Fiber inspection device and fiber inspection method

Abstract

A fiber inspection device 1 establishes, with regard to a fiber group connecting a leaf switch and a spine switch, an inspection path q that includes one uplink and one downlink as fibers to be inspected from the fiber group, inputs test light to the inspection path q, measures a received optical power Pq of the test light that has passed through the inspection path q, and classifies the uplink and downlink as status-unknown fibers if the received optical power Pq is less than a permitted received optical power Pok. The fiber inspection device 1 establishes, with regard to each of the status-unknown fibers, an inspection path q including only one status-unknown fiber as the fiber to be inspected, inputs test light to the inspection path q, measures the received optical power Pq of the test light that has passed through the inspection path q, and classifies the status of the status-unknown fiber on the basis of the received optical power Pq.

Description

Fiber inspection device and fiber inspection method

[0001] The present disclosure relates to a fiber inspection device and a fiber inspection method.

[0002] In recent years, there has been an increasing number of cases where Fiber Cross Connect (FXC) or Optical Circuit Switch (OCS) (hereinafter referred to as optical switch) is introduced into the switch fabric of data centers in order to accommodate traffic generated within the data centers with low power consumption and low cost. Optical switches do not involve photoelectric conversion when switching signals, and do not cause delays due to buffering. By arranging optical switches according to a topology called a CLOS network or a TF-CLOS network, a non-blocking fiber switching network can be constructed.

[0003] Large-scale data center networks (DCNs) and AI clusters need to accommodate a large number of terminals. The large-scale DCN described in Non-Patent Document 1 accommodates approximately 30,000 terminals.

[0004] L. Poutievski et.al., “Jupiter evolving: Transforming Google's datacenter network via optical circuit switches and software-defined networking,” in Proc. ACM SIGCOMM Conf., Aug, pp. 66-85 2022. K. Chen et.al, “DAC: Generic and automatic address configuration for data center networks,” in IEEE / ACM Transactions on Networking, vol. 20, no. 1, pp.84-99, 2011C. Jian, et.al,“MTR: Fault tolerant routing in Clos data center network with miswiring links,” in 2014 IEEE 20th International Workshop on Local & Metropolitan Area Networks (LANMAN), pp. 1-6, 2014X. Ma, et al.“Error tolerant address configuration for data center networks with malfunctioning devices,”in IEEE 32nd International Conference on Distributed Computing Systems, 2012.

[0005] 70% of problems that occur in data centers are caused by human error, with incorrect wiring accounting for a large proportion of these. Expanding a fiber switching network requires manual rewiring, making it prone to human error. Furthermore, there is a risk of damaging the fiber during the wiring work. Therefore, an efficient method for inspecting the fiber wiring and quality is required.

[0006] In a conventional Ethernet Switching Network (ESN), electrical processing is possible at intermediate switches, so faulty wiring can be detected by utilizing protocols such as Link Layer Discovery Protocol (LLDP) and ICMP, or existing technologies (Non-Patent Documents 1-3). However, optical switches process signals directly from light without converting them to electricity, so conventional mechanisms and technologies cannot be applied.

[0007] One possible method is to build a fiber switching network using optical switches equipped with power meters and send test signals to each fiber to easily identify faulty wiring locations. However, introducing optical switches equipped with power meters is not desirable because it increases costs, insertion loss, and power consumption.

[0008] One possible method is to connect a test terminal between two optical switches, send a test signal from one optical switch, and receive the test signal from the other optical switch via only one fiber to be tested, thereby testing each fiber connecting two optical switches one by one. However, when testing a fiber connecting two other optical switches, manual reconnection of the test terminal is required, which makes it impossible to avoid human error. Furthermore, testing each fiber one by one takes a very long time. When testing two or more fibers at once, it may not be possible to determine the condition of each fiber.

[0009] The present disclosure has been made in view of the above, and aims to efficiently and correctly inspect the status of all fibers in a fiber switching network.

[0010] A fiber inspection device according to one aspect of the present disclosure establishes, for a fiber group connecting a pair of switches, a first inspection path that includes one uplink and one downlink from the fiber group as fibers to be inspected, inputs test light into the first inspection path, measures a first received power of the test light that has passed through the first inspection path, and classifies the uplink and the downlink as fibers with an unspecified status if the first received power is less than a threshold value, establishes, for each of the fibers with an unspecified status, a second inspection path that includes only the single fiber with an unspecified status as a fiber to be inspected, inputs test light into the second inspection path, measures a second received power of the test light that has passed through the second inspection path, and classifies the status of the fibers with an unspecified status based on the second received power.

[0011] According to the present disclosure, the status of all fibers in a fiber switching network can be inspected efficiently and accurately.

[0012] FIG. 1 is a diagram illustrating an example of a fiber switching network. FIG. 2 is a diagram illustrating an example of an inspection path. FIG. 3 is a diagram illustrating an example of a path between transceivers. FIG. 4 is a diagram illustrating an example of the configuration of a fiber inspection device. FIG. 5 is a flowchart illustrating an example of the processing flow of the fiber inspection device. FIG. 6 is a flowchart illustrating an example of the processing flow for inspecting a fiber between a leaf switch and a spine switch. FIG. 7 is a flowchart illustrating an example of the processing flow for inspecting a fiber whose status is unknown. FIG. 8 is a diagram illustrating an example of fiber inspection. FIG. 9 is a diagram illustrating an example of fiber inspection. FIG. 10 is a diagram illustrating an example of fiber inspection. FIG. 11 is a diagram illustrating an example of fiber inspection. FIG. 12 is a diagram illustrating an example of fiber inspection. FIG. 13 is a diagram illustrating an example of fiber inspection. FIG. 14 is a diagram illustrating an example of fiber inspection. FIG. 15 is a diagram illustrating an example of fiber inspection. FIG. 16 is a diagram illustrating an example of fiber inspection. FIG. 17 is a diagram illustrating an example of the relationship between the number of quality-degraded fibers and abnormal fibers and the inspection time. FIG. 18 is a diagram illustrating an example of the relationship between the time required to replace a test terminal and the inspection time. FIG. 19 is a diagram illustrating an example of the relationship between probing time and inspection time. FIG. 20 is a diagram showing an example of the relationship between the number of fibers between a switch pair and inspection time. FIG. 21 is a diagram showing an example of the relationship between the ratio of degraded fibers and abnormal fibers and inspection time. FIG. 22 is a diagram showing an example of the number of fibers that resulted in false positives when the minimum loss of each fiber in the fiber length and the dispersion associated with the fiber loss were changed. FIG. 23 is a diagram showing an example of the number of fibers that resulted in false positives when the transceiver budget, the maximum loss of the optical switch, and the dispersion associated with the fiber loss were changed. FIG. 24 is a diagram showing an example of the number of fibers that resulted in false positives when the LC connector loss and the dispersion associated with the fiber loss were changed. FIG. 25 is a diagram showing an example of the number of fibers that resulted in false positives when the propagation loss coefficient and the dispersion associated with the fiber loss were changed. FIG. 26 is a diagram showing an example of the hardware configuration of a fiber inspection device.

[0013] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0014] An example of a fiber switching network to be inspected by the fiber inspection device of this embodiment is shown in Fig. 1. The configuration (topology) of the fiber switching network to be inspected is TF-CLOS.

[0015] A fiber switching network has k leaf switches I = {1,...,k} and m spine switches J = {k+1,...,k+m}. Leaf switches provide connections between end devices and networks within a data center. Spine switches are connected to all leaf switches within the data center by fibers e and forward traffic within the data center. Fibers that transmit data from leaf switches to spine switches are called uplinks, and fibers that transmit data from spine switches to leaf switches are called downlinks. The number of uplinks / downlinks connecting leaf switches and spine switches is v.

[0016] The set of all switches is represented as S = I∪J. The set of fibers connecting leaf switches and spine switches is represented as E. The fiber to be inspected is e∈E. Hereinafter, when there is no distinction between leaf switches and spine switches, they will simply be referred to as optical switches.

[0017] The physical wiring between switches that the operator envisions (wants to build) is called the envisioned topology. The envisioned topology describes the connection information between the ports of each optical switch. The operator builds a network based on the envisioned topology. The fiber inspection device inspects the envisioned topology to check for wiring errors and for abnormalities in the quality of each fiber (ensure that it can be used during operation).

[0018] It is assumed that all optical switches are normal, that there is no miswiring between the same optical switches or between optical switches of the same layer, and that there is no miswiring between Tx ports or Rx ports. Furthermore, in typical network construction work, only a small portion of fibers exhibit abnormalities in wiring or loss, and most fibers are correctly wired and have normal loss. It is assumed that more than half of the fibers are correctly wired and have normal loss.

[0019] During testing, a test terminal 50 is connected to a transmit / receive port of a leaf switch. In the example of FIG. 1, the test terminal 50 is connected to the leaf switch I1. The test terminal 50 inputs test light to a test path established in the fiber switching network, receives the test light that has passed through the test path, and measures the received optical power of the test light. The test path is the path along which the test light passes through the fiber switching network and is received by the test terminal 50. The test terminal 50 may be connected to any leaf switch. If a spine switch has an available port, the test terminal 50 may be attached to the spine switch. A terminal other than the test terminal 50 may input test light to the test path. The test terminal 50 does not need to be reconnected until all fibers to be tested have been tested.

[0020] [Signal Propagation Model] Next, a propagation model of the test signal will be described.

[0021] The test terminal 50 transmits test light to the fiber switching network and receives the test light that has passed through an inspection path q established in the fiber switching network. In the example of Fig. 2, the test light transmitted from the test terminal 50 passes through the leaf switch I1, fiber e1, spine switch J3, fiber e2, and leaf switch I1, and is received by the test terminal 50.

[0022] The loss on the test path q is due to the insertion loss of the optical switches I1 and J3, the propagation loss of the fibers e1 and e2, and the connectors and patch panels on the fibers e1 and e2. If the transmission power of the test light is P0, the received power P q is expressed by the following equation:

[0023]

[0024] Here, S q is the set of optical switches on the inspection path q, L s is the insertion loss of the optical switch s, E q is the set of fibers on the inspection path q, and L(e) is the loss in fiber e. The insertion loss L between each port of the optical switch s is provided by the vendor. The operator determines the insertion loss L between the ports of the optical switch. s L(e) includes propagation loss and connector / patch panel loss. The propagation loss of a typical single-mode fiber is 0.35 dB / km at a wavelength of 1310 nm. In a typical data center, the fiber length is about 500 m. Also, the LC connector loss is generally expressed as an average μ c =0.10dB, dispersion σ c =0.0049.

[0025] Each fiber e ∈ E on the inspection path q q The maximum possible loss L e,q max can be defined as follows:

[0026]

[0027] Here, L min is the minimum loss (including propagation loss and patch panel loss) that each fiber must have. q is the total loss incurred on inspection path q. min For example, if all fibers are accommodated in a patch panel, an average loss of 0.20 dB occurs, so the average value is halved to obtain L min =0.10dB.

[0028] The loss of fiber e on inspection path q is L e,q max The fiber e is connected to multiple inspection paths q1, q2,...,q n When inspected with, the maximum possible loss of fiber e is L e max is expressed by the following equation:

[0029]

[0030] As shown in FIG. 3, when a path is established between the transceivers 60A and 60B during operation, the loss on the path is within the budget P b If the transceiver budget is not less than q, communication between the transceivers 60A and 60B cannot be guaranteed. The transceiver budget is the optical signal loss that can be tolerated for the transceiver to function properly. Therefore, the loss on any path q is less than the transceiver budget P b The following equation must be satisfied:

[0031]

[0032] As shown in FIG. 3, the number of optical switches |S on an arbitrary path q established between transceivers 60A and 60B is q | is 3, the number of fibers |E q | is 2. The maximum loss value among all the optical switches is L s max Then, the maximum loss value L of the fiber e for always being able to communicate between the transceivers 60A and 60B is e max must satisfy the formula.

[0033]

[0034] The fiber inspection device of this embodiment inspects the state of all fibers and extracts a set of fibers that do not satisfy the above formula, that is, a set of fibers with degraded quality, miswired fibers, or broken fibers.

[0035] Here, if the total loss of all fibers on the test path q is L, the received optical power P at the test terminal 50 is q can be expressed as follows, where each fiber e∈E on the inspection path q q The loss L(e) will not exceed L.

[0036]

[0037] The allowable loss of all fibers on the path is L ok The allowable light receiving power P ok is expressed by the following equation:

[0038]

[0039] When the inspection path q is tested, the received light power P q is the allowable light receiving power P ok If it is greater than 1, then each fiber e∈E on the inspection path q q The loss L(e) is the allowable loss L ok Smaller than.

[0040]

[0041] The contrapositive of number 8 is expressed as follows:

[0042]

[0043] In other words, the loss L(e) is the allowable loss L ok If there is a fiber e of at least q is the allowable light receiving power P ok The following is the result.

[0044] If the inspection path q includes only one state-unspecified fiber e′ and all other fibers are normal fibers, then P q ≦P ok Then, L(e')≧L ok Therefore, the unspecified fiber can be considered as a deteriorated fiber. q >P ok Then, L(e') < L ok Therefore, the state-unspecified fiber can be regarded as a normal fiber. q <P ε (P ε is the sensitivity of the test terminal 50), that is, if the test terminal 50 cannot receive the test light, the state-unspecified fiber can be considered to be an abnormal fiber (miswiring or disconnection).

[0045] In this way, when the inspection path q includes only one fiber with an unspecified state and the other fibers are all normal fibers, the state of the fiber with an unspecified state is determined as follows: q can be uniquely determined based on

[0046] In addition, the allowable loss L ok is expressed as follows using the right side of Equation 5.

[0047]

[0048] The proposed method is designed to prevent false negatives (oversights). However, false positives (judging normal to be abnormal) may occur. A false positive occurs when any of the fibers e on the inspection path q satisfies the following equation:

[0049]

[0050] In other words, the loss L(e) of fiber e is the allowable loss L ok and the maximum loss L that the fiber e can have e max is the allowable loss L ok In the above cases, a false positive occurs.

[0051] 2, fiber e1 is a normal fiber that has been inspected, and fiber e2 is a fiber whose status is unknown. The total loss of fibers e1 and e2 on the inspection path q is 0.25 dB, and fiber e2 is a normal fiber with no loss due to damage or failure.

[0052] The loss of the optical switches I1 and J3 is assumed to be 0.8 dB. q From equation 1, P q =0.0-2.4-0.5=-2.9.

[0053] The maximum loss L that can be taken by the fibers e1 and e2 e1 max , L e2 max From equation 2, L e1 max =L e2 max = 0.0 - (-2.9) - 2.4 - 0.1 = 0.4. The minimum loss of each fiber, L min The actual loss of fiber e2 is 0.25 dB, but it is estimated to be 0.4 dB, which is a large amount.

[0054] The maximum loss value of the optical switch is L s maxis 1.4 dB, and the transceiver budget P b Let's say that the allowable loss L is 5 dB. ok From equation 10, L ok =(5-3*1.4) / 2=0.4. This does not satisfy the condition of equation 5, so fiber e2 is considered to be an abnormal fiber even though it is a normal fiber.

[0055] [Configuration of Fiber Inspection Device] An example of the configuration of the fiber inspection device 1 of this embodiment will be described with reference to Fig. 4. The fiber inspection device 1 shown in the figure includes a test terminal control unit 11, a switch control unit 12, an inspection unit 13, a path calculation unit 14, a path loss calculation unit 15, a topology information storage unit 16, and a retention unit 17.

[0056] The test terminal control unit 11 controls the test terminal 50 connected to the fiber switching network.

[0057] The switch control unit 12 controls the internal connection state of each optical switch included in the fiber switching network to establish an inspection path. The switch control unit 12 controls the optical switches so that ports not on the inspection path are not connected to any other ports.

[0058] In step 1, the inspection unit 13 classifies the fibers between the leaf switch to which the test terminal 50 is connected and each spine switch. After step 1, in step 2, the inspection unit 13 classifies the fibers between each leaf switch to which the test terminal 50 is not connected and each spine switch. In step 2, an inspection path that passes through the fiber classified as a normal fiber in step 1 is used. After step 2, the inspection unit 13 outputs information on quality-degraded fibers and abnormal fibers.

[0059] The inspection unit 13 measures the received optical power of the test light received by the test terminal 50 when test light is input to the inspection path, and classifies the fibers on the inspection path based on the received optical power. Specifically, the inspection unit 13 measures the received optical power of the fibers connecting the leaf switch and the spine switch along an inspection path including one uplink and one downlink. The inspection unit 13 classifies the uplinks and downlinks whose received optical power meets the standard as normal fibers, and the uplinks and downlinks whose received optical power does not meet the standard as fibers of unknown status. The inspection unit 13 then measures the received optical power of each fiber of unknown status along an inspection path including only one fiber of unknown status, and classifies the fibers of unknown status based on the received optical power. During the inspection, each fiber is classified into normal fiber, degraded fiber, and abnormal fiber. Normal fiber is a collection of fibers with no quality abnormalities and correct wiring. Degraded fiber is a collection of fibers with such high loss that they cannot be used for operation. Abnormal fiber is a collection of fibers that do not pass signals due to incorrect wiring or a break.

[0060] The route calculation unit 14 calculates a test path along which the test light from the test terminal 50 passes through the fiber switching network and returns to the test terminal 50, based on the assumed topology provided by the operator.

[0061] The path loss calculation unit 15 calculates the transmission power P of the test light, the received power P of the test light after passing through the test path q, q , the insertion loss of each optical switch, and the minimum loss of each fiber L min Based on this, the loss of the fiber on the test path is determined.

[0062] The topology information storage unit 16 stores the assumed physical wiring status given by the operator.

[0063] The storage unit 17 stores information necessary for testing the fibers of the fiber switching network. For example, the storage unit 17 stores the transmission power P0 and sensitivity P1 of the test terminal 50. ε , propagation loss coefficient α of the test light, fiber length d(e), insertion loss L for each port pair of the optical switchs , the time required for the control required for switching the optical switch T t , the budget P of the transceiver used during operation b , and the minimum loss of each fiber L min etc.

[0064] [Operation of Fiber Inspection Apparatus] An example of the processing flow of the fiber inspection apparatus 1 will be described with reference to the flowchart of FIG.

[0065] In step S1 (procedure 1), the fiber inspection device 1 inspects the fibers connecting the leaf switch to which the test terminal 50 is connected and each spine switch. For example, in the example of the fiber switching network of Fig. 1, in procedure 1, the fibers connecting the leaf switch I to which the test terminal 50 is connected and the spine switches J and J are inspected.

[0066] Specifically, the fiber inspection device 1 limits the number of fibers included in the inspection path to one uplink and one downlink, a total of two, and sequentially inspects pairs of uplinks and downlinks connecting the leaf switch and the spine switch. The fiber inspection device 1 classifies pairs of uplinks and downlinks whose received optical power at the test terminal 50 satisfies a threshold as normal fibers, and classifies pairs of uplinks and downlinks whose received optical power does not satisfy the threshold as status-unspecified fibers.

[0067] Next, the fiber inspection device 1 establishes an inspection path including only one fiber whose status is unspecified for each fiber whose status is unspecified, and inspects the status of each fiber whose status is unspecified one by one.

[0068] In step S2 (procedure 2), the fiber inspection device 1 establishes an inspection path including the fiber classified as a normal fiber in procedure 1, and inspects the fibers connecting each leaf switch and each spine switch to which the test terminal 50 is not connected. For example, in the example of the fiber switching network in FIG. 1, in procedure 2, the fibers connecting between the leaf switch I2 and the spine switches J4 and J5, and between the leaf switch I3 and the spine switches J4 and J5 are inspected.

[0069] Specifically, in procedure 2, as in procedure 1, the fiber inspection device 1 limits the fibers to be inspected included in the inspection path to a total of two, one uplink and one downlink between the leaf switch to be inspected and the spine switch, and sequentially inspects the pairs of uplink and downlink connecting the leaf switch to be inspected and the spine switch. The fiber inspection device 1 classifies pairs of uplink and downlink whose received optical power at the test terminal 50 satisfies a threshold as normal fibers, and classifies pairs of uplink and downlink whose received optical power does not satisfy the threshold as fibers of unknown status.

[0070] Next, similarly to step 1, the fiber inspection device 1 establishes an inspection path including only one fiber with an unspecified status for each fiber with an unspecified status, and inspects the status of each fiber one by one in turn.

[0071] By performing steps 1 and 2, the state of the fiber group between the optical switches can be identified for all combinations of leaf switches and spine switches. The fiber inspection device 1 outputs the degraded fiber and abnormal fiber.

[0072] Consider creating an arbitrary test path on an assumed topology. If the number of test uplinks and test downlinks on the test path is a maximum of one each, paths other than the test path considered on the assumed topology will not be created due to faulty wiring or other reasons. In other words, a path is generated according to the assumed topology. Conversely, considering the possibility of faulty wiring, if the number of test uplinks and test downlinks is greater than one, the fiber condition cannot be uniquely identified. Therefore, it is necessary to estimate the fiber condition using only test paths that contain a maximum of one test uplink and one test downlink. Since the test paths generated in steps 1 and 2 above satisfy the above, if the received optical power at the test terminal is as expected, it can be guaranteed that there are no faulty wiring in any of the fibers on the test path and that there are no quality issues.

[0073] An example of inspection processing of a fiber group between a leaf switch and a spine switch by the fiber inspection device 1 will be described with reference to the flowcharts in Figures 6 and 7. The processing in Figures 6 and 7 is performed when inspecting a fiber group between a leaf switch and a spine switch in each of steps 1 and 2. In the following description, it is assumed that inspection is performed on a fiber group connecting a leaf switch i and a spine switch j.

[0074] In the process of FIG. 6, the fiber inspection device 1 determines a set E of uplinks from the leaf switch i to the spine switch j. ij and a set E of downlinks from spine switch j to leaf switch i. ji and input the uplink / downlink pair (e u , e d ∈E ij , E ji ) is examined one by one. For example, in step 1, the set E of fibers connecting the leaf switch 1 to which the test terminal is connected and each spine switch j∈{k+1,...,k+m} is examined. 1j , E j1 In step 2, a set of fibers E connecting each leaf switch i∈{2,...,k} except for leaf switch 1 and each spine switch j∈{k+1,...,k+m} is input. ij , E ji Enter.

[0075] In step S11, the fiber inspection device 1 controls each optical switch to inspect the fiber e u , e d The test terminal 50 inputs test light to the test path q, receives the test light that has passed through the test path q, and outputs a received power P q In step 1, the inspection path q is the fiber e to be inspected. u , e d In step 2, the test path q passes through the leaf switch 1 to which the test terminal 50 is connected, and does not include any fibers classified as normal fibers in step 1 and the test target fiber e. u , e d Includes.

[0076] In step S12, the fiber inspection device 1 detects the received light power P q is the allowable light receiving power P ok Determine whether it is greater than or equal to the value.

[0077] P q >P ok In this case, in step S13, the fiber inspection device 1 u , e d is classified as a normal fiber. q >P ok In the case of L(e u )<L ok , L(e d )<L ok is.

[0078] P q ≦P ok In this case, in step S14, the fiber inspection device 1 u , e d At this point, the fiber e u , e d It is unclear which of these is normal or abnormal.

[0079] The above process is executed for each pair of uplink / downlink connecting leaf switch i and spine switch j.

[0080] 6 is completed, all fibers connecting leaf switch i and spine switch j are classified as either normal fibers or fibers with an unspecified status. After the processing of FIG. 6, the fiber inspection device 1 proceeds to the processing of FIG. 7.

[0081] In the process of FIG. 7, the fiber inspection device 1 identifies the state of each of the fibers that are classified as state-unspecified fibers among the fibers connecting the leaf switch i and the spine switch j.

[0082] In step S21, the fiber inspection device 1 selects one fiber e to be inspected from the fibers whose status is not specified and one fiber in the opposite direction from the normal fibers. For example, if fiber e is an uplink, a downlink fiber classified as a normal fiber is selected, and if fiber e is a downlink, an uplink fiber classified as a normal fiber is selected.

[0083] In step S22, the fiber inspection device 1 controls each optical switch to establish an inspection path q including fiber e. The inspection path q includes one fiber e whose status is not specified and one or more normal fibers. The test terminal 50 inputs test light to the inspection path q, receives the test light that has passed through the inspection path q, and calculates the received optical power P of the test light. q Measure.

[0084] In step S23, the fiber inspection device 1 detects the received light power P q is the allowable light receiving power P ok Determine whether it is greater than or equal to the value.

[0085] P q >P ok In this case, in step S24, the fiber inspection device 1 classifies the fiber e as a normal fiber. q >P ok In the case of L(e)<L ok is.

[0086] P q ≦P ok In this case, in step S25, the fiber inspection device 1 determines whether or not the test light has been received. ε Then, P q <P ε In this case, the test terminal 50 cannot receive the test light.

[0087] P q ≧P ε In this case, in step S26, the fiber inspection device 1 classifies the fiber e as a quality-degraded fiber.

[0088] P q <P εIn this case, in step S27, the fiber inspection device 1 classifies the fiber e as an abnormal fiber.

[0089] The above process is executed for each state-unspecified fiber between leaf switch i and spine switch j.

[0090] When the process of FIG. 7 is completed, all fibers connecting leaf switch i and spine switch j are classified as normal fibers, degraded fibers, or abnormal fibers.

[0091] [Example] An example of fiber inspection using the fiber inspection device 1 will be described with reference to Figs.

[0092] Figure 8 shows the fiber switching network to be tested. The fiber switching network shown in the figure includes leaf switches I1 and I2 and spine switches J3 and J4. Leaf switch I1 is connected to spine switch J3 by four fibers e1-e4. Leaf switch I1 is connected to spine switch J4 by two fibers e5 and e6. Leaf switch I2 is connected to spine switch J3 by two fibers e7 and e8. Leaf switch I2 is connected to spine switch J4 by four fibers e9-e12. A test terminal 50 is attached to leaf switch I1.

[0093] Assume that fiber e4 connecting leaf switch I1 and spine switch J3 is a deteriorated fiber.

[0094] In step 1, the fiber inspection device 1 inspects fibers e1-e4 connecting the leaf switch I1 and the spine switch J3, and fibers e5 and e6 connecting the leaf switch I1 and the spine switch J4. After inspecting the fibers for each uplink and downlink pair using the process in Fig. 6, the fiber inspection device 1 inspects each of the fibers classified as status-unspecified fibers using the process in Fig. 7.

[0095] First, as shown in Fig. 9, the fiber inspection device 1 inspects the fibers e1 and e2, which are a pair of uplink and downlink connecting the leaf switch I1 and the spine switch J3. Since both the fibers e1 and e2 are normal, the test terminal 50 can receive test light with sufficient intensity. q >P ok and L(e1)<L ok , L(e2)<L ok The fiber inspection device 1 classifies the fibers e1 and e2 as normal fibers.

[0096] 10, the fiber inspection device 1 inspects another uplink and downlink pair of fibers e3 and e4 connecting the leaf switch I1 and the spine switch J3. Because the fiber e4 is deteriorated, the test terminal 50 receives test light with a lower intensity than expected. q ≦P ok At this point, it is not possible to distinguish whether fiber e3 or fiber e4 is deteriorated. The fiber inspection device 1 classifies fibers e3 and e4 as fibers whose condition is not specified.

[0097] 11, the fiber inspection device 1 inspects the fibers e5 and e6, which are a pair of uplink and downlink connecting the leaf switch I1 and the spine switch J4. Since both fibers e5 and e6 are normal, the test terminal 50 can receive test light with sufficient intensity. q >P ok and L(e5) < L ok , L(e6)<L ok The fiber inspection device 1 classifies the fibers e5 and e6 as normal fibers.

[0098] 6 is completed, the fiber inspection device 1 identifies the states of fibers e3 and e4, which are fibers with an unspecified state, by the process of Fig. 7. To identify the fibers with an unspecified state, fibers e3 and e4 are inspected in combination with normal fibers e1 and e2.

[0099] 12, the fiber inspection device 1 inspects the downlink fiber e3 paired with the uplink fiber e2. Since the fiber e3 is normal, the test terminal 50 can receive the test light with sufficient intensity. q >P ok and L(e3)<L ok The fiber inspection device 1 classifies the fiber e3 as a normal fiber.

[0100] 13, the fiber inspection device 1 inspects the uplink fiber e4 paired with the downlink fiber e1. Since the fiber e4 is deteriorated, the test terminal 50 receives test light with a lower intensity than expected. q ≦P ok and L(e4)≧L ok Since fiber e1 is known to be normal, fiber inspection device 1 classifies fiber e4 as a degraded fiber.

[0101] This is the end of the processing in step 1, and inspection of all fibers connecting the leaf switch I1 and the spine switches J3 and J4 is completed.

[0102] Next, in procedure 2, the fiber inspection device 1 inspects fibers e7 and e8 connecting the leaf switch I2 and the spine switch J3, and fibers e9-e12 connecting the leaf switch I2 and the spine switch J4. In procedure 2, fibers e1, e2, e5, and e6 classified as normal fibers in procedure 1 are used to inspect fibers e7-e12 connecting the leaf switch I2 and the spine switches J3 and J4, respectively.

[0103] First, as shown in Fig. 14, the fiber inspection device 1 establishes an inspection path via fibers e1 and e2, and inspects fibers e7 and e8, which are a pair of uplink and downlink connecting the leaf switch I2 and the spine switch J3. Since both fibers e7 and e8 are normal, the test terminal 50 can receive test light with sufficient intensity. q >P okand L(e7)<L ok , L(e8)<L ok The fiber inspection device 1 classifies the fibers e7 and e8 as normal fibers.

[0104] 15, the fiber inspection device 1 establishes an inspection path via fibers e5 and e6, and inspects fibers e9 and e10, which are a pair of uplink and downlink connecting the leaf switch I2 and the spine switch J4. Since both fibers e9 and e10 are normal, the test terminal 50 can receive test light with sufficient intensity. q >P ok and L(e9)<L ok , L(e10)<L ok The fiber inspection device 1 classifies the fibers e9 and e10 as normal fibers.

[0105] 16, the fiber inspection device 1 establishes an inspection path via fibers e5 and e6, and inspects fibers e11 and e12, which are a pair of uplink and downlink connecting the leaf switch I2 and the spine switch J4. Since both fibers e11 and e12 are normal, the test terminal 50 can receive test light with sufficient intensity. q >P ok and L(e11)<L ok , L(e12)<L ok The fiber inspection device 1 classifies the fibers e11 and e12 as good fibers.

[0106] In procedure 2, since there are no fibers with unspecified states, the process of Fig. 7 is not executed. If there are fibers with unspecified states, the fibers with unspecified states are inspected one by one by the process of Fig. 7, as in procedure 1.

[0107] The fiber inspection device 1 outputs the inspection result that the fiber e4 is a quality-degraded fiber and there is no abnormal fiber.

[0108] [Evaluation of Testing Time] We compared the testing time between the conventional method and the proposed method. The conventional method involves attaching a test terminal to the leaf switch and spine switch, and testing each fiber connected between the leaf switch and spine switch. When changing the pair of leaf switch and spine switch, it is necessary to reconnect the test terminal.

[0109] The total test time required for the conventional method can be calculated as follows: The number of switch pairs, which are a combination of k leaf switches and m spine switches, is km. Since each switch pair is connected by v fibers, the number of tests is kmv. The time required to install the test terminal is T t , the time required for one test (probe time) is T p Then, the total inspection time is kmT. t +kmvT p This becomes:

[0110] The total inspection time required for the proposed method can be calculated as follows. If the average number of inspections required to identify the state of an unidentified fiber in the processes of Figures 6 and 7 is |U|, then the number of inspections for a switch pair is v + |U|. Since the number of switch pairs is km, the total number of inspections is km(v + |U|). Assuming that the number of quality-degraded or abnormal fibers is sufficiently small and |U| << v, the number of inspections can be approximated as kmv. The total inspection time is T t +kmvT p This becomes:

[0111] Simulations were performed to determine the inspection times for the conventional and proposed methods. The various parameters used in the simulations were set as shown in the following table, taking into account the actual device.

[0112]

[0113] Below, we show the results of comparing the inspection time between the proposed method (Ours) and the conventional method (Baseline) while changing the above parameters. Each figure shows the simulation results (a) for a fiber switching network (a) where (k, m, v) = (24, 24, 16) assuming a large-scale DCN, and the simulation results (b) for a fiber switching network (b) where (k, m, v) = (32, 16, 16) assuming an AI cluster. In the fiber switching network (a), MEMS is assumed for switching the optical switches, and the probe time T p The fiber switching network (b) assumes that the optical switch uses a Robot OCS (ROME max-t) for switching, and the probe time is T p The time was set to 30s and the insertion loss range was set to [0.5, 1.0].

[0114] Figure 17 shows the relationship between the number of degraded and abnormal fibers and the inspection time. The horizontal axis represents the number of degraded and abnormal fibers, and the vertical axis represents the inspection time. With the proposed method, the inspection time depends on the number of degraded and abnormal fibers, and this is its weak point, so the inspection time was calculated by changing the number of degraded and abnormal fibers. The proposed method can complete the inspection at least 1.5 times faster than the conventional method.

[0115] FIG. 18 shows the time T t The horizontal axis shows the time required to change the test terminal. t In the fiber switching network (a), T t When the time is 600 seconds, the proposed method completes the inspection in 23 hours, while the conventional method takes 217 hours (more than a week). The proposed method is approximately 13.9 times faster than the conventional method.

[0116] In FIG. 19, the probe time T p The horizontal axis shows the relationship between the probe time T pThe vertical axis shows the time required for each probe, and the vertical axis shows the inspection time. The time required for one probe depends on the control / switching time of the optical switch. The proposed method completes the inspection more than 1.8 times faster than the conventional method.

[0117] Figure 20 shows the relationship between the number of fibers v between switch pairs and the inspection time. The horizontal axis represents the number of fibers v, and the vertical axis represents the inspection time. This is consistent with the results of theoretical analysis, and the inspection time increases as the number of fibers v increases. The proposed method is able to complete the inspection in a shorter time than the conventional method.

[0118] Fig. 21 shows the ratio r of quality-degraded fibers and abnormal fibers. m The horizontal axis shows the ratio of quality-degraded fiber to abnormal fiber, r m The vertical axis shows the test time. m The inspection time did not change even when the time was changed. The proposed method was able to complete the inspection in a shorter time than the conventional method.

[0119] In this way, we confirmed that the proposed method can significantly improve inspection efficiency compared to conventional methods. While the existing method can shorten inspection time by having multiple people work in parallel, it requires multiple people and multiple devices, which incurs costs in terms of labor and equipment. The proposed method can automatically perform inspections using a single test terminal 50, so it is advantageous not only in terms of time reduction but also in terms of cost.

[0120] [Evaluation of accuracy] The accuracy of the proposed method was evaluated by the number of fibers that resulted in false positives.

[0121] In the accuracy evaluation, we performed a simulation on a fiber switching network with (k, m, v) = (24, 24, 16). With the settings in Table 1 above, no false positives occurred.

[0122] Below, we show the results of evaluating the number of fibers that resulted in false positives while changing the parameters.

[0123] Figure 22 shows the fiber length d(e) and the minimum loss L min , and the dispersion associated with the fiber loss σ cThe results of evaluating the number of fibers that became false positives by changing d(e) are shown below. For three fiber lengths d(e) = 1.0 km, 1.5 km, and 2.0 km, the longer the fiber, the lower the minimum loss L min Two L per fiber length are used so that min Set the LC connector loss μ c =0.10dB, dispersion σ c Simulations were performed for 0.005, 0.0075, and 0.010. Note that the transceiver budget P b = 7.0 dB, the maximum loss value of the optical switch L s max =1.4dB.

[0124] Generally, the LC connector loss μ c =0.10dB, dispersion σ c =0.0049, and it was confirmed that false positives occurred only under conditions significantly exceeding these. The maximum number of fibers that became false positives was 60 or less. min If we can grasp this more accurately, false positives will not occur. For example, at d(e) = 2.0 km, there is a loss of 0.90 dB due to the propagation loss and two patch panels, but L min It has been confirmed that no false positives occur at values ​​of 0.74 dB or higher.

[0125] FIG. 23 shows the budget P b , the maximum loss value of the optical switch L s max , and the dispersion associated with the fiber loss σ c The results of evaluating the number of fibers that became false positives by changing the budget P are shown below. b = 5dB, 6dB, 7dB, budget P b The maximum loss value L s max Adjust the LC connector loss μ c =0.10dB, dispersion σ c Simulations were performed for =0.005, 0.0075, and 0.010.

[0126] Generally, the LC connector loss μ c =0.10dB, dispersion σ c = 0.0049, and it was confirmed that false positives occurred only under conditions significantly exceeding these limits. The maximum number of fibers that became false positives was 35 or less.

[0127] FIG. 24 shows the LC connector loss μ c and variance σ c The results of evaluating the number of fibers that became false positives by changing the LC connector loss μ c For =0.122dB, 0.121dB, and 0.120dB, variance σ c Simulations were performed with ρ = 0.005, 0.0075, and 0.010.

[0128] Generally, the LC connector loss μ c =0.10dB, dispersion σ c = 0.0049, and it was confirmed that false positives occurred only under conditions significantly exceeding these limits. The maximum number of fibers that became false positives was 30 or less.

[0129] Figure 25 shows the propagation loss coefficient α and dispersion σ c The results of evaluating the number of fibers that became false positives by changing the propagation loss coefficient α = 0.53, 0.52, 0.51, 0.50, and the dispersion σ c Simulations were performed with ρ = 0.005, 0.0075, and 0.010.

[0130] It was confirmed that false positives occur when the propagation loss coefficient α is greater than 0.50. Generally, the propagation loss coefficient α is less than 0.35 to 0.40, and the variance σ c = 0.0049, and it was confirmed that false positives occurred only under conditions significantly exceeding these limits. The maximum number of fibers that became false positives was 25 or less.

[0131] Thus, the LC connector loss μ c>0.12dB, dispersion σ c False positives occur when the propagation loss coefficient α is greater than 0.50 and the LC connector loss μ c =0.10dB, dispersion σ c =0.0049, and the propagation loss coefficient α≦0.35, and no false positives occurred under these conditions.

[0132] As described above, the fiber inspection device 1 of this embodiment establishes an inspection path q including one uplink and one downlink from a fiber group connecting a leaf switch and a spine switch as fibers to be inspected, inputs test light to the inspection path q, and measures the received power P of the test light that has passed through the inspection path q. q Measure the received light power P q is the allowable light receiving power P ok If the value does not satisfy the condition, the uplink and downlink are classified as fibers with an unspecified state. The fiber inspection device 1 establishes an inspection path q including only one fiber with an unspecified state as an inspection target fiber for each fiber with an unspecified state, inputs test light to the inspection path q, and measures the received power P of the test light that has passed through the inspection path q. q Measure the received light power P q The state of the unspecified fiber is classified based on the

[0133] The above-described fiber inspection device 1 can be implemented, for example, as a general-purpose computer system including a central processing unit (CPU) 901, a memory 902, a storage 903, a communication device 904, an input device 905, and an output device 906, as shown in Fig. 26. In this computer system, the fiber inspection device 1 is realized by the CPU 901 executing a predetermined program loaded onto the memory 902. This program can be recorded on a computer-readable non-transitory recording medium such as a magnetic disk, an optical disk, or a semiconductor memory, or can be distributed via a network.

[0134] REFERENCE SIGNS LIST 1 Fiber inspection device 11 Test terminal control unit 12 Switch control unit 13 Inspection unit 14 Path calculation unit 15 Path loss calculation unit 16 Topology information storage unit 17 Holding unit 50 Test terminal 60A, 60B Transceiver

Claims

1. A fiber inspection device, which, for a fiber group connecting a pair of switches, establishes a first inspection path that includes one uplink and one downlink from the fiber group as fibers to be inspected, inputs test light into the first inspection path, measures a first received power of the test light that has passed through the first inspection path, and classifies the uplink and downlink as fibers with an unspecified status if the first received power is less than a threshold value, and establishes a second inspection path that includes only one of the fibers with an unspecified status as fibers to be inspected, inputs test light into the second inspection path, measures a second received power of the test light that has passed through the second inspection path, and classifies the status of the fibers with an unspecified status based on the second received power.

2. A fiber inspection device as claimed in claim 1, which inspects the condition of the fiber in each fiber group connecting each switch in a first switch group and each switch in a second switch group, comprising: connecting a test terminal that transmits and receives the test light to one of the switches in the first switch group; classifying, as a first step, the condition of the fiber connecting between a first switch in the first switch group connected to the test terminal and each switch in the second switch group; classifying, as a second step, the condition of the fiber connecting between each switch in the first switch group other than the first switch and each switch in the second switch group; and establishing, in the second step, the first inspection path and the second inspection path including the fiber classified as normal in the first step.

3. A fiber inspection device according to claim 1, wherein the threshold value is set based on the budget of a transceiver used during operation and the insertion loss of the switch.

4. A fiber inspection method using a fiber inspection device, comprising the steps of: establishing a first inspection path for a fiber group connecting a pair of switches, the first inspection path including one uplink and one downlink from the fiber group as fibers to be inspected; inputting test light into the first inspection path; measuring a first received power of the test light that has passed through the first inspection path; and classifying the uplink and downlink as fibers of an unspecified state if the first received power is less than a threshold value; establishing a second inspection path for each of the fibers of an unspecified state, the second inspection path including only the single fiber of an unspecified state as a fiber to be inspected; inputting test light into the second inspection path; measuring a second received power of the test light that has passed through the second inspection path; and classifying the state of the fibers of an unspecified state based on the second received power.

Images