JP5485740B2 - Failure detection device - Google Patents

Description

  The present invention relates to a failure detection device that detects a failure of a network device.

  Among network devices belonging to various communication networks such as a wired network and a wireless network, those having a monitoring function for detecting a failure of the own device or other network devices in the network are known. This network device detects a failure of its own device or another network device at an early stage by a monitoring function such as polling, and repairs the detected failure at an early stage, thereby preventing deterioration in service quality.

  By the way, the failure of the network device includes not only a hardware failure but also a failure in a part of processing due to a bug in software or firmware. In this case, although the monitoring function as described above can detect a hardware failure, it may be difficult to detect the failure by software or the like. Specifically, in a failure detection target network device, if a failure occurs only in traffic processing due to a software bug, the failure may be detected in order to react normally to a monitoring function such as polling. It was difficult.

  Conventionally, failure detection apparatuses that detect failures by various methods including such failures have been studied.

  For example, a failure is detected by setting a threshold for the number of incoming / outgoing calls, the number of successful calls in call connection processing, the success rate, etc. (Patent Literature 1 and Patent Literature 2), and a failure is detected using a frame error rate Devices (Patent Document 3), devices that fail a device that does not accept traffic for a certain period of time (Patent Document 4), and the like have been studied.

  In addition, in a wireless communication system, a failure detection device that detects a failure by using electric field strength received by a terminal device is also being studied. For example, a device that detects a failure based on information used for transmission power control (Patent Document 5), a device that detects a failure based on a change in electric field strength in an area (Patent Document 6), and an electric field from a plurality of areas A technique for detecting a failure by creating a database of strength (Patent Document 7) and the like has been studied.

  Further, a method of performing a detection process periodically by arranging a terminal device in a network device has been studied (Patent Document 8).

JP 2008-227618 A JP 2000-253148 A Japanese Patent Laid-Open No. 08-274709 Japanese Patent Application Laid-Open No. 07-046658 JP 2008-048160 A JP 11-163784 A JP-A-11-146443 Japanese Patent Laid-Open No. 05-103368

  However, the failure detection method according to the above prior art has the following problems.

  Specifically, as in Patent Documents 5 to 7, it is difficult to detect a failure that does not affect the electric field strength in the failure detection method based on the electric field strength. Also, as disclosed in Patent Document 8, it is difficult to install a terminal device for detection processing near a network device due to arrangement conditions or the like, and the terminal device cannot be installed in all network devices. .

  In addition, as disclosed in Patent Documents 1 to 4, in the failure detection method based on traffic information such as the number of successful calls, the traffic greatly fluctuates depending on the time zone and the operation configuration of the network device. It may not be detected properly. In particular, when traffic information fluctuates with time in accordance with the wireless environment as in a wireless network, there is a problem in that the accuracy of detecting a failure of the network device is reduced.

  The present invention has been made in view of such a point, and an object of the present invention is to provide a failure detection device capable of improving the failure detection accuracy of a network device when the traffic information fluctuates greatly in time. And

The failure detection device of the present invention is a failure detection device for detecting a failure of a network device, and among the traffic information generated when communicating using the network device, two pieces of traffic information having a statistical linear relationship, Alternatively, a traffic information acquisition unit that acquires two pieces of traffic information in which conversion information by a conversion function has a statistical linear relationship, an error calculation unit that calculates an error of a statistical linear relational expression indicating the statistical linear relationship, and the error calculation unit A failure detection unit that detects a failure of the network device, and a learning unit that learns the two traffic information following time variation, and the error calculation unit includes: The slope of the statistical linear relational expression is calculated based on the two traffic information learned by the learning unit, and has the calculated slope And calculating the error of the statistical linear relations for linear relations.

  According to this configuration, traffic is greatly fluctuated in time by detecting a failure of the network device based on an error of a statistical linear relational expression indicating a statistical linear relation of traffic information or traffic conversion information. In addition, it is possible to improve the detection accuracy of the failure of the network device.

  ADVANTAGE OF THE INVENTION According to this invention, when traffic information fluctuates large temporally, the failure detection apparatus which can improve the detection accuracy of the failure of a network apparatus can be provided.

1 is a schematic configuration diagram of a radio communication system according to a first embodiment of the present invention. It is a functional lineblock diagram of a failure detection device concerning a 1st embodiment of the present invention. It is a flowchart which shows operation | movement of the failure detection apparatus which concerns on the 1st Embodiment of this invention. It is a figure for demonstrating the periodic steadiness of the traffic information which concerns on the 2nd Embodiment of this invention. It is a flowchart which shows operation | movement of the failure detection apparatus which concerns on the 2nd Embodiment of this invention.

  Hereinafter, embodiments of the present invention will be described. In the following embodiment, an example in which the failure detection device according to the present invention is arranged in a wireless communication system including a wireless section between a wireless terminal device and a wireless base station will be described. However, the present invention is not limited to this configuration. It is not something. Any communication system can be used as long as the failure detection device can be arranged. For example, a wireless section such as the Internet or a public switched telephone network (PSTN) to which a fixed terminal device such as a personal computer or a fixed telephone is connected. It is good also as a structure which arrange | positions a failure detection apparatus in the wired communication system which does not contain.

[First Embodiment]
<Overall schematic configuration of wireless communication system>
FIG. 1 is a configuration diagram of a wireless communication system according to the first embodiment. As shown in FIG. 1, the wireless communication system includes a wireless terminal device 10 such as a mobile phone terminal, a wireless base station device 20 to which the wireless terminal device 10 is wirelessly connected, a relay device 30 such as a router device and a switch device, and a wireless base station. The server device 40 and the failure detection device 50 communicate with the wireless terminal device 10 via the device 20 and the relay device 30.

  The failure detection device 50 is a network device in the wireless communication system, that is, the wireless base station device 20, the relay device 30, the server device 40, and the wireless base station device 20 and the relay device 30 that are not shown. Traffic information such as a radio network controller (RNC) is monitored, and a failure in each network device is detected based on the traffic information.

  Although the failure detection device 50 is connected as an external device of each network device, it may be provided inside as a part of each network device. In addition, the failure detection device 50 may be configured to detect all failures in each network device (that is, the wireless base station device 20, the relay device 30, the server device 40, a wireless control device (not shown), etc.) It may be configured to detect a failure of only the network device.

  Moreover, although the failure detection apparatus 50 demonstrates the example comprised so that traffic information may be acquired from each network apparatus in a radio | wireless communications system itself, it is not limited to this structure. For example, a traffic information collection device (not shown) that collects traffic information of each network device is separately provided in the wireless communication system, and the failure detection device 50 is configured to acquire the traffic information collected by the traffic information collection device. May be.

<Fault detection principle>
Next, the failure detection principle according to the first embodiment will be described. In the failure detection apparatus 50 according to the first embodiment, two pieces of traffic information x, y having a statistical linear relationship among traffic information generated when communication is performed using a network device that is a failure detection target, or The failure of the network device is detected using the two pieces of traffic information x and y in which the traffic conversion information f (x) and g (y) based on the conversion functions f and g have a statistical linear relationship.

Here, the two pieces of traffic information x and y having the statistical linear relationship are the traffic information x and y having the relationship of Expression (1).

The traffic conversion information f (x) and g (y) is obtained by converting the traffic information x and y using the conversion functions f and g. The two pieces of traffic information x and y in which the traffic conversion information f (x) and g (y) have a statistical linear relationship are the traffic information x and y in the relationship of Expression (2).

  In equations (1) and (2), a represents the slope of the linear relational expression, b represents a constant, and e represents an error. The slope a and the constant b fluctuate only to such an extent that the value does not fluctuate in a short time interval, but fluctuate with time in a long time interval. The error e also varies with time. Thus, the linear relationship in which the slope a, the constant b, and the error e fluctuate following the time variation of the traffic information x and y is called a statistical linear relationship, and the statistical linear relationship is expressed by equations (1) and (2). The line format shown is called a statistical linear relational expression. The error e indicates an error of a statistical linear relational expression with respect to a linear relational expression having a slope a, and is used for detecting a failure.

  Examples of the traffic information x and y satisfying the statistical linear relational expressions shown in the equations (1) and (2) include the number of calls accepted and the number of calls accepted as the number of calls accepted by the network device as a failure detection target. The number of successful calls, which is the number of successful calls, is used. During normal operation of a network device that is a failure detection target, the relationship between the number of accepted calls and the number of successful calls is usually a linear relationship in which an error of standard deviation between the number of accepted calls and the number of successful calls is added. It becomes. For this reason, when the number of accepted calls and the number of successful calls are used as the traffic information x and y during normal operation of the network device, the calculated value of the error e in the statistical linear relational expressions shown in the equations (1) and (2) is normal. Within the range. On the other hand, when the network device is faulty, the difference between the number of accepted calls and the number of successful calls increases, and the calculated value of error e shown in equations (1) and (2) falls outside the normal range. Therefore, when the calculated value of the error e is out of the normal range, it can be determined that a potential failure has occurred inside the network device, and the network device cannot normally perform traffic processing. Device failure can be detected.

  As the traffic information x, y, the number of accepted calls and the number of successful calls per unit time may be used. Here, the unit time may be 1 second, 1 minute, or 1 hour. However, the longer the unit time, the larger the tracking error. Therefore, it is preferable to shorten the time unit as much as possible.

  Further, as the traffic information x, y, even if the number of accepted calls and the number of successful calls of the entire network device that is the target of failure detection are used, the number of accepted calls and the number of successful calls for each device included in the network device May be used. Here, the devices included in the network device are, for example, a plurality of network interface ports included in the switch device, and the number of accepted calls and the number of successful calls of the entire switch device are used as the traffic information x, y. Alternatively, the number of accepted calls and the number of successful calls for each network interface port included in the switch device may be used.

  In the failure detection apparatus 50, the types of the traffic information x and y may be changed according to time zones such as morning, noon, and night, and traffic states such as a congestion state, a normal state, and a low traffic state. For example, as the traffic information x, y, the number of accepted calls and the number of successful calls may be used in a normal traffic state, and the number of accepted calls and the number of failed calls may be used in a congestion state.

  The failure detection device 50 preferably uses traffic information x, y that minimizes the variance of the error e in Equation (1) or Equation (2) for failure detection in each network device.

  As described above, the failure detection device 50 calculates the error e of the statistical linear relational expressions shown in the equations (1) and (2) in order to detect a failure of the network device. Below, the calculation method of the error e in Formula (1) (2) is demonstrated concretely.

(1) Calculation method of error e when traffic information x, y is not assumed to vary with time First, a method of calculating error e when traffic information x, y is not assumed to vary with time will be described. That is, a method for calculating the error e will be described on the assumption that the traffic information x, y and the traffic conversion information f (x), g (y) are fixed values. In the following, an example of calculating the error e in the above equation (2) will be described. However, as will be described later, the following calculation method can be applied to the equation (1).

Based on equation (2), error e is given by equation (3).

すると、式（４）〜（６）により、式（７）のように誤差ｅの２乗平均（分散）を算出することができる。

Here, the mean square (dispersion) of the traffic conversion information f (x) is defined as in the equation (4), and the mean square (dispersion) of the traffic conversion information g (y) is defined as in the equation (5). And the covariance of the traffic transformation information f (x) and g (y) is defined as in equation (6).

Then, the mean square (variance) of the error e can be calculated from Equations (4) to (6) as shown in Equation (7).

この結果、トラヒック変換情報ｇ（ｙ）の推定値ｇ（ｙ）_ｅｓｔが、式（９）により与えられる。

By transforming equation (7), a slope a that minimizes the mean square (variance) of error e is given by equation (8).

As a result, the estimated value g (y) _{est of} the traffic conversion information g (y) is given by Expression (9).

From the above, the error e with respect to the linear relational expression having the inclination a given by the equation (8) is given by the equation (10).

As described above, in equation (2), the error e is calculated on the assumption that the traffic information x, y and the traffic conversion information f (x), g (y) are fixed. It is also possible to apply to the formula (1). When the above calculation method is applied to equation (1), a slope a that minimizes the mean square (variance) of error e is given by equation (11). Further, an error e with respect to the linear relational expression having the inclination a given by Expression (11) is given by Expression (12).

(2) Method of calculating error e (T) when assuming time variation of traffic information x, y Next, the time variation of traffic information x, y is assumed by applying the method of calculating error e described above. A method for calculating the error e (T) in the case will be described. In the failure detection device 50, statistical information necessary for calculating the error e (T) following the time variation of the traffic information x, y is learned.

  Here, in order to calculate the error e (T) following the time variation of the traffic information x, y, it is necessary to learn the statistical information of the traffic information x, y based on the equation (13) described later. The statistical information is the statistical information of the mean square (variance) of the traffic information x learned by the equation (15), the statistical information of the covariance of the traffic information x, y learned by the equation (16), the equation ( 17) is the statistical information of the expected value of the traffic information x learned in 17), and the statistical information of the expected value of the traffic information y learned in Expression (18).

  As the statistical information learning method as described above, exponential weighting is used, but methods other than exponential weighting (for example, the Holt-Winters method, the Kalman filter, etc.) may be used.

For example, when the statistical information of the traffic information x, y is learned using exponential weighting, it is calculated at the time i−1 by the traffic information x, y collected at the time i as shown in the equation (13). By updating the statistical information of the traffic information x and y, the statistical information of the traffic information is learned so as to follow the time variation.

  In equation (13), t (i) represents traffic information x, y at time i (i> 0), α (i) represents statistical information of traffic information x, y at time i, and time i −1 traffic information x, y statistical information α (i−1) and traffic information x, y at time i (here, t (i)). Λ is a forgetting factor and is a constant of 1 or less. The statistical information α (i) of the traffic information x, y at time i depends on the latest traffic information t (i) as the value of the forgetting factor λ decreases. α (0) is an initial value of the statistical information of the traffic information x and y, and is a predetermined constant. For example, statistical information of traffic information x and y flowing on an adjacent network may be set as α (0).

ここで、傾きa（Ｔ−１）が次式により与えられる。

By the way, assuming the time variation of the traffic information x and y in the above equation (12), the error e (T) is given by equation (14).

Here, the inclination a (T-1) is given by the following equation.

Here, using the above-described exponential weighting, the root mean square (variance) of the traffic information x at time T is defined as in equation (15), and the traffic information x, y at time T as in equation (16). Is defined, statistical information of expected value of traffic information x at time T is defined as in equation (17), and expected value of traffic information y at time T is defined as in equation (18). Define statistical information.

Based on the statistical information learned by the equations (15) to (18), the error e (T) at the time T is given by the equation (14), and the error e (T) is within the normal range −δ by the equation (19). _- and by + [delta] ₊ whether within, the failure in the network device is detected.

Note that the threshold values δ ₋ and δ _{+ in} the normal range in the equation (19) are positive values determined in advance. Further, in the equations (15) to (19), the statistical information at the time T-1 immediately before is not necessarily used. As long as the traffic state is within a range that can be regarded as almost steady, statistical information at a time (for example, time T-2) that is more than one minute ago may be used.

  Further, the error e (T) at time T in the equation (19) may be obtained by changing the error e (T) at time T based on some function. For example, the error e (T) at the time T is the mean square deviation of the error e (T) at the time T, or the error e (T) itself at the time T is the standard of the error e (T−1) at the time T. It may be changed to one normalized by the deviation or one obtained by normalizing the standard deviation of the error e (T) at the time T by, for example, statistical information of traffic information.

For example, instead of using the error e (T) itself at the time T, the error e (T) at the time T normalized by the standard deviation of the error e (T-1) at the time T−1 is used. 19) can be changed to equation (20).

Here, the error e (T) at time T is given by equation (14). Note that the variance (root mean square) of the error e (T) at time T is learned as in equation (21).

When the error e (T) at the time T has a normal distribution or a distribution shape close thereto, it is desired to set δ ₋ , δ ₊ ∈ [2, 3] to an appropriate value. However, the threshold values δ ₋ and δ ₊ need not be set to the same value. The threshold values δ ₋ and δ ₊ may be changed according to time zones such as morning, noon, and night, and traffic states such as a congestion state, a normal state, and a low traffic state.

  Note that the forgetting factor λ used in the equations (15) to (18) and (21) may be changed according to the statistical information calculated by each equation. Further, the forgetting factor λ may be changed depending on the performance of the failure detection device 50 or the like. In addition, the initial values (that is, the values when time T = 0) of Expression (15) to Expression (18) and Expression (21) may be numerical values other than 0 described above. It may be a numerical value at the same time adjacent to the network device.

  As described above, the method of calculating the error e (T) following the time variation of the traffic information x and y in the equation (12) has been described. However, the traffic can be obtained by applying the above calculation method to the equation (10). It is also possible to calculate an error e (T) following the time variation of the conversion information f (x) and g (y).

  Thus, according to the failure detection principle of the present invention, whether or not the error e (T) following the time variation of the traffic information x, y or the traffic conversion information f (x), g (y) is within the normal range. Therefore, the failure detection accuracy of the network device can be improved.

<Configuration of failure detection device>
Next, the configuration of the failure detection device that detects a failure based on the failure detection principle as described above will be specifically described. The failure detection apparatus is physically an apparatus including an antenna, a modem, a CPU, a memory, and the like. FIG. 2 is a functional configuration diagram of the failure detection apparatus according to the first embodiment.

  As shown in FIG. 2, the failure detection apparatus 50 includes a traffic information acquisition unit 51, a traffic information storage unit 52, an error calculation unit 53, a statistical information storage unit 54, a failure detection unit 55, a detection condition storage unit 56, and a notification unit 57. It comprises.

  The traffic information acquisition unit 51 includes two pieces of traffic information x, y or traffic conversion information f (x) having a statistical linear relationship among traffic information generated when communication is performed using a network device that is a failure detection target. ) And g (y) are obtained two pieces of traffic information x and y having a statistical linear relationship. Specifically, the traffic information acquisition unit 51 periodically acquires traffic information x and y satisfying the above-described formula (1) or formula (2).

  The traffic information storage unit 52 stores the traffic information x and y acquired by the traffic information acquisition unit 51.

  The error calculation unit 53 calculates an error e of a statistical linear relational expression indicating a statistical linear relation satisfied by the traffic information x, y or the traffic conversion information f (x), g (f). Specifically, the error calculation unit 53 calculates the slope a of the statistical linear relational expression as shown in Expression (8), and calculates the error e from the linear relational expression having the calculated slope a. The error calculation unit 53 calculates, as the slope a of the statistical linear relational expression, a slope that minimizes the mean square (variance) of the error e, as shown in Expression (8).

  Further, the error calculation unit 53 may calculate an error e (T) of a statistical linear relational expression indicating a statistical linear relation that the traffic information x, y or the traffic conversion information f (x), g (f) satisfies. Good. Specifically, the error calculation unit 53 uses the last-time statistical information stored in the statistical information storage unit 54 to be described later and the traffic information newly acquired by the traffic information acquisition unit 51 to use the traffic information x , Y, and statistical information (formulas (13) and (15) to (18)) necessary for calculating the error e (T) following the time fluctuation. The error calculation unit 53 calculates the slope a (T) of the statistical linear relational expression using the learned statistical information, and calculates the error e from the linear relational expression having the calculated slope a (T).

  The statistical information storage unit 54 stores the statistical information learned by the error calculation unit 53. Specifically, the error calculation unit 53 includes statistical information on the traffic information x and y learned using Expression (13), and statistical information on the mean square of the traffic information x learned using Expression (15). , Statistical information on the covariance of traffic information x and y learned using equation (16), statistical information on expected values of traffic information x learned using equation (17), and equation (18) The statistical information of the expected value of the learned traffic information y is stored.

The failure detection unit 55 detects a failure of the network device based on the error e calculated by the error calculation unit 53. Specifically, the failure detection unit 55 uses the normal range (−) in which the value of the error e (T) calculated by the error calculation unit 53 is stored in the detection condition storage unit 56, as shown in Expression (19). [delta] _- and determines whether there the - [delta ₊ within). Alternatively, as shown in the equation (20), the failure detection unit 55 has a value obtained by normalizing the error e (T) calculated by the error calculation unit 53 with the standard deviation of the error e until time T−1. storage unit 56 to the stored normal range (- [delta _- and - [delta ₊ limit) whether may determine whether the. If the standard deviation of error e until time T or the value obtained by normalizing error e (T) with the standard deviation of error e until time T-1 is not within the normal range, Detects faults.

  Further, the failure detection unit 55 initializes the statistical information stored in the statistical information storage unit 54 when a network device failure is detected. On the other hand, if no failure of the network device is detected, the failure detection unit 55 does not initialize the statistical information stored in the statistical information storage unit 54, and continues the learning of statistical information by the error calculation unit 53.

The detection condition storage unit 56 stores detection conditions for the failure detection unit 55 to detect a failure. Specifically, the detection condition storage unit 56 stores the values of the normal range threshold values δ ₋ and δ ₊ shown in Expression (19) or Expression (20). The detection condition storage unit 56 may store different threshold values δ ₋ and δ ₊ according to time zones such as morning, noon, and night, and traffic states such as a congestion state, a normal state, and a low traffic state.

  When the failure detection unit 55 detects a network device failure, the notification unit 57 notifies the network device from which the failure has been detected.

<Operation of failure detection device>
Next, the operation of the failure detection apparatus according to the first embodiment configured as described above will be described. FIG. 3 is a flowchart showing a failure detection operation of the failure detection apparatus according to the first embodiment. Note that steps S102 to S108 in FIG. 3 are repeated at predetermined time intervals.

  At the start of this operation, the types of traffic information x and y acquired by the traffic information acquisition unit 51 are determined, and the statistical information stored in the statistical information storage unit 54 is initialized (step S101). Specifically, the types of traffic information x and y (for example, the number of accepted calls and the number of successful calls) are determined according to the time zone and traffic state, and the statistical information stored in the statistical information storage unit 54 is expressed by the formula ( 13) Initialized according to the initial values shown in (15) to (18).

  Although the initial value is set to 0 in equations (13) and (15) to (18), the initial value is set based on the statistical information of the adjacent network device adjacent to the network device that is the target of failure detection. May be. Here, the adjacent network device is another network device that belongs to the same higher level network device as the network device that is a failure detection target. For example, when the network device that is the failure detection target is a “sector”, the adjacent network device The device is another sector belonging to the same radio base station.

  The traffic information acquisition unit 51 includes two types of traffic information x, y or traffic conversion information f (x) and g (y) having a statistical linear relationship among the types of traffic information determined in step S101. Two pieces of traffic information x and y that are related are acquired (step S102). In the following, a case where two pieces of traffic information x and y having a statistical linear relationship are acquired will be described. However, in this operation, two pieces of traffic in which the traffic conversion information f (x) and g (y) have a statistical linear relationship are described. The present invention is also applicable when information x and y are acquired.

  Based on the traffic information x, y newly acquired by the traffic information acquisition unit 51 and the statistical information stored in the statistical information storage unit 54, the error calculation unit 53 follows the time variation of the traffic information x, y. Statistical information necessary for calculating the error e (T) is learned, and the learned statistical information is stored in the statistical information storage unit 54 (step S103). Specifically, the error calculation unit 53 is based on the traffic information x (T), y (T) at the time T and the statistical information at the time T-1, and the above equations (13), (15) to (15)-( 18) is used to learn statistical information at time T.

  The error calculation unit 53 calculates the slope a (T-1) of the statistical linear relational expression at time T-1 based on the statistical information at time T-1 stored in the statistical information storage unit 54 at time T. (Step S104). Specifically, the error calculation unit 53 calculates the statistical linearity at time T-1 based on the statistical information on the covariance of the traffic information x and y at time T-1 and the mean square (variance) of the traffic information x. The inclination a (T-1) of the relational expression is calculated.

  The error calculation unit 53 calculates an error e (T) of the statistical linear relational expression with respect to the linear relational expression having the slope a (T−1) calculated in step S104 (step S105). Specifically, the error calculation unit 53 calculates the error of the statistical linear relational expression at time T based on the slope a (T-1) of the statistical linear relational expression at time T-1 as shown in Expression (14). e (T) is calculated.

The failure detection unit 55 detects a failure of the network device based on the error e (T) calculated in step S105 (step S106). Specifically, the failure detecting unit 55, as shown in equation (19), the value of the error e (T) at time T is the normal range _- determines whether the (- [delta and + [delta] ₊ within) . Alternatively, the failure detection unit 55 normalizes the error e (T) at time T with the standard deviation σ _e (T−1) of the error e (T−1) at time T−1 as shown in Expression (20). phased value normal range _- may determine whether the (- [delta and + [delta] ₊ within).

  When the failure detection unit 55 detects a failure of the failure detection target network device, the notification unit 57 notifies the network device to that effect (step S107). When a failure notification is sent to the network device, the statistical information stored in the statistical information storage unit 54 is initialized (step S108), and the operation returns to step S102.

<Action and effect>
According to the wireless communication system according to the first embodiment, the network is based on the error e of the statistical linear relational expression indicating the statistical linear relation between the traffic information x, y or the traffic conversion information f (x), g (y). By detecting a device failure, it is possible to improve the detection accuracy of a network device failure even when the traffic fluctuates greatly over time.

  According to the wireless communication system according to the first embodiment, since the error e (T) following the time variation of the traffic information x, y is calculated, the failure detection accuracy of the network device based on the error e (T) is increased. It can be improved further.

[Second Embodiment]
Next, a wireless communication system according to the second embodiment will be described focusing on differences from the first embodiment. The wireless communication system according to the second embodiment is different from the first embodiment in that the error of the statistical linear relational expression is calculated using the periodic steadiness of traffic information.

<Fault detection principle>
A failure detection principle according to the second embodiment will be described. In the failure detection apparatus 50 according to the second embodiment, two pieces of traffic information x and y having a statistical linear relationship, or traffic conversion information f (x) and g (y) based on conversion functions f and g are statistically linear. A failure of the network device is detected by using the periodic stationarity of the two pieces of traffic information x and y that are related.

  FIG. 4 is a diagram for explaining periodic steadiness of traffic information in the wireless communication system according to the second embodiment. As shown in FIG. 4, the fluctuation of traffic information has a periodicity of period m in the time direction. Thus, a stochastic process in which statistical information such as expected values and standard deviations becomes a periodic function is called a cyclic stationary random process with a period m. Hereinafter, on the assumption that the traffic information has periodic stationarity, a failure detection principle using the period stationarity will be described.

  As shown in FIG. 4, when focusing on a certain time t, traffic information x (t), x (at time t, t + (m-1), t + 2 (m-1), ..., t + N (m-1). t + (m−1)), x (t + 2 (m−1)),..., x (t + N (m−1)) are stationary sequences. Therefore, the tracking errors A ′, B ′, C ′ between the traffic information of the previous cycle can be suppressed to be smaller than the tracking errors A, B, C between the traffic information of the previous cycle, and as a result, the network device This makes it possible to improve the accuracy of fault detection.

ここで、傾きａ（Ｔ−ｍ＋１）が次式により与えられる。

Next, a method for calculating the error e (T) when the periodic steadiness of the traffic information x and y is used will be described. When the periodic steadiness of the traffic information x and y is used, the error e (T) is given by equation (22).

Here, the inclination a (T−m + 1) is given by the following equation.

Here, using the statistical information one cycle before (that is, statistical information at time T−m + 1 when one cycle is m), statistical information on the variance of traffic information x at time T as shown in Equation (23). , Defining statistical information on the variance of traffic information y at time T as in equation (24), defining statistical information on expected values of traffic information x at time T as in equation (25), As shown in (26), the statistical information of the expected value of the traffic information y at time T is defined.

Based on the statistics that are learned by the formula (23) to (26), the error e (T) is given at time T by the equation (27), the error e (T) is the normal range - [delta _- and + [delta] ₊ within A failure in the network device is detected depending on whether it is in the network.

When the error e (T) at time T is normalized by the standard deviation of the error e at time T−m + 1 one cycle before, equation (27) can be changed to equation (28).

Here, the mean square (variance) of the error e (T) is learned as shown in Equation (29).

  Note that the above equations (22) to (29) are examples, and the error e (T) may be calculated using the periodic steadiness of the traffic information x and y using other equations. The period m can be set to, for example, one day or one week, but is not limited thereto. In the above formulas (22) to (29), the statistical information of one cycle before is used. However, if the traffic can be regarded as almost steady, the time before and after the cycle (for example, , Time T−m + 2, time T−m, etc.) may be used.

  Further, β used in equations (23) and (24), λ used in equations (25) and (26), and γ used in equation (29) are forgetting factors. The forgetting factors β, λ, γ may be changed depending on the performance of the failure detection device 50 or the like.

<Configuration of failure detection device>
Next, the configuration of the failure detection apparatus 50 according to the second embodiment will be described. The failure detection apparatus 50 has the same configuration as that of the first embodiment, but the statistical information storage unit 54 is learned by the error calculation unit 53 so that the error calculation unit 53 can refer to the statistical information of the previous cycle. The history of learned statistical information as well as the last statistical information is stored.

<Operation of failure detection device>
Next, the operation of the failure detection apparatus according to the second embodiment configured as described above will be described. FIG. 5 is a flowchart showing a failure detection operation of the failure detection apparatus according to the second embodiment. Note that steps S202 to S208 in FIG. 5 are repeated at predetermined time intervals. Steps S201 and S202 are the same as steps S101 and S102 in FIG. In FIG. 5, m represents one cycle.

  Based on the traffic information x, y newly acquired by the traffic information acquisition unit 51 and the statistical information at time Tm + 1 one cycle before stored in the statistical information storage unit 54, the error calculation unit 53 performs traffic information. Statistical information necessary for calculating the error e (T) following the time variation of x and y is learned, and the learned statistical information is stored in the statistical information storage unit 54 (step S203). Specifically, the error calculation unit 53 is based on the traffic information x (T), y (T) at the time T and the statistical information at the time T-m + 1 one cycle before, from the above formulas (23) to (23). 26) is used to learn statistical information at time T.

  Based on the statistical information at the time T-m + 1 one cycle before stored at the time T at the time T, the error calculation unit 53 calculates the slope a (T−T) of the statistical linear relational expression at the time T−m + 1. m + 1) is calculated (step S204). Specifically, the error calculation unit 53 calculates the statistical linearity at time T−m + 1 based on the statistical information on the covariance of the traffic information x and y at time T−m + 1 and the mean square (variance) of the traffic information x. The inclination a (T−m + 1) of the relational expression is calculated.

  The error calculation unit 53 calculates an error e (T) of the statistical linear relational expression with respect to the linear relational expression having the inclination a (T−m + 1) calculated in step S204 (step S205). Specifically, the error calculation unit 53 calculates the error of the statistical linear relational expression at time T based on the slope a (Tm + 1) of the statistical linear relational expression at time T−m + 1 as shown in Expression (22). e (T) is calculated.

  Steps S206 to S208 are the same as steps S106 to S108 in FIG.

  According to the wireless communication system according to the second embodiment, the error e (T) is calculated based on various statistical information learned one cycle before using the periodic steadiness of the traffic information x and y. Therefore, it is possible to further improve the accuracy of detecting a failure of the network device based on the error e (T).

[Other Embodiments]
The present invention is not limited to a specific system, and may be applied to any appropriate communication system. For example, the present invention may be applied to ISDN (Integrated Services Digital Network), ADSL (Asymmetric Digital Subscriber Line), NGN (Next Generation Network) systems, and the like. Further, the present invention is a mobile communication system, a W-CDMA (Wideband Code Division Multiple Access) system, an HSDPA / HSUPA (High Speed Downlink Packet Access / High Speed Uplink Packet Access) system, an LTE (Long Term Evolution). ) System, LTE-Advanced system, IMT-Advanced (International Mobile Telecommunication-Advanced) system, WiMAX (Worldwide Interoperability for Microwave Access), Wi-Fi system, and the like.

  Although the present invention has been described with reference to particular embodiments, they are merely illustrative and those skilled in the art will appreciate various variations, modifications, alternatives, substitutions, and the like. . Although specific numerical examples have been described in order to facilitate understanding of the invention, these numerical values are merely examples and any appropriate values may be used unless otherwise specified. The division of the embodiments or items is not essential to the present invention, and the matters described in two or more embodiments or items may be used in combination as necessary. For convenience of explanation, an apparatus according to an embodiment of the present invention has been described using a functional block diagram, but such an apparatus may be realized by hardware, software, or a combination thereof. The present invention is not limited to the above embodiments, and various modifications, modifications, alternatives, substitutions, and the like are included in the present invention without departing from the spirit of the present invention.

DESCRIPTION OF SYMBOLS 10 ... Radio | wireless terminal apparatus, 20 ... Radio base station apparatus, 30 ... Relay apparatus, 40 ... Server apparatus, 50 ... Failure detection apparatus, 51 ... Traffic information acquisition part, 52 ... Traffic information storage part, 53 ... Error calculation part, 54 ... Statistical information storage unit, 55 ... Fault detection unit, 56 ... Detection condition storage unit, 57 ... Notification unit

Claims (7)

A failure detection device for detecting a failure of a network device,
Of the traffic information generated when communicating using the network device, two pieces of traffic information having a statistical linear relationship or two pieces of traffic information in which conversion information by a conversion function has a statistical linear relationship is acquired. An acquisition unit;
An error calculating unit for calculating an error of the statistical linear relational expression indicating the statistical linear relation;
A failure detection unit that detects a failure of the network device based on the error calculated by the error calculation unit;
A learning unit that learns the two traffic information following time variation,
The error calculation unit calculates an inclination of the statistical linear relational expression based on the two traffic information learned by the learning unit, and calculates an error of the statistical linear relational expression with respect to the linear relational expression having the calculated inclination. A failure detection apparatus characterized by calculating .   The failure detection apparatus according to claim 1, wherein the traffic information acquisition unit acquires the two types of traffic information according to a time zone or a traffic state. The slope failure detection device according to claim 1 or 2, characterized in that the mean square of the error of the statistical linear relationship is the slope to minimize. The error calculating unit, based on the two traffic information learned before moment by the learning section, any of claims 1 to 3, characterized in that to calculate the slope of the statistical linear relations before moment failure detecting apparatus according to an item or. The error calculation unit calculates the slope of the statistical linear relationship based on the two traffic information learned one cycle before by the learning unit using the periodic steadiness of the two traffic information. The failure detection apparatus according to any one of claims 1 to 3 . The failure detection unit detects a failure of the network device when the value of the error or a value obtained by normalizing the error with a standard deviation of the error is outside a normal range. Item 6. The failure detection device according to any one of Items 1 to 5 . The failure detection apparatus according to claim 6 , wherein the threshold value of the normal range is changed according to a time zone or a traffic state.

Images