Autonomous decentralized event trigger detection device

Abstract

In the present invention, traffic passing between a transmitting device and a receiving device disposed in separated locations is aggregated separately at both the transmitting end and the receiving end, and the transmitting end and the receiving end each autonomously detect an event trigger such as an addition of an optical path in a decentralized manner on the basis of the results of the traffic aggregation. In this case, the transmitting end and the receiving end are synchronized in terms of time, and then a model that accounts for differences in aggregation timing between the transmitting end and the receiving end is used to detect occurrence of an event and execute processing corresponding to a pre-registered event condition. The aggregation timing is re-adjusted so as to minimize the error therebetween on the basis of the results of the traffic aggregation. The probability that the transmitting end and the receiving end return the same condition determination result is calculated as a normal detection probability. In consideration of the differences in aggregation timing, an interval with traffic observed by both the transmitting end and the receiving end is distinguished from an interval with traffic observed only by the transmitting end or the receiving end to perform the calculation.

Description

Autonomous decentralized event trigger detection device

[0001] The present invention relates to an autonomous distributed event trigger detection device.

[0002] The amount of data traffic flowing through a network varies depending on the user's network service usage patterns. When transferring data, routers and network interface devices are powered by electricity, so the amount of power consumed by the devices is usually proportional to the amount of traffic they process.

[0003] When designing a network, routers and links are used that can accommodate peak traffic volumes, which results in excess processing capacity and the associated power consumption when traffic volumes decrease. In order to reduce these cost-increasing factors, progress is being made in the development of technologies that can adjust router processing capacity and link bandwidth according to the traffic volume being processed within the network.

[0004] A common method for adjusting link bandwidth in an optical network is to change the number of optical paths set between two points. Here, we assume a case in which optical gateways 10 and 20 at two points are connected by an optical network 15, and communication is carried out via two optical paths 15a and 15b, as in the optical communication system 100 shown in Figure 1. In this example, optical gateways 10 and 20 for the optical network are located at each point, and data is normally transferred via one optical path 15a allocated between the optical gateways 10 and 20. One optical gateway 10 has optical transceivers 11 and 12 as interfaces, and the other optical gateway 20 has optical transceivers 21 and 22.

[0005] In the optical communication system 100 shown in Fig. 1, for example, time-series changes in traffic volume as shown in Fig. 2 are monitored by at least one of the optical gateways 10 and 20. When an increase in traffic is detected, the system transitions from the upper state shown in Fig. 3 to the lower state. That is, a new optical path 15b is established, and link aggregation of the multiple optical paths 15a and 15b is performed, bundling the two optical paths 15a and 15b to make them available for data transfer. On the other hand, when the traffic volume decreases to the point where only the single optical path 15a can handle all traffic, one of the two optical paths 15a and 15b is stopped, and data transfer continues via the remaining path.

[0006] 1, it is necessary to set the optical interface on / off settings to the same value in both optical gateways 10 and 20. When an optical path addition / removal event is detected, the optical interface on / off settings in both optical gateways 10 and 20 are completed at the same time, which makes it possible to add / remove optical paths without causing problems such as packet loss.

[0007] There are two types of event detection methods: one is a centralized management type detection method and the other is an autonomous distributed type detection method. For example, an example of the centralized management type detection method is described in Non-Patent Document 1.

[0008] "OVERVIEW", Prometheus, Internet<URL:https: / / prometheus.io / docs / introduction / overview / >

[0009] In the centralized management detection method, as in the optical communication system 100A shown in Fig. 4A, it is assumed that an event is detected in one of the two optical gateways 10, 20, and an event notification message after the detection is sent to the other gateway, which then issues an instruction to set up an optical path for the event and sets up its own optical path. In the example shown in Non-Patent Document 1, event detection is performed by the server to be monitored, and the notification is sent to the controller that monitors the server.

[0010] On the other hand, when an autonomous distributed detection method is adopted, as in the optical communication system 100B shown in Figure 4B, it is assumed that event detection is performed by both optical gateways 10 and 20, and after detection, optical path setting for the event is performed by each gateway.

[0011] On the other hand, when setting up optical paths at the same time, the two detection methods mentioned above, the centralized management type and the autonomous distributed type, each have advantages and disadvantages as shown below. The advantage of the centralized management type is that only one packet sample is used for event detection, and both gateways are set for the same sample, which prevents malfunctions such as one gateway not setting up after an event is detected.

[0012] The disadvantage of the centralized management type is that the timing of lightpath setup at both gateways is delayed due to message transfer. That is, when one gateway detects an event, it must send a message to notify the other gateway of the event occurrence, which requires time for message creation and message transfer. As a result, a delay occurs due to transmission delays in the timing at which the receiving gateway learns of the event occurrence when the message arrives.

[0013] Such a timing discrepancy can lead to a situation similar to that of the optical communication system 100C shown in Figure 4C. That is, even though the optical transceiver 12 of the optical gateway 10 that detected the event switches from off to on and establishes an additional optical path 15b, the optical transceiver 22 of the optical gateway 20 remains off. Therefore, when the optical gateway 10 erroneously recognizes that optical path 15b is available and attempts to transmit a packet through optical path 15b, packet loss occurs. Furthermore, the optical transceiver 12 wastes power even though optical path 15b is not yet available.

[0014] One way to avoid the timing discrepancy described above is to share the lightpath setup execution time between both gateways. However, this method is based on the lightpath setup event detection result based on traffic information before the setup time. Therefore, in situations where traffic fluctuates rapidly, a lag occurs between the detection of the event and the actual lightpath setup, which can result in a traffic loss due to the increased traffic not being accommodated.

[0015] The advantage of the autonomous distributed architecture is that, depending on the conditions, it is possible to synchronize message detection at both gateways. In the case of event detection using data traffic statistics, this statistical information is obtained by periodically aggregating information such as packet counters. The start timing of the aggregation can be set in 10 millisecond increments starting from exactly midnight on New Year's Day. The starting times for these start times can be synchronized using time synchronization protocols such as NTP (Network Time Protocol) and PTP (Precision Time Protocol), with an error range of milliseconds for the former and an error range of nanoseconds for the latter.

[0016] The disadvantage of the autonomous distributed system is that the timing of statistical information collection by both gateways is different due to the aforementioned errors and data transfer delays. In other words, both gateways cannot detect events based on the same packet arrival pattern sample.

[0017] On the other hand, events that trigger the establishment or deletion of lightpaths between two gateways must be given to both gateways simultaneously. In other words, in the case of traffic measurement, both gateways must detect events from the results of observing the same packets.

[0018] In general event detection, for example, as shown in Figure 5, traffic volume that fluctuates over time is observed and aggregated between the start time t1 and the end time t2 of the packet observation period T0, and traffic statistical data such as the average packet arrival rate is repeatedly calculated, for example, every 10 msec.

[0019] Although both gateways can observe the same packets, the observed packets will differ due to the aforementioned errors in the counting time and packet transfer delays. For example, if you try to observe packets that arrive within a certain observation time interval on the sending side on the receiving side, due to errors in the counting time, it is possible that the packet before that observation time interval will be observed in the receiving side's observation time interval, and the packet just before the end of that sending side's observation time interval will be observed in the next receiving side's observation time interval.

[0020] Due to the time difference between packets to be observed and the timing difference between statistical data collection in such an autonomous distributed system, two gateways may show different event detection results for the same packet time series.

[0021] 6, the difference between the observation interval T11 on the sending side and the observation interval T12 on the receiving side is small, so the aggregated result for the observation interval T11 (e.g., 91 Gbps) and the aggregated result for the observation interval T12 (e.g., 90 Gbps) are almost the same. As a result, the optical gateways 10 and 20 on both the sending and receiving sides observe traffic volumes above their respective thresholds, and the event detections on both the sending and receiving sides coincide.

[0022] 7, there is a large discrepancy between the observation interval T21 on the sending side and the observation interval T22 on the receiving side, resulting in a large difference between the aggregation result for observation interval T21 (e.g., 91 Gbps) and the aggregation result for observation interval T22 (e.g., 80 Gbps). Therefore, when the aggregation result for observation interval T21 exceeds the threshold in the optical gateway 10 on the sending side and an event is detected, the aggregation result for observation interval T22 does not exceed the threshold in the optical gateway 20 on the receiving side, so the event is not detected. In other words, the event detection fails between the two optical gateways 10 and 20, and only one optical gateway 10 sets up an optical path, so an optical path cannot be established between the optical gateways 10 and 20 and the interface wastes power.

[0023] That is, in the case of devices such as gateways, due in part to the influence of the large geographical distance between the sending side and the receiving side, as shown in FIG. 8, there is a relative time difference between the observation period T31 in which the sending side observes traffic and the observation period T41 in which the receiving side observes traffic, and therefore the sending side and the receiving side cannot observe the same traffic volume.

[0024] Furthermore, the rate at which an optical path cannot be established and an interface wastes power also changes because the situation fluctuates due to factors such as transmission delays, timing synchronization errors, packet generation rates per unit time, and event detection thresholds. Therefore, it is difficult for network designers to correctly evaluate the impact of power waste when autonomous distributed event detection is introduced.

[0025] The present invention has been made in consideration of the above-mentioned circumstances, and aims to provide an autonomous distributed event trigger detection device that can easily perform appropriate processing in accordance with the results of aggregation when multiple transmission devices that process the same traffic employ autonomous distributed event detection to observe and aggregate traffic volume.

[0026] The autonomous distributed event trigger detection device of the present invention utilizes a first device and a second device that transmit or receive the same traffic, monitors the traffic on the first device side as a first traffic time series, monitors the traffic on the second device side as a second traffic time series, collects the monitoring results of the first traffic time series and the second traffic time series at the same period, and has an event trigger identification unit that identifies whether the collected results match a pre-set condition, wherein the event trigger identification unit comprises: a traffic designation unit that designates traffic to be monitored by the first device and the second device; a collection timing synchronization unit that synchronizes the traffic collection timing of the first device and the traffic collection timing of the second device; an event detection unit that detects the occurrence of an event from the results collected at the traffic collection timing of the first device and the second device; and a processing execution unit that executes processing corresponding to the condition when the traffic collection results match a pre-registered event occurrence condition at each timing.

[0027] According to the autonomous distributed event trigger detection device of the present invention, when multiple transmission devices that process the same traffic employ autonomous distributed event detection to observe and aggregate traffic volumes, it becomes easy to perform appropriate processing based on the aggregation results. That is, after synchronizing the aggregation timing between the first device and the second device, it is possible to associate the aggregation results with event occurrence conditions while taking into account the timing difference, and appropriate processing can be performed based on the aggregation results.

[0028] 1 is a block diagram showing an example of the configuration of a main part of an optical communication system. FIG. 2 is a graph showing an example of time-series changes in the amount of traffic transmitted through an optical network. FIG. 3 is a schematic diagram showing an example of changes in the state of an optical communication system. FIG. 4 is a schematic diagram showing an example of a state of an optical communication system. FIG. 5 is a schematic diagram showing an example of a state of an optical communication system. FIG. 6 is a time chart showing an example of time-series changes in the amount of traffic transmitted through an optical network. FIG. 7 is a time chart showing an example of time-series changes in the same traffic on the transmitting side and the receiving side. FIG. 8 is a time chart showing an example of time-series changes in the same traffic on the transmitting side and the receiving side. FIG. 9 is a time chart showing an example of a correspondence relationship between each observation interval on the transmitting side and the receiving side and the traffic amount. FIG. 10 is a schematic diagram showing the relationship between multiple parameters used to calculate the event trigger detection success rate and the calculation results. FIG. 11 is a time chart showing multiple observation intervals in a model that takes into account partial overlapping of the observation intervals of multiple devices. FIG. 11 is a schematic diagram showing how the timings of the observation intervals of multiple devices are mapped on a common time axis. FIG. 12 is a time chart showing the ranges of each interval divided into multiple sections on a common time axis onto which the timings of multiple devices are mapped. FIG. 1 is a schematic diagram showing a normal detection pattern in which a plurality of devices output the same detection result at their respective data aggregation timings. FIG. 2 is a schematic diagram showing a normal detection pattern in which a plurality of devices output the same detection result at their respective data aggregation timings. FIG. 3 is a schematic diagram showing a normal detection pattern in which a plurality of devices output the same detection result at their respective data aggregation timings. FIG. 4 is a block diagram showing an example configuration of an event trigger detection device. FIG. 5 is a logical block diagram showing an example configuration of an event trigger detection device. FIG. 6 is a flowchart showing an overview of the operation of an event trigger detection device. FIG. 7 is a block diagram showing an example configuration of an autonomous distributed optical transmission device. FIG. 8 is a flowchart showing the operation at the time of startup of an autonomous distributed optical transmission device. FIG. 9 is a flowchart showing the optical path setting event detection operation of an autonomous distributed optical transmission device. FIG. 10 is a time chart showing an example of timing changes before and after performing aggregation timing readjustment. FIG. 11 is a flowchart showing the operation of aggregation timing readjustment. FIG. 12 is a time chart showing an example of aggregation timing when statistical data is acquired with a time difference δ shifted.1 is a time chart showing an example of aggregation timing when statistical data for multiple periods is acquired at different timings. FIG. 2 is a schematic diagram showing an example of the configuration of statistical data for multiple periods acquired at different timings. FIG. 3 is a time chart showing an example of minimum error pattern 1 when aggregation timing is readjusted. FIG. 4 is a time chart showing an example of minimum error pattern 2 when aggregation timing is readjusted. FIG. 5 is a time chart showing an example of minimum error pattern 3 when aggregation timing is readjusted. FIG. 6 is a block diagram showing an example of a device configuration that can be used to acquire statistical data of synchronization errors. FIG. 7 is a flowchart showing an example of operation when the result of trigger detection success rate calculation is reflected in control.

[0029] An embodiment of the present invention will be described below with reference to the drawings. <Parameters Used to Calculate Success Rate> Fig. 9 shows the relationship between a plurality of parameters used to calculate the "event trigger detection success rate" and the calculation results.

[0030] For example, assuming that an autonomous distributed detection control system, such as the optical communication system 100B shown in Figure 4B, is adopted and event trigger detection is performed based on the amount of traffic passing through each of the two optical gateways 10 and 20, the percentage of times that the event trigger detection results (detection / no detection) at the two optical gateways 10 and 20 match can be defined as the "event trigger detection success rate."

[0031] 9 estimates a probability corresponding to the "event trigger detection success rate" based on a plurality of input parameters Pm, and outputs the calculation result 39. In this embodiment, the parameters Pm are a data transfer delay Pm11, a synchronization error Pm12, a counting interval Pm2, and an average packet generation rate Pm3.

[0032] The data transfer delay Pm11 represents the length of the delay time from when the transmitting side transmits a packet until the packet is observed by the receiving side. The synchronization deviation Pm12 represents the timing deviation between the transmitting side and the receiving side that occurs after the transmitting side optical gateway 10 and the receiving side optical gateway 20 are time-synchronized using a reference time.

[0033] In reality, due to the influence of the geographical distribution of the locations of the gateways, it is inevitable that time deviations of the order of nanoseconds will occur when synchronizing time using PTP, and of the order of milliseconds when synchronizing time using NTP. Due to the influence of the data transfer delay Pm11 and synchronization deviation Pm12 described above, for example, as shown in Figure 8, a relative time deviation will occur between the observation section T31 on the sending side and the observation section T41 on the receiving side.

[0034] The counting interval Pm2 represents the length of one period of the interval over which the average packet generation rate Pm3 is counted in each of the optical gateways 10 and 20. For example, the time length of each of the observation periods T31 and T41 shown in Fig. 8 corresponds to the counting interval Pm2. The average packet generation rate Pm3 represents the result of counting and statistical processing in each of the observation periods T31 and T41 of each of the optical gateways 10 and 20, i.e., the observed value of the traffic volume (average value per period).

[0035] <Model of Observation Section> In this embodiment, when the calculation processing 30 calculates the "event trigger detection success rate", the calculation is performed by adopting a model such as that shown in FIG. 10, taking into account that the observation sections of multiple optical gateways 10, 20 partially overlap.

[0036] 10, it is assumed that the transmitting side performs observations in the interval between times ts1 and ts2 for each observation period T01, and the receiving side performs observations in the interval between times td1 and td2 for each observation period T02 (the same length as T01). This model means that the entire system can be divided into three observation periods TA, TB, and TC for evaluation.

[0037] That is, although the entire observation intervals on the transmitting and receiving sides are shifted in time by the amount of data transfer delay Pm11 and synchronization error Pm12, the observation interval TB is an area where the transmitting side interval and the receiving side interval overlap, so the number of arriving packets common to both the transmitting side and receiving side observation intervals can be observed. Meanwhile, the observation interval TA can observe the number of arriving packets observed only on the transmitting side. Furthermore, the observation interval TC can observe the number of arriving packets observed only on the receiving side.

[0038] <Calculation Prerequisites> In this embodiment, the calculation process 30 calculates the "event trigger detection success rate" under the following conditions (1) and (2): (1) A queuing model is used for the calculation. Also, the packet arrival process is assumed to be a Poisson process. That is, it is assumed that the packet arrival interval is exponentially distributed. (2) The rate of the number of packets is used for the calculation. That is, since calculation using bit rates is difficult, calculation is performed on a packet-by-packet basis. Also, in a situation where packet length can be assumed to be approximately constant, such as when jumbo frames are the main traffic, calculation using the number of packets is expected to improve the accuracy of the calculation results.

[0039] <Details of calculation process> - <Modeling of deviation of observed object> Fig. 11 shows how the timing of each observation section of multiple devices is mapped onto a common time axis. In a network such as the optical communication system 100B, there is a time deviation between the timing of the observation sections on the transmitting side and the receiving side, as shown in Fig. 8, and therefore, the aggregation timing is mapped taking this time deviation into consideration, as shown in Fig. 11.

[0040] That is, the optical gateway 10 on the transmitting side counts the number of arriving packets in one period for each transmitting-side ts counting timing for the observed packet time series Px on the time axis tx. The optical gateway 20 on the receiving side counts the number of arriving packets in one period for each receiving-side td counting timing for the observed packet time series Py on the time axis ty. Therefore, each transmitting-side ts counting timing and observed packet time series Px on the time axis tx, and each receiving-side td counting timing and observed packet time series Py on the time axis ty are mapped onto a common time axis tz. This makes it possible to model the influence of the timing difference Te on the counting between the transmitting and receiving sides on the common time axis tz.

[0041] Calculation process 30 counts the number of packets arriving on the receiving side in the interval "t11-t12" on the common time axis tz, and counts the number of packets arriving on the transmitting side in the interval "t21-t22." Then, it calculates the probability that the number of arriving packets per cycle satisfies or does not satisfy the event detection condition.

[0042] -<Deriving the Probability that the Transmitter and Receiver Show the Same Detection Result> The ranges of each of the multiple sections divided on a common time axis tz onto which the timings of multiple devices are mapped are shown in Fig. 12. Calculation process 30 calculates the probability that the transmitter and receiver will return the same detection result for each observation section after time synchronization, using the method shown below.

[0043] The calculation process 30 arranges the lightpath passing packet generation events in chronological order and assumes that the timing of the events follows a Poisson process, which makes the calculation probability independent of the start time of the aggregation timing.

[0044] The calculation process 30 maps the aggregation timing of the sending side onto a time series on the time axis tz, and maps the aggregation timing of the receiving side onto a time series on the same time axis tz at a timing shifted by the synchronization error and transfer delay from the aggregation timing of the sending side.

[0045] The sending and receiving devices each perform observations at the same traffic counting interval "I", so that for each counting period, there are three types of intervals: time t11 to t21, time t21 to t12, and time t12 to t22, as shown in Figure 12.

[0046] In the section "D" in length from time t11, the observation timing of the transmitting side, to time t21, the observation timing of the receiving side, the transmitting side observes the number of packets arriving in that section, and the receiving side observes the number of packets arriving in the next observation section. Also, in the section "I-D" in length from time t21, the observation timing of the receiving side, to time t12, the observation timing of the transmitting side, both the transmitting side and the receiving side observe the number of packets arriving in that section. Also, in the section "D" in length from time t12, the observation timing of the transmitting side, to time t22, the observation timing of the receiving side, the receiving side observes the number of packets arriving in that section, and the transmitting side observes the number of packets arriving in the previous observation section.

[0047] -<Normal Detection Patterns> Three types of normal detection patterns that represent situations in which the transmitting and receiving devices output the same detection results at their respective data collection timings are shown in FIGS. 13A to 13C.

[0048] 13A corresponds to a situation in which the number of packets arriving is equal to or greater than the threshold "H" between times t21 and t12, i.e., within the observation period Tc common to both the sending and receiving sides. Therefore, in normal detection pattern P11, it is determined that an event has occurred that satisfies the conditions on both the sending and receiving sides, i.e., "there has been an increase in traffic volume."

[0049] 13B corresponds to a situation in which the total number of arriving packets in either the observation period Tx or Ty on the sending or receiving side and in the common observation period Tc is equal to or exceeds the threshold value "H." Therefore, in normal detection pattern P12, it is determined that an event that satisfies the conditions has occurred on both the sending and receiving sides.

[0050] 13C corresponds to a situation where the total number of packets in the observation periods Tx and Tc on the sending side is less than the threshold value "H," and the total number of packets in the observation periods Ty and Tc on the receiving side is less than the threshold value "H." Therefore, in normal detection pattern P13, it is determined that no events satisfy the conditions on either the sending or receiving side.

[0051] The calculation process 30 calculates the occurrence probability of each of the three normal detection patterns P11, P12, and P13, and determines the probability that the sending side and receiving side will show the same event detection result. The details of this process are explained below.

[0052] -<Calculation of Probability of Normal Detection> The normal detection probability P0 at which the sending side and receiving side show the same event detection result is expressed as the sum of the probabilities P1, P2, and P3 of the three normal detection patterns P11, P12, and P13, as shown in the following equation (1).

[0053] A ｒｉ : the number of arriving packets in the observation section Tx from time t11 to t21 of the i-th period A ｃｉ : the number of arriving packets in the observation section Tc from time t21 to t12 of the i-th period A ｓｉ : The number of arriving packets in the observation section Ty from time t12 to t22 in the i-th period. Furthermore, the probabilities P1, P2, and P3 can be calculated using the following equations (2), (3), and (4), respectively.

[0054] Furthermore, assuming that the arrival packet generation process is a Poisson process, the probability that "a" packets of traffic with a packet arrival frequency λ per unit time arrive at the monitored link within a time length "t" can be calculated using the following equation (5).

[0055] The probability that "a" packets arrive in each observation interval is calculated using the following equation (6).

[0056] Furthermore, assuming that the time for data transfer and propagation delay between the sending and receiving sides is "Tr," the synchronization deviation between the sending and receiving sides regarding the timing of statistical data collection is "E," and the length of the traffic collection interval "I" is constant, the time deviation of the observation start packet, i.e., the length of the section observed by only one side, "D," can be calculated using the following equation (7).

[0057] <Configuration of the Device for Calculating the Normal Detection Probability P0> Fig. 14 shows an example configuration of an event-triggered detection device 200 that calculates the normal detection probability P0 at which the sending and receiving sides show the same event detection result. The event-triggered detection device 200 shown in Fig. 14 is connected to each of the two optical gateways 10 and 20 of the optical communication system 100B shown in Fig. 4B via an external interface 53. The event-triggered detection device 200 then obtains traffic aggregation results from each of the optical gateways 10 and 20 to calculate the normal detection probability P0. The calculated normal detection probability P0 can be used by network designers to correctly evaluate the impact of power waste when, for example, an autonomous distributed detection method is introduced.

[0058] 14 includes a calculation device 51, a storage medium 52, and an external interface 53. The calculation device 51 includes a success rate calculation function 51a that calculates the normal detection probability P0 using the method described above. This success rate calculation function 51a can be configured with computer hardware and a calculation program.

[0059] The storage medium 52 can store a calculation program 52a and calculation input data 52b. The area of ​​the calculation input data 52b can store and manage the data on the data transfer delay Pm11, synchronization deviation Pm12, traffic counting interval Pm2, average packet generation rate Pm3, and event trigger determination threshold "H" shown in Fig. 9. The latest value of the data on the average packet generation rate Pm3 of the monitored link can be periodically and repeatedly obtained from each of the optical gateways 10 and 20 via the external interface 53.

[0060] <Configuration of Event Trigger Detection Device> Fig. 15 shows a configuration example of an event trigger detection device 200 according to an embodiment of the present invention. In Fig. 15, an optical gateway 10, which is a first device, and an optical gateway 20, which is a second device, are connected by an optical network 15. Each of the optical gateways 10 and 20 can be configured as, for example, the above-mentioned optical transmission device 60. Therefore, for example, a packet transmitted as an optical signal by the optical gateway 10 is received by the optical gateway 20 via an optical path on an optical fiber 203.

[0061] The event trigger detection device 200 shown in Fig. 15 monitors the same traffic passing through the optical gateways 10 and 20, and is equipped with an event trigger identification unit 210 to detect a predetermined event trigger. The event trigger detection device 200 is connected to the outside of the optical gateways 10 and 20. Note that the functions of the event trigger identification unit 210 can be built into one or both of the optical gateways 10 and 20, or

[0062] This event trigger identification unit 210 utilizes optical gateways 10 and 20 through which the same traffic is transmitted or received, monitors the traffic on the optical gateway 10 side as a first traffic time series, monitors the traffic on the optical gateway 20 side as a second traffic time series, and has the function of aggregating the monitoring results of the first traffic time series and the second traffic time series at the same interval and identifying whether the aggregation results match predetermined conditions.

[0063] Specifically, the event trigger identification unit 210 includes a traffic designation unit 211, a counting timing synchronization unit 212, an event detection unit 213, a processing execution unit 214, an event occurrence condition database 215, an error minimization processing unit 216, and a normal detection probability calculation unit 217.

[0064] The traffic designation unit 211 designates specific traffic to be monitored that is common to the optical gateways 10 and 20. The counting timing synchronization unit 212 controls so as to synchronize the traffic counting timing of the optical gateway 10 with the traffic counting timing of the optical gateway 20. For example, by having the optical gateways 10 and 20 synchronize with each other with respect to a reference time and counting the traffic after synchronization, it becomes possible to reduce the difference in counting results that occurs between the optical gateways 10 and 20.

[0065] The event detection unit 213 detects the occurrence of an event from the results of counting the traffic of the optical gateways 10 and 20. For example, the occurrence of an event can be detected by comparing the average packet generation rate Pm3 counted at timings such as the observation periods T31 and T41 shown in FIG. 8 with the event trigger determination threshold "H."

[0066] When the traffic counting result matches a pre-registered event occurrence condition at each timing, the processing execution unit 214 executes the processing corresponding to the condition. For example, the processing execution unit 214 executes the following processes (1) to (4) according to each condition. (1) When both optical gateways 10 and 20 detect an increase in traffic volume above a predetermined level, a new optical path is added between the transmitter and receiver. (2) When both optical gateways 10 and 20 detect a decrease in traffic volume below a predetermined level, one of the optical paths connecting the transmitter and receiver is opened. (3) Similar to the operation shown in FIG. 29 (described later), the normal detection probability P0 is calculated, and depending on the result of comparing the calculated normal detection probability P0 with a threshold, the automatic distributed detection method and the centralized detection method are switched. (4) The normal detection probability P0 is calculated, and depending on the result of comparing the calculated normal detection probability P0 with a threshold, the normal detection probability P0 is notified to the system administrator.

[0067] The event occurrence condition database 215 holds predetermined data representing the occurrence conditions for each event as shown in (1) to (4) above. The error minimization processing unit 216 performs the processing shown in Fig. 21. That is, after time-synchronizing the timing of the optical gateways 10 and 20, it further readjusts the timing as shown in Fig. 20 to minimize the error between the statistical data compiled by the optical gateway 10 and the statistical data compiled by the optical gateway 20.

[0068] The normal detection probability calculation unit 217 calculates the normal detection probability P0 as a calculation result 39 based on the traffic count results of the optical gateways 10 and 20, similar to the calculation process 30 shown in FIG.

[0069] <Operation of the Device for Calculating the Normal Detection Probability P0> An overview of the operation of the event trigger detection device 200 shown in Fig. 14 is shown in Fig. 16. The arithmetic unit 51 of the event trigger detection device 200 executes a calculation program 52a stored in a storage medium 52, thereby realizing a success rate calculation unit 51a.

[0070] As shown in FIG. 17, the success rate calculation unit 51a inputs calculation input data 52b required for calculation in step S01, and assigns these data as parameters to be used in the calculation in step S02.

[0071] That is, the value "λ" corresponding to the average packet generation rate Pm3 for each link, the event trigger determination threshold "H", the value "I" of the traffic counting interval Pm2, and the length "D" of the section observed by only one of the sending and receiving sides, i.e., the sum of the data transfer delay Pm11 and the synchronization error Pm12, are identified. Note that the average packet generation rate Pm3 of the monitored link may be calculated using data acquired in real time from the optical gateways 10 and 20, or it may be calculated using statistical data that has been recorded and saved in the past.

[0072] The success rate calculation unit 51a uses the values ​​of the parameters identified in step S02 to perform calculations such as the above-mentioned equations (1) to (6) in step S03, and outputs the calculation results in step S04.

[0073] <Configuration of Autonomous Distributed Optical Transmission Device> An example configuration of an optical transmission device 60 that performs autonomous distributed event detection is shown in Fig. 17. For example, the optical transmission device 60 in Fig. 17 can be used as each of the optical gateways 10 and 20 in the optical communication system 100B shown in Fig. 4B.

[0074] 17, an optical transmission device 60 includes a calculation device 61, a storage medium 62, and an external interface 63. The calculation device 61 also includes a device management function 61a, an optical path setting event detection function 61b, an optical path setting function 61c, a packet forwarding device 61d, a packet statistics function 61e, and an optical transmission module 61f.

[0075] The storage medium 62 can hold and manage various function execution programs 62a, packet statistical data 62b, and an optical path setting event trigger database 62c.

[0076] <Start-up Operation of Autonomous Distributed Optical Transmission Device> Start-up operation of the optical transmission device 60 in Fig. 17 is shown in Fig. 18. The operation shown in Fig. 18 will be described below.

[0077] In step S11, the optical transmission devices 60 on the sending and receiving sides are connected by optical fibers so that optical communication can be performed between these devices. This operation may be performed manually or by switching.

[0078] In step S12, the administrator or the event trigger detection device 200 designates whether each of the two optical transmission devices 60 arranged as the transmitting side and the receiving side operates as the transmitting side or the receiving side.

[0079] In step S13, the administrator or the event trigger detection device 200 specifies, for each of the two optical transmission devices 60, the links to be monitored among the traffic passing between the two optical transmission devices 60 arranged as the transmitting and receiving sides.

[0080] In step S14, a period for acquiring statistical data in the link to be monitored is determined for the two optical transmission devices 60 arranged as the transmitting and receiving sides. This period is specified by an input operation by an administrator for each optical transmission device 60, or by an instruction from the event trigger detection device 200.

[0081] In step S15, synchronization processing is performed between the two optical transmission devices 60 arranged as the transmitting and receiving sides, and each optical transmission device 60 processes so that the timing of acquiring statistical data in the monitored link is synchronized between the transmitting and receiving sides.

[0082] In step S16, the two optical transmission devices 60 arranged as the transmitting and receiving sides acquire statistical data of the monitored link at timings synchronized between the transmitting and receiving sides. The operation of acquiring statistical data is repeatedly executed at the period determined in step S14.

[0083] <Optical Path Setting Event Detection Operation of Autonomous Distributed Optical Transmission Device> Fig. 19 shows the optical path setting event detection operation in the optical transmission device 60 in Fig. 17. The operation shown in Fig. 19 is described below. In step S21, in each optical transmission device 60 on the sending and receiving sides, the packet statistics function 61e acquires traffic statistical data of the monitored link at the statistical data acquisition timing. Specifically, in step S21, the average packet generation rate per period is repeatedly acquired at each timing of the period determined in step S14.

[0084] In step S22, each of the optical transmission devices 60 on the sending and receiving sides compares the traffic statistical data acquired in step S21 with its threshold. The thresholds to be compared are pre-registered in the optical path setting event trigger database 62c for each type of event, and therefore the optical transmission device 60 selects an appropriate threshold for each target event.

[0085] If the comparison result of step S22 indicates that the traffic statistical data matches the event trigger conditions, each optical transmission device 60 proceeds to processing from step S23 to S24 and identifies whether the type of matched event is "setting up (adding) an optical path" or "opening an optical path."

[0086] If the type of the matched event is "optical path setting," the optical path setting function 61c of the transmitting-side optical transmission device 60 processes to allocate a new optical path as the transmitting side in step S25. Also, the optical path setting function 61c of the receiving-side optical transmission device 60 processes to allocate a new optical path as the receiving side.

[0087] If the type of the matched event is "opening of an optical path," then in step S26, the optical path setting function 61c in each of the optical transmission devices 60 on the sending and receiving sides processes the optical path to be opened.

[0088] When each optical transmission device 60 on the transmitting and receiving sides performs the operation shown in Figure 19, for example, in the optical communication system 100 shown in Figure 1, it becomes possible to autonomously switch between using only one of the two optical paths 15a, 15b of the optical network 15 and using both, in accordance with the increase or decrease in the traffic volume actually observed.

[0089] <Re-adjusting the Counting Timing Using Packet Measurement Results> - <Example of Timing Change Before and After Readjustment> Figure 20 shows an example of timing change before and after readjustment of the counting timing. When the optical transmission devices 60 on the sending and receiving sides count traffic volume 73 at regular intervals, even after these devices are time-synchronized, there is an influence of the data transfer delay Pm11 and synchronization error Pm12 described above, resulting in a relative deviation in the counting timing on the sending and receiving sides, as shown in timing 71 before readjustment in Figure 20. However, by comparing the changes in the traffic counting results on the sending and receiving sides, it is possible to readjust the counting timing so that the deviation is reduced, as shown in timing 72 after readjustment in Figure 20.

[0090] In the example shown in Figure 20, the sender counting timing ts in pre-readjustment timing 71 has been corrected to the sender counting timing tsc in post-readjustment timing 72. Therefore, the difference between the corrected sender counting timing tsc and the receiver counting timing td is reduced in post-readjustment timing 72. Furthermore, on a time axis common to both the sender and receiver, the length of the observation period TB, during which the sender and receiver counting periods overlap, is increased to observation period TB2 in post-readjustment timing 72. Furthermore, the lengths of the observation periods TA and TC, during which only one of the sender and receiver counts, are reduced.

[0091] Therefore, by re-adjusting the timing, it is possible to synchronize the sender and receiver more precisely, which can lead to a significant improvement in the probability of correct detection, P, where the sender and receiver show the same event detection result.

[0092] -<Counting Timing Readjustment Operation> The counting timing readjustment operation in this embodiment is shown in Fig. 21. A part or all of the functions of the "timing readjustment unit" that performs this operation can be located and executed in, for example, at least one of the optical transmission devices 60 on the sending side and the receiving side, or in the event trigger detection device 200. The operation shown in Fig. 21 will be described below.

[0093] After detecting in step S31 that initial time synchronization has been completed between the optical transmission devices 60 on the transmitting and receiving sides, the "timing readjustment unit" proceeds to step S32 and subsequent steps. Then, in step S32, traffic statistical data is acquired on both the transmitting and receiving sides for a plurality of observation interval candidates whose timings are shifted by an integer multiple of the infinitesimal time "δ" from the initial counting timing that serves as the reference. The time width by which the timing is shifted in step S32 is determined as "-aδ" and "aδ" using a plurality of integer constants "a: 1, 2, 3, ...".

[0094] An example of the tallying timing when statistical data is acquired with a time difference of δ is shown in Fig. 22. In the example shown in Fig. 22, the sending-side tallying timings ts1 and ts2, which serve as the reference before the timing shift, are shifted by time differences of "-δ" and "δ", respectively, and the results are used as candidate sending-side tallying timings ts11, ts12, ts21, and ts22. The receiving-side timing may be adjusted in the same way as in Fig. 22.

[0095] Furthermore, the "timing readjustment unit" repeatedly acquires statistical data multiple times with the timings shifted from each other in step S33 of Fig. 21. This statistical data acquisition is performed for both the transmitting side and the receiving side.

[0096] An example of the timing of aggregation when statistical data for multiple periods is acquired at different times is shown in Fig. 23. Also, an example of the structure of statistical data for multiple periods acquired at different times is shown in Fig. 24.

[0097] As shown in FIG. 24, even when the sending side collects statistical data at the same sending side collection timing ts, the first, second, and third collections collect statistical data A of traffic volume for parts 731, 732, and 733 of different traffic volumes 73. ｓ，０，１ , A ｓ，０，２ , A ｓ，０，３ In addition, by aggregating statistical data at timings shifted by the time differences "-δ" and "δ", the n-th statistical data A of the traffic volume for the traffic volumes 73 of different parts can be obtained. ｓ，－δ，ｎ , A ｓ，０，ｎ , A ｓ，＋δ，ｎ can be obtained respectively.

[0098] Similarly to the above, even when the receiving side tally the statistical data at the same receiving side td, the first, second and third tallying times are used to compile the statistical data A of the traffic volume for the portions 741, 742 and 743 of the traffic volume 73, which are different from each other. ｄ，０，１ , A ｄ，０，２ , A ｄ，０，３In addition, by aggregating statistical data at timings shifted forward and backward by the time differences "-δ" and "δ", the n-th statistical data A of the traffic volume for the traffic volumes 73 of different parts can be obtained. ｄ，－δ，ｎ , A ｄ，０，ｎ , A ｄ，＋δ，ｎ can be obtained respectively.

[0099] Therefore, as shown in FIG. 24, the sending side obtains the statistical data DSi of the traffic volume collected for the i-th time as the statistical data A ｓ，－δ，１ , A ｓ，－δ，２ , A ｓ，－δ，３ , A ｓ，０，１ , A ｓ，０，２ , A ｓ，０，３ , A ｓ，＋δ，１ , A ｓ，＋δ，２ , A ｓ，＋δ，３ The receiving side can obtain the statistical data A as the statistical data DDi of the traffic volume collected for the i-th time. ｄ，－δ，１ , A ｄ，－δ，２ , A ｄ，－δ，３ , A ｄ，０，１ , A ｄ，０，２ , A ｄ，０，３ , A ｄ，＋δ，１ , A ｄ，＋δ，２ , A ｄ，＋δ，３ can be obtained respectively.

[0100] In other words, by repeating steps S32 to S34 shown in Fig. 21, it is possible to aggregate traffic volume 73 while fine-tuning the relative timing between the sender and receiver, as shown in Fig. 22. Furthermore, by collecting statistical data multiple times as shown in Fig. 23, it is possible to obtain data such as that shown in Fig. 24, which allows one to grasp the differences in the aggregation results for various parts of traffic volume 73 for each aggregation time window.

[0101] When statistical data acquisition for all candidate counting windows with different timings has been completed, the timing readjustment unit proceeds from steps S34 to S35. In step S35, the timing readjustment unit compares the acquired data from the sending and receiving sides for each of the multiple candidates, and relatively readjusts the receiving-side counting timing ts and the sending-side counting timing td for the observation window so that the counting results are closer together. This enables traffic volume 73 to be counted at more appropriate timing.

[0102] The details of the process of step S35 in Fig. 21 are explained below. The "timing readjustment unit" shares the statistical data DSi observed and compiled on the transmitting side and the statistical data DDi observed and compiled on the receiving side. Then, these statistical data are compared, and the timings at which these compilation results are close are set as the new receiving-side compilation timing ts and the transmitting-side compilation timing td on the transmitting side and the receiving side. In this embodiment, the least squares method is applied to compare the transmitting-side statistical data DSi with the receiving-side statistical data DDi and to determine the closeness of these compilation results.

[0103] The comparison targets are three different cases: (1), (2), and (3). (1) When the observation timing on the sending side is delayed by "δ", it approaches the timing on the receiving side. (2) When the observation timing on the sending side is advanced by "δ", it approaches the timing on the receiving side. (3) When the observation timing on the sending side and the receiving side are almost synchronized.

[0104] In the above cases (1), (2), and (3), the relationships expressed by the following formulas (8), (9), and (10) are established, respectively.

[0105] In the case of the above case (1), the relationship of the above equation (8) is applied to the following equation (11), and the square error ε ｓ，＋δ In this formula (11), it is assumed that the number of counts is three. Here, the statistical data shown on the upper and lower sides of the formula (8) above are assigned to the arguments "a" and "b" shown in formula (11), respectively. Furthermore, the magnitude of other squared errors ε ｓ，－δ , ε ｓ，０ is calculated in the same way as above.

[0106] In the case of the above case (2), the relationship of the above equation (9) is applied to the above equation (11), and the square error ε ｓ，＋δ Here, the statistical data shown in the upper and lower parts of the above formula (9) are assigned to the arguments "a" and "b" shown in formula (11), respectively.ｓ，－δ , ε ｓ，０ is calculated in the same way as above.

[0107] In the case of the above case (3), the relationship of the above formula (10) is applied to the above formula (11), and the square error ε ｓ，＋δ Calculate the magnitude of the other squared error ε ｓ，－δ , ε ｓ，０ is calculated in the same way as above.

[0108] - <Specific Example of Counting Timing Readjustment> (1) An example of "Minimum Error Pattern 1" when the "timing readjustment unit" readjusts the counting timing is shown in Fig. 25. In the case of "Minimum Error Pattern 1" shown in Fig. 25, the time difference between the sending-side counting timing ts1, ts2 and the receiving-side counting timing td1, td2 before timing readjustment is almost zero, so the error is minimized as it is. Therefore, even after timing readjustment, resynchronization is performed so as to maintain the same timing as before adjustment.

[0109] (2) FIG. 26 shows an example of "minimum error pattern 2" when the "timing readjustment unit" readjusts the tally timing. In the case of "minimum error pattern 2" shown in FIG. 26, if the transmitting-side tally timings ts1 and ts2 before timing readjustment are advanced by "δ," the time difference between the transmitting-side tally timings ts1 and ts2 and the receiving-side tally timings td1 and td2 becomes almost zero. Therefore, by timing readjustment, the transmitting-side tally timings ts1 and ts2 are advanced by "δ," and the transmitting and receiving sides are resynchronized with the new transmitting-side tally timings ts1c and ts2c. As a result, the time difference between the transmitting-side tally timings ts1c and ts2c after resynchronization and the receiving-side tally timings td1 and td2 becomes almost zero. Of course, the transmitting-side timing may be fixed and the receiving-side timing may be readjusted.

[0110] (3) FIG. 27 shows an example of "minimum error pattern 3" when the "timing readjustment unit" readjusts the tally timing. In the case of "minimum error pattern 3" shown in FIG. 27, if the transmitting-side tally timings ts1 and ts2 before timing readjustment are delayed by "δ," the time difference between the transmitting-side tally timings ts1 and ts2 and the receiving-side tally timings td1 and td2 becomes almost zero. Therefore, the transmitting-side tally timings ts1 and ts2 are delayed by "δ" through timing readjustment, and the transmitting and receiving sides are resynchronized with the new transmitting-side tally timings ts1c and ts2c. As a result, the time difference between the transmitting-side tally timings ts1c and ts2c after resynchronization and the receiving-side tally timings td1 and td2 becomes almost zero. Of course, the transmitting-side timing may be fixed and the receiving-side timing may be readjusted.

[0111] -<Method for reducing the amount of calculation> In the above description, the "timing readjustment unit" calculates the squared errors ε ｓ，＋δ，ｉ，ａ , ε ｓ，＋δ，ｉ，ｂ However, if it can be guaranteed that these multiple squared errors are almost the same, the squared error ε ｓ，＋δ，ｉ，ａ By calculating only one of the two, the amount of calculation and the calculation time can be reduced.

[0112] -<Method of determining the time difference δ> The value of the time difference δ is determined based on statistical data of the synchronization error. For example, a method is envisaged in which half the value of the synchronization error calculated from the statistical data is adopted as the time difference δ.

[0113] An example of a device configuration that can be used to acquire statistical data related to the above-mentioned synchronization error is shown in Fig. 28. In the example shown in Fig. 28, a transmitting port 80a and a receiving port 80b of one error measurement device 80 are connected by an optical fiber 82 of a certain length. The length of this optical fiber 82 is set to, for example, exactly 100 km.

[0114] That is, the error measurement device 80 can send an optical signal from a transmitting port 80a, and receive the optical signal at a receiving port 80b after it has passed through a path that is folded back on the optical fiber 82. By using a local clock to time-synchronize the timing of the transmitting port 80a and the receiving port 80b, to which both ends of the optical fiber 82 are connected, the error measurement device 80 can measure the time difference between transmitting an optical signal and receiving the optical signal that has passed through the optical fiber 82. Furthermore, by comparing the time of the local clock with the time of a global clock 81 managed by the error measurement device 80, statistical data on the synchronization error that occurs when an optical signal is transmitted over a distance of 100 km can be obtained.

[0115] <Example of control reflecting the calculation result of the trigger detection success rate> An example of system operation when the calculation result of the trigger detection success rate is reflected in control is shown in Fig. 29. For example, when each optical gateway 10, 20 employs an autonomous distributed detection method, as in the optical communication system 100B shown in Fig. 4B, the above-mentioned event trigger detection device 200 can be used to obtain the normal detection probability P0 (equivalent to the trigger detection success rate) at which the sending side and receiving side show the same event detection result.

[0116] Furthermore, in an optical communication system 100B such as that shown in FIG. 4B, if the normal detection probability P0 is small, there is a high probability that the event detection results on the sending side and the receiving side will not match, which can easily lead to situations such as packet loss and unnecessary power consumption, as in the optical communication system 100C shown in FIG. 4C.

[0117] Therefore, in this embodiment, the system is configured so that the optical transmission device 60 of each optical gateway 10, 20 can selectively use both the autonomous distributed detection method and the centralized management (message transmission) detection method. The system then autonomously switches between the autonomous distributed detection method and the centralized management detection method depending on the situation. Specifically, the autonomous distributed detection method is adopted in the initial state, and when it is detected that the normal detection probability P0 is small, the system switches to the centralized management detection method.

[0118] The operation example shown in Fig. 29 will be described below. In this embodiment, similar to the optical communication system 100B in Fig. 4B, for example, optical gateways 10 and 20 are provided on the transmitting and receiving sides, respectively. Furthermore, each optical transmission device 60 provided in each optical gateway 10 and 20 is connected to an external event trigger detection device 200 via a control communication line, and the event trigger detection device 200 controls each optical transmission device 60 on the transmitting and receiving sides.

[0119] The threshold value of the normal detection probability P0 required for determining whether to switch control is registered in advance in the event trigger detection device 200. In step S41, the event trigger detection device 200 acquires statistical traffic data from the transmitting and receiving optical transmission devices 60. In step S42, the event trigger detection device 200 calculates the above-mentioned normal detection probability P0 based on the acquired statistical data.

[0120] The event trigger detection device 200 compares the calculated normal detection probability P0 with a predetermined threshold in step S43. If the normal detection probability P0 is equal to or greater than the threshold, the process proceeds from step S43 to S45, and if the normal detection probability P0 is less than the threshold, the process proceeds from step S43 to S44.

[0121] In step S45, the event trigger detection device 200 controls the transmitting and receiving optical transmission devices 60 so that they each employ an autonomous distributed detection method to detect an event trigger. The event trigger detection device 200 also instructs the transmitting and receiving devices to synchronize the link to be monitored, the traffic counting interval "I," and the counting timing.

[0122] In step S44, the event trigger detection device 200 controls the optical transmission devices 60 on the sending and receiving sides to detect event triggers using a centralized detection method. The event trigger detection device 200 also selects either the sending or receiving side as an event detection device, and notifies that event detection device of the link to be monitored, the traffic counting interval "I," and information necessary for notification to the opposing device. The information necessary for notification to the opposing device includes the port number of the opposing side's optical path setting function, etc.

[0123] <Characteristics of the Autonomous Distributed Event Trigger Detection Device> The following [1] to [4] list the characteristic features of the autonomous distributed event trigger detection device of the present invention. [1] A first device (optical gateway 10) and a second device (optical gateway 20) that transmit or receive the same traffic are used, and the first device monitors the traffic as a first traffic time series, and the second device monitors the traffic as a second traffic time series, and the monitoring results of the first traffic time series and the second traffic time series are collected at the same period, and an event trigger identification unit (210) is provided that identifies whether the collected results match a predetermined condition, wherein the event trigger identification unit (210) comprises: a traffic designation unit (211) that designates traffic to be monitored by the first device and the second device; a collection timing synchronization unit (212) that synchronizes the traffic collection timing of the first device and the traffic collection timing of the second device; and an event detection unit (213) that detects an event occurrence from the results of the collection at the traffic collection timing of the first device and the second device, An autonomous distributed event trigger detection device comprising: a processing execution unit (214) that executes processing corresponding to a condition when a traffic aggregation result matches a pre-registered event occurrence condition at each timing.

[0124] According to the autonomous distributed event trigger detection device having the configuration [1] above, when the first device and the second device perform event detection in an autonomous distributed manner, it is possible to perform appropriate processing while taking into consideration the possibility that a timing discrepancy may cause a difference between the traffic count results on the first device side and the traffic count results on the second device side. For example, when adding a new optical path between the first device and the second device, it is possible to prevent a situation in which only one device completes the optical path addition while the other device does not. In other words, it is possible to prevent packet loss and unnecessary power consumption at the interfaces of each device.

[0125] [2] The event trigger identification unit prepares a plurality of aggregation timing time series that have the same period but are offset in timing from each other, aggregates traffic for each of the plurality of aggregation timing time series, shares the aggregation results of the plurality of aggregation timing time series between the first device and the second device, calculates the error between the aggregation results of the first device and the second device from the shared aggregation results, and uses the result of aggregating traffic for a specific aggregation timing time series that minimizes the error among the plurality of aggregation timing time series (step S35), in the autonomous distributed event trigger detection device described in [1] above.

[0126] According to the autonomous distributed event trigger detection device having the configuration [2] above, the occurrence of an event can be detected after the difference in the counting timing between the first device and the second device is readjusted to a small value, for example, as shown in FIG. 20. Therefore, it is expected that the normal detection probability P0 will be significantly improved compared to when only time synchronization is performed between the first device and the second device. Therefore, when processing an optical path setting event or the like is performed, it is possible to suppress the occurrence of packet loss and unnecessary power consumption at the interface of each device.

[0127] [3] The autonomous distributed event trigger detection device according to [1] above, wherein the event trigger identification unit has a function (calculation process 30) of, after executing synchronization processing between the first device and the second device, estimating a difference between the traffic used for aggregation in the first device and the traffic used for aggregation in the second device based on the time difference between the closest aggregation timings in time between the first device and the second device and the data transmission and propagation delay times between the first device and the second device, and calculating (steps S01 to S04) the probability that the first device and the second device will return the same condition judgment result for the monitored time series (normal detection probability P0) using the estimated traffic difference, the traffic aggregation period, and statistical parameters related to the traffic as input.

[0128] According to the autonomous distributed event trigger detection device having the configuration [3] above, a network designer or the like can grasp the probability that the first device and the second device will return the same condition judgment result for the monitored time series. For example, when an evaluation index is needed to select between a message detection method and an autonomous distributed method as a means of sharing event occurrence information between two optical gateways, the probability calculated by the event trigger identification unit can be used for evaluation.

[0129] [4] The autonomous distributed event trigger detection device according to [3] above, wherein the event trigger identification unit has the function of calculating, based on a difference in counting timing between the first device and the second device (length D of the section observed by only one of them) and the traffic monitoring cycle (traffic counting interval I), a first time section (time t21 to t12) in which traffic observed by both the first device and the second device exists, and a second observation section (time t11 to t21, time t12 to t22) in which traffic observed by only one of the first device and the second device exists, of the entire monitoring time section used for condition judgment, calculating an occurrence probability for each amount of traffic occurring in the first time section and the second observation section, and calculating a probability that the first device and the second device will return the same condition judgment result for a monitored time series.

[0130] According to the autonomous distributed event trigger detection device having the configuration [4] above, the probability is calculated based on a model that assumes that there is a time lag in the aggregation timing between the first device and the second device, so that probability information that is useful as an evaluation index when an autonomous distributed system is introduced into a network can be obtained.

[0131] REFERENCE SIGNS LIST 10 Optical gateway (first device) 20 Optical gateway (second device) 11, 12, 21, 22 Optical transceiver 15 Optical network 15a, 15b Optical path 30 Calculation processing 39 Calculation result 51 Arithmetic unit 51a Success rate calculation unit 52 Storage medium 52a Calculation program 52b Calculation input data 53 External interface 60 Optical transmission device 61 Arithmetic unit 61a Device management function 61b Optical path setting event detection function 61c Optical path setting function 61d Packet forwarding device 61e Packet statistics function 61f Optical transmission module 62 Storage medium 62a Various function execution programs 62b Packet statistics data 62c Optical path setting event trigger database 63 External interface 71 Timing before readjustment 72 Timing after readjustment 73 Traffic volume 80 Error measurement device 80a: Sending port 80b: Receiving port 81: Global clock 82: Optical fiber 100, 100A, 100B, 100C: Optical communication system 200: Event trigger detection device (autonomous distributed event trigger detection device) 203: Optical fiber 210: Event trigger identification unit 211: Traffic designation unit 212: Counting timing synchronization unit 213: Event detection unit 214: Processing execution unit 215: Event occurrence condition database 216: Error minimization processing unit 217: Normal detection probability calculation unit Ari, Aci, Asi: Number of arriving packets D: Length of section observed by only one side I: Traffic counting interval H: Event trigger determination threshold P0: Normal detection probability P11, P12, P13: Normal detection pattern Pm: Parameter Pm11: Data transfer delay Pm12: Synchronization deviation Pm2: Counting interval Pm3 Average packet generation rate Px, Py, Pz Observed packet time series td Receiver count timing ts Sender count timing tx, ty, tz Time axis T01, T02 Observation period T31, T41 Observation interval TA, TB, TC Observation interval Tc, Tx, Ty Observation interval Te Timing offset λ Average packet generation rate

Claims

1. An autonomous distributed event trigger detection device that utilizes a first device and a second device that transmit or receive the same traffic, monitors the traffic on the first device side as a first traffic time series, monitors the traffic on the second device side as a second traffic time series, and has an event trigger identification unit that aggregates the monitoring results of the first traffic time series and the second traffic time series at the same period and identifies whether the aggregation results match a pre-set condition, wherein the event trigger identification unit comprises: a traffic designation unit that designates traffic to be monitored by the first device and the second device; an aggregation timing synchronization unit that synchronizes the traffic aggregation timing of the first device and the traffic aggregation timing of the second device; an event detection unit that detects the occurrence of an event from the aggregation results at the traffic aggregation timing of the first device and the second device; and a process execution unit that executes a process corresponding to the condition when the traffic aggregation results match a pre-registered event occurrence condition at each timing.

2. The autonomous distributed event trigger detection device of claim 1, wherein the event trigger identification unit prepares a plurality of aggregation timing time series that have the same period but are offset in timing from each other, aggregates traffic for each of the plurality of aggregation timing time series, shares the aggregation results of the plurality of aggregation timing time series between the first device and the second device, calculates the error between the aggregation results of the first device and the second device from the shared aggregation results, and uses the result of aggregating traffic for a specific aggregation timing time series that minimizes the error among the plurality of aggregation timing time series.

3. The autonomous distributed event trigger detection device according to claim 1, wherein the event trigger identification unit has the function of, after executing synchronization processing between the first device and the second device, estimating the difference between the traffic used for aggregation by the first device and the traffic used for aggregation by the second device based on the time difference between the closest aggregation timings in time between the first device and the second device and the data transmission and propagation delay times between the first device and the second device, and calculating the probability that the first device and the second device will return the same condition judgment result for the monitored time series using the estimated traffic difference, the traffic aggregation period, and statistical parameters related to the traffic as input.

4. The autonomous distributed event trigger detection device according to claim 3, wherein the event trigger identification unit has the function of calculating, based on the difference in aggregation timing between the first device and the second device and the traffic monitoring period, a first time interval in which traffic observed by both the first device and the second device exists, and a second observation interval in which traffic observed by only one of the first device and the second device exists, out of the entire monitoring time interval used for condition judgment, calculating the occurrence probability for each amount of traffic occurring in the first time interval and the second observation interval, and calculating the probability that the first device and the second device will return the same condition judgment result for the monitored time series.

Images