and the test result analysis unit 150 are connected to a target network 200 as shown in the figure.

[0071] The processing steps of the network fault location identification device 100 having the above configuration will be described with reference to the flowchart of FIG.

[0072] In S101, first, data and parameters (prior distribution Pr(s), U, N B , B, ε, etc. In S102, the state number reduction unit 120 determines ξ based on these in accordance with the second line of Algorithm 1 (Procedure 2), and passes it to the test execution unit 130.

[0073] In S103, the test execution unit 130 executes a test on the target network using a connectivity check program such as ping. In S102 again, the state number reduction unit 120 determines the next ξ according to Procedure 2 based on the obtained result.

[0074] After steps S102 and S103 have been performed a set number of times, in step S104, test optimization unit 140 creates m according to the greedy algorithm or another optimization algorithm in lines 5 to 10 of Algorithm 1. This is passed to test execution unit 130, which executes a test on the target network in step S105 (corresponding to line 11 of Algorithm 1).

[0075] The result is passed to the test result analysis unit 150, which calculates the posterior distribution according to the Bayesian estimation framework in S106 (corresponding to line 13 of Algorithm 1).

[0076] The obtained posterior distribution is passed to the test optimization unit 140 engine, which repeats the above steps (S104 to S106) according to the loop of Algorithm 1 or another optimization algorithm. After a predetermined number of steps, an estimated state is output as the maximum likelihood value from the final posterior distribution to the output UI 160. In S107, the output UI 160 outputs the estimated state.

[0077] (Summary of the embodiment) As described above, the network fault location identification device 100 according to this embodiment locates the fault location by repeating, one or more times, the process in which the path measurement device performs a test related to path measurement and obtains the results based on the network routing topology information, dynamically changeable routing information, and constraint information of the path measurement device, as well as parameter information of network components whose communication state varies probabilistically.

[0078] The network fault location identification device 100 includes a state number reduction unit 120 for reducing candidate network states in advance in order to reduce the amount of calculations, a test optimization unit 140 for calculating the optimal path measurement test to be executed, a test execution unit 130 for actually executing the test, and a test result analysis unit 150 for narrowing down the network states according to the test results.

[0079] As parameter information of the network components, for example, the communication probability of each link or each node, or a vector expressed as a set of these communication probabilities, can be used.

[0080] The state number reduction unit 120, for example, calculates the mutual information between a random variable representing the result of performing a certain measurement and a random variable representing the state of the network, actually performs a test including a measurement that approximately or exactly maximizes the mutual information, and reduces the number of states based on the posterior distribution calculated as a result.

[0081] (Effects of the embodiment) The technology according to the present embodiment described above makes it possible to identify a fault location as accurately as possible with a limited number of path measurements in a situation where network components exhibit probabilistic behavior in addition to probabilistic routing. With this technology, even in a probabilistic environment, path measurements that are effective for fault identification are prioritized, so the number of path measurements required to identify a fault is reduced, and it is expected that the time required to identify a fault will be shortened and the network load will be reduced.

[0082] The evaluation was carried out using the three network data sets in Figure 6 [5]. In Figure 6, #bad edges represents the number of edges that are simultaneously in a bad state, and |S| is the total number of possible states. We also considered two cases: ε=0.01 and ε=0.05. We considered the following settings for each network.

[0083] The number of groups is |C|=3|V|. For each node, we set it as an S node, and randomly selected three nodes as D nodes to determine node pairs (groups). For each node pair, the shortest path, the second shortest path, and the third shortest path were considered as monitoring paths (|A|=9|V|). For each group, the i-th shortest path was selected with the following probability (l i is the path length).

[0084]

number

[0085] In addition to the technique according to the present embodiment (referred to as PM), we also performed Random, LS [6], and LASSO [7] for comparison. In the Random method, ξ is selected randomly without using mutual information in the method of the present invention.

[0086] LS and LASSO are existing methods proposed as non-adaptive approaches for network tomography in probabilistic routing. For LASSO, we performed the experiment with the hyperparameter λ changed to 0.0001, 0.001, and 0.01.

[0087] The accuracy rate was used as the evaluation index. For a state with two bad branches, the answer was considered correct only when both were identified. Experiments were performed on 30 randomly selected true state patterns, and the accuracy rate was calculated.

[0088] The results of the experiment are shown in Figs. 7 to 12. In each figure, the horizontal axis is the total number of tests N, and the vertical axis is the accuracy rate. It can be seen that the method PM according to the present embodiment exhibits higher performance than the existing methods LS and LASSO. For example, in order to exceed a accuracy rate of 0.9, in the case of ε=0.01, the best existing method requires 160 (Missouri), 320 (ION), and 320 (Ntelos) tests, while 26, 26, and 26 tests are sufficient for PM, respectively. In addition, in the case of ε=0.05, the best existing method requires 640 (Missouri), 1280 (ION), and 640 (Ntelos) tests, while 26, 68, and 68 tests are sufficient for PM, respectively.

[0089] (Hardware configuration example) The network failure point identification device 100 can be realized, for example, by causing a computer to execute a program. This computer may be a physical computer or a virtual machine on the cloud.

[0090] That is, the network fault location identification device 100 can be realized by using hardware resources such as a CPU and memory built into a computer to execute a program corresponding to the processing performed by the network fault location identification device 100. The above program can be recorded on a computer-readable recording medium (such as a portable memory) and stored or distributed. The above program can also be provided via a network such as the Internet or email.

[0091] Fig. 13 is a diagram showing an example of the hardware configuration of the computer. The computer in Fig. 13 includes a drive device 1000, an auxiliary storage device 1002, a memory device 1003, a CPU 1004, an interface device 1005, a display device 1006, an input device 1007, an output device 1008, and the like, all of which are interconnected via a bus BS.

[0092] The program for realizing the processing on the computer is provided by a recording medium 1001 such as a CD-ROM or a memory card. When the recording medium 1001 storing the program is set in the drive device 1000, the program is installed from the recording medium 1001 to the auxiliary storage device 1002 via the drive device 1000. However, the program does not necessarily have to be installed from the recording medium 1001, but may be downloaded from another computer via a network. The auxiliary storage device 1002 stores the installed program as well as necessary files, data, etc.

[0093] The memory device 1003 reads out and stores the program from the auxiliary storage device 1002 when an instruction to start the program is received. The CPU 1004 realizes functions related to the network fault location identification device 100 in accordance with the program stored in the memory device 1003. The interface device 1005 is used as an interface for connecting to a network, etc. The display device 1006 displays a GUI (Graphical User Interface) or the like according to a program. The input device 1007 is composed of a keyboard and mouse, buttons, a touch panel, etc., and is used to input various operation instructions. The output device 1008 outputs the calculation results.

[0094] The following supplementary notes are further disclosed regarding the above embodiment.

[0095] <Additional Notes> (Additional note 1) A network fault location identification device for identifying a fault location in a target network having network elements whose communication state changes stochastically, comprising: a state number reduction unit that reduces in advance candidates of network states represented by probability values; a test optimization unit that calculates an optimal path measurement test to be performed on the target network based on the network state candidates reduced by the state number reduction unit; a test execution unit that executes a test on the target network; a test result analysis unit that narrows down the network state according to the results of the test performed by the test execution unit; A network fault location identification device comprising: (Additional note 2) The network state is represented by a communication probability in the network element. 2. A network fault location identification device according to claim 1. (Additional note 3) The state number reduction unit actually executes a test including a measurement that approximately or exactly maximizes mutual information between a random variable representing a result of execution of a certain measurement and a random variable representing a network state by the test execution unit, and reduces the number of network states based on a posterior distribution calculated based on the result of the test. 3. A network fault location identification device according to claim 1 or 2. (Additional note 4) A network fault location identification method executed by a network fault location identification device for identifying a fault location in a target network having network elements whose communication state changes stochastically, comprising: a state reduction step for reducing in advance the number of network state candidates represented by probability values; a test optimization step of calculating an optimal path measurement test to be performed on the target network based on the candidates of the network state reduced by the state number reduction step; a test execution step of executing a test on the target network; a test result analysis step of narrowing down the network state according to the result of the test performed in the test execution step; A network fault location identification method comprising: (Additional note 5) A non-transitory storage medium storing a program for causing a computer to function as each unit in the network fault location identification device according to any one of claims 1 to 3.

[0096] Although the present embodiment has been described above, the present invention is not limited to such a specific embodiment, and various modifications and variations are possible within the scope of the gist of the present invention described in the claims. [References] [1] N. Duffield, "Simple network performance tomography," in Proceedings of the 3rd ACM SIGCOMM conference on Internet measurement. ACM, 2003, pp. 210-215. [2] Ikeuchi, Hiroki, Hiroshi Saito, and Kotaro Matsuda. "Network Tomography based on Adaptive Measurements in Probabilistic Routing." IEEE INFOCOM 2022-IEEE Conference on Computer Communications. IEEE, 2022. [4] Ross, Brian C. "Mutual information between discrete and continuous data sets." PloS one 9.2 (2014): e87357. [5] S. Knight, H. X. Nguyen, N. Falkner, R. Bowden, and M. Roughan, "The internet topology zoo," IEEE Journal on Selected Areas in Communications, vol. 29, no. 9, pp. 1765-1775, 2011. [6] H. Herodotou, B. Ding, S. Balakrishnan, G. Outhred, and P. Fitter, "Scalable near real-time failure localization of data center networks," in Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining. ACM, 2014, pp. 1689-1698. [7] R. Tagyo, D. Ikegami, and R. Kawahara, "Network tomography using routing probability for undeterministic routing," IEICE Transactions on Communications, vol. E104.B, no. 7, pp. 837-848, 2021. [Explanation of symbols]

[0097] 100 Network fault location identification device 110 Input UI 120 State reduction section 130 Test Execution Department 140 Test Optimization Department 150 Test Results Analysis Department 160 Output UI 200 Target Networks 1000 Drive Device 1001 Recording media 1002 Auxiliary storage device 1003 Memory device 1004 CPU 1005 Interface device 1006 Display device 1007 Input Device 1008 Output device

Claims

1. A network fault location identification device for identifying a fault location in a target network having network elements whose communication state changes stochastically, comprising: a state number reduction unit that reduces in advance candidates of network states represented by probability values; a test optimization unit that calculates an optimal path measurement test to be performed on the target network based on the network state candidates reduced by the state number reduction unit; a test execution unit that executes a test on the target network; a test result analysis unit that narrows down the network state according to the results of the test performed by the test execution unit; A network fault location identification device comprising:

2. The network state is represented by a communication probability in the network element. The network fault location identification device according to claim 1.

3. The state number reduction unit actually executes a test including a measurement that approximately or exactly maximizes mutual information between a random variable representing a result of execution of a certain measurement and a random variable representing a network state by the test execution unit, and reduces the number of network states based on a posterior distribution calculated based on the result of the test. The network fault location identification device according to claim 1.

4. A network fault location identification method executed by a network fault location identification device for identifying a fault location in a target network having network elements whose communication state changes stochastically, comprising: a state reduction step for reducing in advance the number of network state candidates represented by probability values; a test optimization step of calculating an optimal path measurement test to be performed on the target network based on the candidates of the network state reduced by the state number reduction step; a test execution step of executing a test on the target network; a test result analysis step of narrowing down the network state according to the result of the test performed in the test execution step; A network fault location identification method comprising:

5. A program for causing a computer to function as each unit in the network fault location locating device according to any one of claims 1 to 3.

Images