Monitoring system, monitoring method, and program
Patent Information
- Application Number
- JP2025523236
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-07-26
- Filing Date
- 2023-09-04
- Publication Date
- 2024-12-05
- Estimated Expiration
- 2043-09-04
AI Technical Summary
Conventional monitoring systems struggle to identify which status data is related to an abnormality among various status data regarding a device's status, making it difficult to detect, restore, or analyze anomalies in services effectively.
A monitoring system that includes a status data acquisition unit to gather status data and a relationship determination unit to identify which data is related to an abnormality by analyzing time-series changes and saturation periods, enabling quicker detection and recovery from anomalies.
The system accurately determines which status data is related to an abnormality, facilitating faster anomaly detection and recovery processes, even when multiple devices are involved in the abnormality.
Abstract
Description
Monitoring system, monitoring method, and program
[0001] The present disclosure relates to a monitoring system, a monitoring method, and a program.
[0002] Conventionally, status data regarding the status of monitored devices has been analyzed for various purposes, such as detecting, recovering from, or analyzing anomalies in services. For example, Patent Document 1 describes a system that determines whether or not a device has an anomaly based on a model that learns the relationship between time-series changes in the status data of a device and the actual operating status of the device. Patent Document 2 describes a system that estimates whether or not future memory usage of the device will exceed a threshold based on the device's past memory usage history and the device's current memory usage. For example, Patent Document 3 describes identifying an effective recovery procedure for recovering from a system failure (an example of an anomaly) based on a computational model that calculates the correlation between status data regarding the status of the device in the system and a recovery procedure for recovering from the system failure.
[0003] International Publication No. 2017 / 169949 Special Publication No. 2022-511938 Japanese Patent Application Laid-Open No. 2021-015321
[0004] In the above-described technology, it is required to execute processing for abnormalities in the service based on status data relating to the status of devices that are the subject of monitoring in the service.
[0005] One of the objectives of the present disclosure is to execute processing related to an abnormality in a service based on status data related to the status of a device that is a monitoring target in the service.
[0006] The monitoring system according to the present disclosure includes a status data acquisition unit that acquires status data relating to the status of a device to be monitored in a service, and executes processing relating to an abnormality in the service based on the status data.
[0007] According to the present disclosure, processing related to an abnormality in a service can be executed based on status data related to the status of a device that is a monitoring target in the service.
[0008] 1 is a diagram illustrating an example of the hardware configuration of a monitoring system. FIG. 1 is a diagram illustrating an example of changes in the communication volume of a communication device over time when no abnormality occurs in the communication service. FIG. 2 is a diagram illustrating an example of changes in the communication volume of a communication device over time when an abnormality is detected in the communication service. FIG. 3 is a diagram illustrating an example of a notification displayed on an administrator terminal. FIG. 4 is a diagram illustrating an example of functions implemented in the monitoring system. FIG. 5 is a diagram illustrating an example of a status database. FIG. 6 is a diagram illustrating an example of a period set by a period setting unit. FIG. 7 is a diagram illustrating an example of processing executed in the monitoring system. FIG. 8 is a diagram illustrating an example of functions implemented in the monitoring system of Modification 1-1. FIG. 9 is a diagram illustrating an example of changes in the communication volume of a communication device over time when an abnormality is detected in the communication service. FIG. 10 is a diagram illustrating an example of a notification displayed on an administrator terminal. FIG. 11 is a diagram illustrating an example of functions implemented in the monitoring system. FIG. 12 is a diagram illustrating an example of a status database. FIG. 13 is a diagram illustrating an example of a result of execution of clustering. FIG. 14 is a diagram illustrating an example of processing executed in the monitoring system.
[0009] [1. First Embodiment] A first embodiment, which is an example of an embodiment of a monitoring system according to the present disclosure, will be described. In the first embodiment, an example of processing related to an abnormality in a service will be described. In conventional technology, various status data exist for various devices. Therefore, if a service provider can identify status data related to an abnormality in the service, this is very useful because it can be used for various purposes such as detecting, recovering from, or analyzing the abnormality in the service. However, conventional technology cannot identify which status data from various status data related to the device status is related to the abnormality. Therefore, the monitoring system of the first embodiment determines whether status data is related to the abnormality in the service.
[0010] [1-1. Hardware configuration of a monitoring system] Fig. 1 is a diagram showing an example of the hardware configuration of a monitoring system. For example, the monitoring system 1 includes a server 10, a communication device 20, a user terminal 30, and an administrator terminal 40. Each of the server 10, the communication device 20, the user terminal 30, and the administrator terminal 40 can be connected to a network N. For example, the network N is the Internet, a public communication line, or a LAN.
[0011] The server 10 is a server computer. In this embodiment, a case where the server 10 is managed by a telecommunications carrier is taken as an example. For example, the server 10 includes a control unit 11, a storage unit 12, and a communication unit 13. The control unit 11 includes at least one processor. The storage unit 12 includes at least one of a volatile memory such as RAM and a non-volatile memory such as flash memory. The communication unit 13 includes at least one of a communication interface for wired communication and a communication interface for wireless communication.
[0012] The communication device 20 is a device that communicates with other devices. For example, the communication device 20 is a device managed by a telecommunications carrier. The communication device 20 may be a device for wireless communication or a device for wired communication. The communication device 20 relays at least one of user calls and data communications. For example, the communication device 20 is a base station of a public communication line, a private branch exchange, a wireless LAN access point, a router, a hub, a repeater, a modem, or a server computer used in virtualization technology. The communication device 20 may be a device of a telecommunications carrier that provides users with a fully virtualized cloud-native mobile network.
[0013] The user terminal 30 is a computer of a user who uses the communication service. For example, the user terminal 30 is a smartphone, a tablet, a personal computer, or a wearable terminal. For example, the user terminal 30 includes a control unit 31, a memory unit 32, a communication unit 33, an operation unit 34, and a display unit 35. The hardware configurations of the control unit 31, the memory unit 32, and the communication unit 33 may be similar to those of the control unit 11, the memory unit 12, and the communication unit 13, respectively. The operation unit 34 is an input device such as a touch panel or a mouse. The display unit 35 is a display such as a liquid crystal or organic EL display.
[0014] The administrator terminal 40 is a computer of an administrator in the communication service. For example, the administrator terminal 40 is a smartphone, a tablet, a personal computer, or a wearable terminal. For example, the administrator terminal 40 includes a control unit 41, a storage unit 42, a communication unit 43, an operation unit 44, and a display unit 45. The hardware configurations of the control unit 41, the storage unit 42, the communication unit 43, the operation unit 44, and the display unit 45 may be similar to those of the control unit 11, the storage unit 12, the communication unit 13, the operation unit 34, and the display unit 35, respectively.
[0015] The programs stored in the storage units 12, 32, and 42 may be supplied via the network N. Each computer may also include at least one of a reading unit (e.g., a memory card slot) that reads a computer-readable information storage medium and an input / output unit (e.g., a USB port) that inputs and outputs data to and from an external device. For example, a program stored in an information storage medium may be supplied via at least one of the reading unit and the input / output unit.
[0016] Furthermore, the monitoring system 1 is not limited to the example of FIG. 1 as long as it includes at least one computer. For example, the monitoring system 1 does not need to include at least one of the communication device 20, the user terminal 30, and the administrator terminal 40. In this case, the at least one exists outside the monitoring system 1. The monitoring system 1 may include only the server 10. The monitoring system 1 may include the server 10 and another computer not shown in FIG. 1. The monitoring system 1 may include only the administrator terminal 40. The monitoring system 1 may include the administrator terminal 40 and another computer not shown in FIG. 1.
[0017] [1-2. Overview of Monitoring System] In this embodiment, an example is given in which the communication device 20 is a server computer having a containerized network function (CNF). Furthermore, an example is given in which the user terminal 30 is a smartphone. The communication device 20 relays communications between a large number of user terminals 30. For example, if the amount of communication that the communication device 20 must process reaches the amount of communication that the communication device 20 can process, an abnormality may occur in the communication service. The amount of communication can be expressed by a known index such as bps.
[0018] For example, when an abnormality occurs in a communication service, an administrator identifies the communication device 20 that caused the abnormality and performs recovery work. However, since a large number of communication devices 20 exist in a communication service and various abnormalities can occur, it can be difficult for an administrator to identify the communication device 20 that caused the abnormality. In this case, the communication device 20 that caused the abnormality cannot process any more communications, and the communication volume may become saturated.
[0019] Saturation of communication volume means that the communication volume is positive (not zero) and does not change or changes almost unchanged. No change in communication volume means that the change in communication volume is zero. Almost no change in communication volume means that the change in communication volume is less than a threshold. Specific methods for determining saturation will be described later. Although an upper limit may be set for communication volume, there are cases where the communication device 20 can handle a communication volume greater than the upper limit set by the administrator, or where the communication device 20 can only handle a communication volume lower than the upper limit set by the administrator, and further, where the administrator himself does not know the upper limit of the communication device 20. Therefore, in this embodiment, no upper limit is set for communication volume. Therefore, even if the communication volume of the communication device 20 reaches a certain level, an abnormality is not detected.
[0020] FIG. 2 is a diagram showing an example of changes in communication volume over time for a communication device 20 when no abnormality occurs in the communication service. The horizontal axis of FIG. 2 represents time. The vertical axis of FIG. 2 represents communication volume. In the example of FIG. 2, changes in communication volume for each of three communication devices 20A to 20C located at a telecommunications carrier's facility are shown. Hereinafter, when there is no need to distinguish between the communication devices 20A to 20C, they will simply be referred to as communication device 20. As shown in FIG. 2, when no abnormality occurs in the communication service, the communication volume of each communication device 20 is not saturated. In other words, the communication volume of each communication device 20 is, in principle, constantly changing.
[0021] 3 is a diagram showing an example of a change in communication volume over time for a communication device 20 when an abnormality is detected in a communication service. In the example of FIG. 3, the communication volume for communication device 20C is saturated during the period from time t00 to time t01. During this period, communication device 20C cannot process any more communications, so a user terminal 30 attempting to connect to communication device 20C in a specific area may be unable to use the communication service. In such a case, the user may contact the call center of the communication service and inform them that the communication service is unavailable.
[0022] For example, an administrator receives inquiries from multiple users and becomes aware that an abnormality has occurred. At this point, the administrator has not yet identified which communication device 20 is causing the abnormality. Then, at time t01, the communication volume of communication device 20C decreases, and the abnormality resolves naturally. Furthermore, the administrator determines that the abnormality has been resolved when inquiries from users cease to be received. Since the user terminal 30 becomes able to use the communication service at time t01, there is a correlation between the period during which the abnormality occurred and the communication volume of communication device 20C. Note that in this embodiment, the period from the saturation start timing to the saturation end timing is treated as an example of a period during which an abnormality has occurred. However, the period from the saturation start timing to the present time, when saturation has not yet ended, may also be treated as an example of a period during which an abnormality has occurred. For example, the period from the saturation start timing to a time specified by the administrator may also be treated as an example of a period during which an abnormality has occurred.
[0023] For example, after an abnormality occurs and is resolved, the server 10 analyzes the traffic volume of each of the communication devices 20A to 20C. The server 10 determines that the traffic volume of the communication device 20C was saturated during the period when the abnormality occurred. The server 10 stores the characteristics of the abnormality that occurred this time and the fact that the traffic volume of the communication device 20C is related to the abnormality, in association with each other. If a similar abnormality occurs again, the server 10 sends a notification to the administrator terminal 40 urging the administrator terminal 40 to check the status of the communication device 20C.
[0024] 4 is a diagram showing an example of a notification displayed on the administrator terminal 40. For example, when the administrator starts a maintenance tool installed on the administrator terminal 40, an administrator screen SC showing the notification received from the server 10 is displayed on the display unit 45. The administrator screen SC displays a message indicating that the communication volume of communication device 20C was saturated when a similar abnormality was detected in the past, and therefore the status of communication device 20C should be checked again this time.
[0025] For example, suppose that in the past, when the communication volume of communication device 20C was saturated, a call center for a communication service received an inquiry from a user who tried to use the communication service from a specific area X. If the call center currently receives an inquiry from a user who tried to use the communication service from the same area X, it will detect that an abnormality has occurred in the communication service in area X. In this case, server 10 transmits display data for an administrator screen SC to administrator terminal 40, prompting the administrator terminal 40 to check the status of communication device 20C.
[0026] For example, when the administrator terminal 40 receives display data for the administrator screen SC, it causes the display unit 45 to display the administrator screen SC including a message prompting the administrator to check the status of the communication device 20C. The administrator screen SC may display a graph showing the current communication volume of the communication device 20C, or may display a graph showing the communication volume of the communication device 20C when a similar abnormality occurred in the past. The administrator checks the administrator screen SC and performs maintenance on the communication device 20C. The maintenance itself of the communication device 20C may be performed by a known method.
[0027] As described above, when an abnormality occurs in a communication service and is subsequently recovered, the monitoring system 1 determines whether the communication volume of the communication device 20 has been saturated. Based on the determination result of whether the communication volume of the communication device 20 has been saturated, the monitoring system 1 determines whether the communication volume of the communication device 20 is related to the abnormality in the communication service. For example, when a similar abnormality occurs again, the administrator can easily identify the communication device 20 that caused the abnormality, enabling rapid recovery work. Details of this embodiment will be described below.
[0028] [1-3. Functions Realized in the Monitoring System] Fig. 5 is a diagram showing an example of functions realized in the monitoring system 1. Fig. 5 shows functions realized in the server 10 and functions realized in the administrator terminal 40. The functions of the communication device 20 and the user terminal 30 are similar to those of known communication services, and are therefore omitted from Fig. 5. For example, the communication device 20 has a function to transmit status data or part of the status data to the server 10, which will be described later. The user terminal 30 has a function that enables the user to use the communication service.
[0029] [1-3-1. Functions Realized by the Server] For example, the server 10 includes a data storage unit 100, a status data acquisition unit 101, a period setting unit 102, a normalization execution unit 103, a saturation determination unit 104, a relationship determination unit 105, and a recovery process execution unit 106. The data storage unit 100 is realized by the storage unit 12. The status data acquisition unit 101, the period setting unit 102, the normalization execution unit 103, the saturation determination unit 104, the relationship determination unit 105, and the recovery process execution unit 106 are realized by the control unit 11.
[0030] [Data Storage Unit] The data storage unit 100 stores data related to communication services. For example, the data storage unit 100 stores a status database DB.
[0031] 6 is a diagram showing an example of a state database DB. The state database DB is a database in which state data for each of a plurality of communication devices 20 is stored. For example, the state database DB stores identification information for each of the plurality of communication devices 20 and state data indicating the state of the communication device 20. Other data may also be stored in the state database DB. For example, the state database DB may store information indicating the determination result of the saturation determination unit or information indicating the determination result of the relationship determination unit.
[0032] The identification information of the communication device 20 may be any information, such as an IP address, a device name, or a MAC address. The status data may indicate the pinpoint status of the communication device 20 at a certain point in time, but in this embodiment, the status data is data relating to changes in the status of the communication device 20 over time. The status of the communication device 20 can also be referred to as the load on the communication device 20. The status of the communication device 20 may refer to the hardware status or the software status.
[0033] In this embodiment, an example is given in which the communication volume of the communication device 20 corresponds to the state of the communication device 20, but other states of the communication device 20 may also correspond to the state of the communication device 20. Examples of other states may include the amount of resource consumption of the communication device 20, the processing time (response time) for a request for a service, or the number of errors returned in response to a request. The state data may be indicators known as golden signal metrics or other indicators used in known benchmark tests. For example, the state data may be CPU usage, memory usage, power consumption, communication speed, temperature, or a combination thereof.
[0034] The status data may indicate a status related to performance of the communication service or other services. The other services are services that users use via the communication service, such as e-commerce services, travel reservation services, payment services, or financial services. For example, the status data may indicate the number of users currently connected to the communication device 20 (logged in to the communication service or other services), payments made by users via the communication device 20, or orders placed by users via the communication device 20. The greater the number of payments, payment amounts, number of orders, and order amounts, the greater the load on the communication device 20. Therefore, the number of payments, payment amounts, number of orders, and order amounts also correspond to the status of the communication device 20.
[0035] In this embodiment, the status data indicates changes in the status of the communication device 20 over time. For example, the status data includes the date and time when the status of the communication device 20 was acquired and a numerical value indicating the status of the communication device 20 at that date and time. The status of the communication device 20 may be expressed in other formats, such as characters, instead of numerical values. In this embodiment, the communication volume of the communication device 20 corresponds to the status of the communication device 20, so the status data includes the date and time when the communication volume of the communication device 20 was acquired and the communication volume of the communication device 20 at that date and time.
[0036] The data stored in the data storage unit 100 is not limited to the state database DB. The data storage unit 100 may store any data related to the communication service. For example, the data storage unit 100 may store data indicating the determination result of the saturation determination unit 104. The data storage unit 100 may store data indicating the determination result of the relationship determination unit 105. The data storage unit 100 may store data necessary for displaying an administrator screen SC on the administrator terminal 40. The data storage unit 100 may store various thresholds used in the processing described below. The thresholds may be specified by the administrator or may be determined based on machine learning techniques or statistical techniques.
[0037] [Status Data Acquisition Unit] The status data acquisition unit 101 acquires status data. For example, the status data acquisition unit 101 periodically (for example, every 1 to 30 seconds or every 1 to 5 minutes) requests the status of each of the multiple communication devices 20. When the communication device 20 receives a request from the status data acquisition unit 101 from the server 10, the communication device 20 transmits its own identification information and information indicating its own status to the server 10. The status data acquisition unit 101 adds the current date and time and the information received from the communication device 20 to the status data associated with the identification information of the communication device 20.
[0038] Note that the communication device 20 may accumulate status data for a certain period of time and transmit the status data to the server 10 all at once. In this case, the status data acquisition unit 101 acquires the status data all at once from the communication device 20 and stores it in the status database DB. Alternatively, for example, the status data acquisition unit 101 may acquire the status data from the communication device 20 irregularly rather than periodically. The status data acquisition unit 101 may acquire the status data from the communication device 20 at a timing instructed by an administrator.
[0039] The communication device 20 is an example of a device to be monitored in a communication service. The device to be monitored may be any device and is not limited to the communication device 20. In this embodiment, the description of the communication device 20 can be replaced with another device to be monitored. The other device may be a device used in the communication service, such as the server 10, a computer other than the server 10, components such as a CPU, memory, or network card included in the server 10 or another computer, a battery, an antenna, a communication cable, or a sensor.
[0040] In this embodiment, the status data acquisition unit 101 acquires status data when an abnormality is detected. The status data acquired when an abnormality is detected indicates the status during all or part of the period from when the abnormality occurs to when the abnormality is resolved. Hereinafter, this period will be referred to as the abnormality period. The status data may indicate not only all or part of the abnormality period, but also the status during periods other than the abnormality period. The status data acquisition unit 101 may acquire the status data during the period from when the abnormality occurs to when the abnormality is resolved, or may acquire the status data after the abnormality is resolved.
[0041] In this embodiment, the status data does not indicate the status at a specific pinpoint timing, but indicates the status over a certain period of time. For example, the status data acquiring unit 101 acquires status data relating to changes in the status of the device over time over a period set by the period setting unit 102. In this embodiment, three periods, namely, a saturation period, a reference period, and an abnormal period, are set by the period setting unit 102 (described later), and the status data acquiring unit 101 acquires status data indicating the status of the communication device 20 in each of the saturation period, the reference period, and the abnormal period.
[0042] The status data acquisition unit 101 may acquire status data indicating the status of the communication device 20 in periods other than the saturation period, the reference period, and the abnormality period. For example, the status data acquisition unit 101 may acquire status data indicating the status of the communication device 20 before the saturation period. The status data acquisition unit 101 may acquire status data indicating the status of the communication device 20 after the reference period and the abnormality period. The status data acquisition unit 101 may acquire status data indicating the status of the communication device 20 regardless of the timing of the abnormality. The status data acquisition unit 101 can acquire the status data stored in the status database DB at any timing.
[0043] [Period Setting Unit] The period setting unit 102 sets a period related to the status data. The period related to the status data is a period used in the determination by the relationship determination unit 105, which will be described later. Although the entire period in which the status is indicated in the status data may be used in the determination by the relationship determination unit 105, which will be described later, in this embodiment, only a portion of the period in which the status is indicated in the status data is used in the determination by the relationship determination unit 105, which will be described later. Setting a period means determining the start and end points of the period.
[0044] 7 is a diagram showing an example of periods set by the period setting unit 102. In this embodiment, an example is given in which the period setting unit 102 sets three periods, namely, a saturation period T1, a reference period T2, and an abnormal period T3. The time window W in FIG. 7 is a period of the status data that is used in processing by the relationship determination unit 105, etc. In the example of FIG. 7, the combined period of the reference period T2 and the abnormal period T3 is the time window W.
[0045] For example, the period setting unit 102 may set only one or two of the saturation period T1, the reference period T2, and the abnormal period T3. That is, the period setting unit 102 may set only the saturation period T1, only the reference period T2, only the abnormal period T3, only the saturation period T1 and the reference period T2, only the saturation period T1 and the abnormal period T3, or only the reference period T2 and the abnormal period T3. The period setting unit 102 may set only the time window W without setting the saturation period T1, the reference period T2, and the abnormal period T3.
[0046] For example, the period setting unit 102 sets a saturation period T1 for the status data based on the determination result of the saturation determination unit 104. In the example of FIG. 7 , the period setting unit 102 sets the saturation period T1 as the period from a saturation start time t1, at which the status indicated by the status data begins to saturate, to a saturation end time t2, at which the status indicated by the status data ends to saturation. The saturation start time t1 is the date and time at which the saturation determination unit 104 determines that saturation has started among the dates and times at which the status indicated by the status data is determined. The saturation end time t2 is the date and time at which the saturation determination unit 104 determines that saturation has ended among the dates and times at which the status indicated by the status data is determined. Note that the period setting unit 102 may set the saturation period T1 as the period from the saturation start time t1 to the current time at which saturation has not yet ended (saturation is continuing). For example, the period setting unit 102 may set the saturation period T1 as the period from the saturation start time t1 to a time point designated by the administrator.
[0047] For example, when an abnormality is detected, the period setting unit 102 sets a reference period T2 and an abnormal period T3 related to the status data based on the abnormality occurrence timing t3 at which the abnormality occurred. In this embodiment, when an abnormality is detected, the administrator performs an abnormality occurrence operation to indicate that an abnormality has occurred from the administrator terminal 40. For example, when the administrator receives an inquiry from a user indicating that an abnormality has occurred, the administrator performs the abnormality occurrence operation.
[0048] For example, the server 10 determines whether an abnormality has occurred by determining whether abnormality occurrence data indicating an abnormality-causing operation has been received from the administrator terminal 40. The server 10 does not determine that an abnormality has occurred if it does not receive abnormality occurrence data from the administrator terminal 40. The server 10 determines that an abnormality has occurred if it receives abnormality occurrence data from the administrator terminal 40. The timing at which the server 10 receives the abnormality occurrence data is abnormality occurrence timing t3.
[0049] The administrator may perform an abnormality occurrence operation based on other factors rather than a user inquiry. For example, the administrator may visually check the status data of the communication device 20 on the administrator screen SC or another screen and perform an abnormality occurrence operation. When the execution result of a program that estimates the occurrence of an abnormality is displayed on the administrator screen SC or another screen, the administrator may visually check the execution result and perform an abnormality occurrence operation.
[0050] Furthermore, the method by which the server 10 detects the occurrence of an abnormality may be any method and is not limited to an abnormality occurrence operation by an administrator. For example, the server 10 may detect the occurrence of an abnormality based on the results of execution of an abnormality detection program for detecting the occurrence of an abnormality. The abnormality detection program determines whether an abnormality has occurred based on the status data. For example, the abnormality detection program determines that an abnormality has occurred when the status indicated by the status data exceeds a reference range.
[0051] For example, the anomaly detection program may determine that an anomaly has occurred when the state indicated by the status data exceeds a reference range for a predetermined period of time or more. The anomaly detection program may determine that an anomaly has occurred when the number of times or frequency at which the state indicated by the status data exceeds the reference range exceeds a threshold value. In this case, the timing at which the anomaly detection program determines that an anomaly has occurred is the anomaly occurrence timing t3.
[0052] 7 , the period setting unit 102 sets the reference period T2 as the period from a reference timing t4, which is the reference for the reference period T2, to an abnormality occurrence timing t3, at which an abnormality occurs. For example, the reference timing t4 is a timing that is a predetermined time before the abnormality occurrence timing t3. The reference timing t4 may be a timing that is a predetermined time before the current time, or may be a timing specified by the administrator.
[0053] For example, the period setting unit 102 sets the period from the abnormality occurrence timing t3 to the abnormality resolution timing t5 when the abnormality is resolved as the abnormality period T3. The abnormality resolution timing t5 is the timing at which it is determined that the abnormality has been resolved. In this embodiment, when the abnormality is resolved, the administrator performs an abnormality resolution operation from the administrator terminal 40 to indicate that the abnormality has been resolved. For example, the administrator performs the abnormality resolution operation when the administrator no longer receives inquiries from users indicating that an abnormality has occurred.
[0054] For example, the server 10 determines whether the abnormality has been resolved by determining whether abnormality resolution data indicating an abnormality resolution operation has been received from the administrator terminal 40. The server 10 does not determine that the abnormality has been resolved if it does not receive abnormality resolution data from the administrator terminal 40. The server 10 determines that the abnormality has been resolved if it receives abnormality resolution data from the administrator terminal 40. The timing at which the server 10 receives the abnormality resolution data is the abnormality resolution timing t5.
[0055] The administrator may perform an operation to resolve the abnormality based on other factors rather than a user inquiry. For example, the administrator may visually check the status data of the communication device 20 on the administrator screen SC or another screen and perform an operation to resolve the abnormality. When the execution result of a program that estimates the resolution of the abnormality is displayed on the administrator screen SC or another screen, the administrator may visually check the execution result and perform an operation to resolve the abnormality.
[0056] Furthermore, the method by which the server 10 detects the elimination of the abnormality may be any method and is not limited to an administrator's operation to resolve the abnormality. For example, the server 10 may detect the elimination of the abnormality based on the execution result of an abnormality resolution program for detecting the elimination of the abnormality. The abnormality resolution program determines whether the abnormality has been resolved based on the status data. For example, the abnormality resolution program determines that the abnormality has been resolved when the state indicated by the status data falls within a reference range.
[0057] For example, the abnormality resolution program may determine that the abnormality has been resolved when the state indicated by the status data remains within a reference range for a predetermined period of time or more. The abnormality resolution program may also determine that the abnormality has been resolved when the number or frequency at which the state indicated by the status data falls within the reference range exceeds a threshold. In this case, the timing at which the abnormality resolution program determines that the abnormality has been resolved is the abnormality resolution timing t5.
[0058] [Normalization Execution Unit] The normalization execution unit 103 performs normalization on the status data. For example, the normalization execution unit 103 performs normalization based on the status data of each of the multiple communication devices 20. The normalization execution unit 103 performs normalization so that the status numerical values indicated by the status data of each of the multiple communication devices 20 fall within a certain range. The normalization execution unit 103 may perform normalization based on a known normalization method. For example, the normalization method may be a z-score normalization method or a minimum-maximum normalization method.
[0059] [Saturation Determination Unit] The saturation determination unit 104 determines whether the time-series state of the communication device 20 is saturated based on the status data. Saturation refers to the amount of change in the status indicated by the status data falling below a threshold, or the amount of change remaining below the threshold for a predetermined period of time or more. In cases where there is no upper limit, such as with communication volume, the amount of change primarily refers to the amount of increase when communication volume increases. Therefore, saturation refers to the amount of increase in the status indicated by the status data falling below a threshold after the total amount of change increases over a certain period of time, or the amount of change remaining below the threshold for a predetermined period of time or more. The amount of change may be an integrated value over a certain period of time, or the amount of change between one point in time and the next. The amount of change may also refer to the amount of decrease in communication volume. In other words, the saturation determination unit 104 determines saturation based on the time-series change in the amount of change itself indicated by the status data.
[0060] For example, the saturation determination unit 104 determines that the time-series state of the communication device 20 is saturated when the increase in the state indicated by the state data changes from a state where the increase is equal to or greater than a threshold to a state where the increase is less than the threshold, or a state where the increase is less than the threshold continues for a predetermined period of time or more. Even if the increase in the state indicated by the state data changes to a state where the increase is less than the threshold, or a state where the increase is less than the threshold continues for a predetermined period of time or more, the saturation determination unit 104 determines that the time-series state of the communication device 20 is not saturated if the increase was less than the threshold before that. For example, the saturation determination unit 104 may determine that the time-series state of the communication device 20 is saturated when the state indicated by the state data is equal to or greater than a threshold and the increase in the state changes to a state where the increase is less than the threshold, or a state where the increase is less than the threshold continues for a predetermined period of time or more. The saturation determination unit 104 determines that the time-series state of the communication device 20 is not saturated if the increase in the state indicated by the state data is less than a threshold value, or if the increase in the state less than the threshold value continues for a predetermined period of time or more.
[0061] For example, the saturation determination unit 104 determines whether the time-series state of a communication device 20 is saturated based on the state data of each of the multiple communication devices 20 and a predetermined threshold. In other words, the saturation determination unit 104 determines whether the time-series state indicated by the state data of each communication device 20 is saturated based on the predetermined threshold. The threshold referenced by the saturation determination unit 104 when determining saturation may be specified by an administrator or may be calculated from past abnormality trends. The threshold may also be determined based on a model using a machine learning technique.
[0062] The saturation determination unit 104 may determine whether the state of the communication device 20 is saturated based on a model that uses a machine learning technique, without using a threshold value. In this case, the model is trained with training data that receives as input time-series state changes indicated by training state data and outputs labels indicating whether the state is saturated. The saturation determination unit 104 may input the state data to be determined to the trained model, and determine whether the state of the communication device 20 is saturated based on the labels output from the model.
[0063] For example, the saturation determination unit 104 determines whether the time-series state of a communication device 20 is saturated by determining whether the amount of change indicated by the state data of the communication device 20 is equal to or greater than a threshold. If the saturation determination unit 104 determines that the amount of change indicated by the state data of the communication device 20 is equal to or greater than the threshold, the saturation determination unit 104 does not determine that the time-series state of the communication device 20 is saturated. If the saturation determination unit 104 determines that the amount of change indicated by the state data of the communication device 20 is less than the threshold, the saturation determination unit 104 determines that the time-series state of the communication device 20 is saturated.
[0064] For example, the saturation determination unit 104 determines whether the time-series state of a communication device 20 is saturated by determining whether a state in which the amount of change indicated by the status data of the communication device 20 is less than a threshold has continued for a predetermined time or more. If the saturation determination unit 104 determines that a state in which the amount of change indicated by the status data of the communication device 20 is less than a threshold has not continued for a predetermined time or more, the saturation determination unit 104 does not determine that the time-series state of the communication device 20 is saturated. If the saturation determination unit 104 determines that a state in which the amount of change indicated by the status data of the communication device 20 is less than a threshold has continued for a predetermined time or more, the saturation determination unit 104 determines that the time-series state of the communication device 20 is saturated.
[0065] In this embodiment, when an abnormality is detected, the saturation determination unit 104 determines whether the state of the communication device 20 is saturated based on the state data of the communication device 20. That is, after an abnormality occurs, the saturation determination unit 104 determines whether the state of the communication device 20 is saturated based on the state data of the communication device 20. For example, the saturation determination unit 104 determines whether the state of the communication device 20 is saturated based on the state data on which normalization has been performed. The saturation determination unit 104 determines whether the state of the communication device 20 is saturated based on the state indicated by the normalized state data.
[0066] In this embodiment, the saturation determination unit 104 determines whether or not a state indicated by the state data is saturated based on a state within a time window W. The saturation determination unit 104 may determine whether or not a state indicated by the state data is saturated regardless of the time window W. For example, the saturation determination unit 104 may determine whether or not a state indicated by the state data is saturated based on a state within a reference period T2. The saturation determination unit 104 may determine whether or not a state indicated by the state data is saturated based on a state within an abnormal period T3. The saturation determination unit 104 may determine whether or not a state is saturated based on a state throughout the entire period indicated by the state data.
[0067] [Relationship Determination Unit] The relationship determination unit 105 determines whether the status data is related to an anomaly in the communication service based on the determination result of the saturation determination unit 104. The status data being related to an anomaly means that there is a correlation between the anomaly and the status data. If the state indicated by the status data becomes saturated when an anomaly occurs, the anomaly is related to the status data. When an anomaly is detected, the state indicated by the status data does not necessarily have to become saturated; it is sufficient that there is some correlation between the anomaly and the status data. For example, when an anomaly occurs n times (n is a natural number), if the state indicated by the status data becomes saturated k times or more (k is a natural number equal to or less than n), the status data is related to the anomaly. Here, k is assumed to be a predetermined percentage of n (e.g., 70%) or more.
[0068] In this embodiment, when an abnormality is detected, the relationship determination unit 105 determines whether the status data is related to the abnormality. That is, the relationship determination unit 105 determines whether the status data is related to the abnormality after the abnormality occurs. The relationship determination unit 105 may determine whether the status data is related to the abnormality before the abnormality is resolved (i.e., during the abnormality period T3). However, in this embodiment, an example is given in which the relationship determination unit 105 determines whether the status data is related to the abnormality after the abnormality is resolved.
[0069] For example, the relationship determination unit 105 determines whether the status data is abnormally related based on a saturation period T1 during which the status is determined to be saturated among periods during which the status data indicates a change. The period during which the status data indicates a change is the period of the timestamp included in the status data. The saturation period T1 is the period during which the saturation determination unit 104 determines that the status of the communication device 20 is saturated. In other words, the saturation period T1 is the period during which the amount of change in the status indicated by the status data is less than a threshold, or the period during which the amount of change in the status indicated by the status data remains less than the threshold.
[0070] For example, the relationship determination unit 105 may determine whether the status data is related to the abnormality based on the degree of coincidence between an abnormal period T3, in which an abnormality occurs, and a saturation period T1, in which the status is determined to be saturated, among the periods in which a change is indicated in the status data. The degree of coincidence here refers to the length or proportion of the overlap period in which the saturation period T1 and the abnormal period T3 overlap. The more the saturation period T1 and the abnormal period T3 coincide, the longer the overlap period or the higher the proportion of the overlap period. For example, the relationship determination unit 105 calculates the overlap period in which the saturation period T1 and the abnormal period T3 overlap, and determines whether the status data is related to the abnormality based on the overlap period.
[0071] For example, if the length of the overlapping period is less than a threshold, the relationship determining unit 105 does not determine that the status data is related to the abnormality, but if the length of the overlapping period is equal to or greater than the threshold, the relationship determining unit 105 determines that the status data is related to the abnormality. The threshold may be set in advance, or may be set based on the length of at least one of the saturation period T1 and the abnormality period T3. For example, the relationship determining unit 105 may determine that the status data is not related to the abnormality if the overlapping period is less than a predetermined percentage (e.g., 80%) of at least one of the saturation period T1 and the abnormality period T3, and may determine that the status data is related to the abnormality if the overlapping period is equal to or greater than the predetermined percentage.
[0072] The method by which the relationship determination unit 105 determines whether the status data is related to the abnormality based on the saturation period T1 is not limited to the method based on the overlap period. For example, the relationship determination unit 105 may determine whether the status data is related to the abnormality by determining whether the saturation period T1 includes the abnormality occurrence timing t3. In this case, the relationship determination unit 105 may determine that the status data is not related to the abnormality if the saturation period T1 does not include the abnormality occurrence timing t3, and may determine that the status data is related to the abnormality if the saturation period T1 includes the abnormality occurrence timing t3.
[0073] Alternatively, for example, the relationship determination unit 105 may determine whether the status data is related to the abnormality by determining whether the length of the saturation period T1 is equal to or greater than a threshold. The relationship determination unit 105 may determine that the status data is not related to the abnormality if the length of the saturation period T1 is less than the threshold, and may determine that the status data is related to the abnormality if the length of the saturation period T1 is equal to or greater than the threshold.
[0074] For example, the relationship determination unit 105 may determine whether the status data is related to the abnormality based on the highest value of the status in another period before the abnormal period T3 in which the abnormality occurs, among the periods in which a change is indicated in the status data. The other period may be any period before the abnormal period T3. In this embodiment, an example is given in which the reference period T2 corresponds to the other period. The other period may be immediately before the abnormal period T3, or may be a period that is not immediately before the abnormal period T3 but is some time before the abnormal period T3. The other period may be a period before the reference period T2. Note that the abnormal period T3 may include a timing slightly away from the timing t3 at which the abnormality occurs. The highest value is the maximum value of the communication volume within a certain period.
[0075] For example, the relationship determination unit 105 may determine whether the status data is abnormally related based on the highest value of the status in another period (in this embodiment, the reference period T2) and the highest value of the status in a period prior to the other period. The period prior to the other period is a period prior to the reference period T2. For example, the period prior to the other period may be the entire period prior to the reference period T2, or a portion of the period prior to the reference period T2. The highest value is a value that is the peak of a time-series change. The relationship determination unit 105 refers to the status database DB and obtains the highest value in the other period and the highest value of the status in a period prior to the other period.
[0076] For example, the relationship determination unit 105 determines whether the highest state value in the other period (in this embodiment, the reference period T2) is less than the highest state value in the period before the other period. If the highest state value in the other period (in this embodiment, the reference period T2) is equal to or greater than the highest state value in the period before the other period, the relationship determination unit 105 determines that the state data is not related to the abnormality, and if the highest state value in the other period (in this embodiment, the reference period T2) is less than the highest state value in the period before the other period, the relationship determination unit 105 determines that the state data is related to the abnormality.
[0077] For example, the relationship determination unit 105 may determine whether the status data is related to the abnormality based on the highest value of the state indicated by the status data during an abnormal period T3 in which an abnormality occurs, among the periods in which a change is indicated in the status data. The relationship determination unit 105 obtains the highest value of the state indicated by the status data during the abnormal period T3. If the relationship determination unit 105 determines that the highest value is less than a threshold, it does not determine that the status data is related to the abnormality, but if the highest value is determined to be equal to or greater than the threshold, it determines that the status data is related to the abnormality.
[0078] For example, the relationship determination unit 105 may determine whether the status data is related to the abnormality based on the amount of change in the status during a saturation period T1 in which the status is saturated and another period prior to the abnormal period T3 in which the abnormality is occurring, among the periods in which the status data indicates a change. As described above, the reference period T2 is an example of the other period. The relationship determination unit 105 determines whether the difference between the amount of change in the status during the saturation period T1 and the amount of change in the status during the other period is equal to or greater than a threshold. If the relationship determination unit 105 determines that the difference is less than the threshold, it does not determine that the status data is related to the abnormality. However, if the relationship determination unit 105 determines that the difference is equal to or greater than the threshold, it determines that the status data is related to the abnormality.
[0079] The relationship determination unit 105 may determine whether the status data is related to the abnormality based on a determination method other than the determination method described above. For example, if the state indicated by the status data is saturated during the abnormality period T3, but the state is also saturated during periods when no abnormality occurs, the relationship determination unit 105 may determine that the status data is not related to the abnormality because the saturation occurs steadily in the status data. On the other hand, if the state indicated by the status data is not saturated during periods when no abnormality occurs, but the state indicated by the status data is saturated during the abnormality period T3, the relationship determination unit 105 may determine that the status data is not related to the abnormality because the saturation occurs only during the abnormality period T3.
[0080] [Recovery Processing Execution Unit] When an abnormality is detected, the recovery processing execution unit 106 executes recovery processing related to recovery from the abnormality based on the status data determined to be related to the abnormality. The recovery processing may be any processing related to the abnormality that has occurred. In this embodiment, the recovery processing corresponds to the processing of displaying the status data related to the abnormality on the administrator screen SC. The recovery processing may be other processing, such as the processing of executing a prepared recovery program or the processing of sending a notification such as an email to the administrator.
[0081] For example, when an abnormality is detected, the recovery process execution unit 106 transmits display data for the administrator screen SC to the administrator terminal 40, thereby displaying the administrator screen SC on the administrator terminal 40. When an abnormality is detected, the recovery process execution unit 106 displays an image showing status data related to the abnormality on the administrator screen SC. In the example of FIG. 4 , when a certain abnormality occurs, the recovery process execution unit 106 displays on the administrator screen SC the status indicated by the status data determined by the relationship determination unit 105 when the same abnormality occurred in the past. When a certain abnormality occurs and status data related to the abnormality is identified, the recovery process execution unit 106 executes recovery processing based on the status data related to the abnormality when the same abnormality occurs again.
[0082] [1-3-2. Functions Realized by the Administrator Terminal] For example, the administrator terminal 40 includes a data storage unit 400, an operation reception unit 401, and a display control unit 402. The data storage unit 400 is realized mainly by the storage unit 42. The operation reception unit 401 and the display control unit 402 are realized mainly by the control unit 41.
[0083] [Data Storage Unit] The data storage unit 400 stores data necessary for the administrator's work. For example, the data storage unit 400 stores data necessary for displaying the administrator screen SC. The data storage unit 400 may also store maintenance tools necessary for the administrator's work. The maintenance tools themselves may be publicly known tools, and may be, for example, tools capable of monitoring the status of at least one of hardware and software.
[0084] [Operation Acceptance Unit] The operation acceptance unit 401 accepts various operations by the administrator. For example, the operation acceptance unit 401 accepts operations on the administrator screen SC.
[0085] [Display Control Unit] The display control unit 402 displays various screens on the display unit 45. For example, the display control unit 402 displays an administrator screen SC on the display unit 45.
[0086] 8 is a diagram showing an example of processing executed in the monitoring system 1. The control units 11 and 41 execute programs stored in the storage units 12 and 42, respectively, to execute the processing in FIG.
[0087] 8 , the server 10 communicates with each of the multiple communication devices 20, acquires status data, and stores the acquired data in the status database DB (S1). When an abnormality occurs, the administrator terminal 40 accepts an abnormality occurrence operation from the administrator (S2). The administrator terminal 40 transmits abnormality occurrence data indicating that an abnormality occurrence operation has been performed to the server 10 (S3). The server 10 receives the abnormality occurrence data from the administrator terminal 40 (S4). The server 10 records the timing at which the abnormality occurrence data is received as the abnormality occurrence timing t3 at which the abnormality occurred in the storage unit 12 (S5). The administrator may specify the abnormality occurrence timing t3. In this case, the server 10 records the abnormality occurrence timing t3 specified by the administrator in the storage unit 12.
[0088] When the abnormality is resolved, the administrator terminal 40 accepts the administrator's operation to resolve the abnormality (S6). The administrator terminal 40 transmits abnormality resolution data indicating that the abnormality resolution operation has been performed to the server 10 (S7). The server 10 receives the abnormality resolution data from the administrator terminal 40 (S8). The server 10 records the timing at which the abnormality resolution data is received in the storage unit 12 as the abnormality resolution timing t5 at which the abnormality is resolved (S9). The administrator may also specify the abnormality resolution timing t5. In this case, the server 10 records the abnormality resolution timing t5 specified by the administrator in the storage unit 12.
[0089] The server 10 sets the period from reference timing t4, which is a predetermined time before the abnormality occurrence timing t3, to the abnormality occurrence timing t3 as the reference period T2 (S10). The server 10 sets the period from the abnormality occurrence timing t3 to the abnormality resolution timing t5 as the abnormality period T3 (S11). The server 10 references the status database DB and acquires status data for the period including the abnormality occurrence timing t3 as status data to be determined (S12). The status data to be determined is the status data that will be processed in S13 and subsequent steps. The subsequent processing is performed for each status data.
[0090] The server 10 normalizes the status data to be determined (S13). In S13, the server 10 may perform preprocessing such as standardization or smoothing on the status data to be determined in addition to normalization. Alternatively, for example, the server 10 may not perform the processes from S14 onwards if the status indicated by the status data to be determined is less than a predetermined threshold.
[0091] The server 10 determines whether the state indicated by the normalized state data to be determined is saturated (S14). In S14, the server 10 calculates the amount of change over time in the state indicated by the state data. The server 10 determines that the state indicated by the state data is saturated if the amount of change over time is less than a threshold value for a continuous period. The server 10 determines that the state indicated by the state data is not saturated if the amount of change over time is equal to or greater than a threshold value, or if the amount of change over time is not less than the threshold value for a continuous period.
[0092] In S14, if it is determined that the state indicated by the status data to be determined is not saturated (S14: N), the server 10 determines that the status data to be determined is not related to an abnormality (S15), and proceeds to the process of S21. In S14, if it is determined that the state indicated by the status data to be determined is saturated (S14: Y), the server 10 sets the period from the saturation start timing t1 to the saturation end timing t2 as the saturation period T1 (S16).
[0093] The server 10 determines whether the first maximum value, which is the maximum value of the status data in the reference period T2, is equal to or greater than the second maximum value, which is the maximum value of the status data in a period earlier than the reference period T2 (S17). If it is determined in S17 that the first maximum value is equal to or greater than the second maximum value (S17: Y), the server 10 proceeds to S15 and determines that the status data to be determined is not related to an abnormality.
[0094] If it is determined in S17 that the first highest value is less than the second highest value (S17: N), the server 10 determines whether the third highest value in the abnormal period T3 of the status data to be determined is equal to or greater than the first highest value or the second highest value (S18). Here, a case will be described in which the server 10 compares the third highest value with either the first highest value or the second highest value (i.e., a case in which only one of the first highest value or the second highest value is compared), but the server 10 may also compare the third highest value with both the first highest value and the second highest value.
[0095] If it is determined in S18 that the highest value in the abnormal period T3 is less than the first highest value or the second highest value (S18: N), the server 10 proceeds to S15 and determines that the status data being determined is not related to an abnormality. If it is determined in S18 that the third highest value, which is the highest value in the abnormal period T3, is equal to or greater than the first highest value or the second highest value (S18: Y), the server 10 determines whether the degree of coincidence between the saturation period T1 and the abnormal period T3 is equal to or greater than a threshold value (S19). If it is determined in S19 that the degree of coincidence between the saturation period T1 and the abnormal period T3 is less than the threshold value (S19: N), the server 10 proceeds to S15 and determines that the status data being determined is not related to an abnormality.
[0096] If it is determined in S19 that the degree of coincidence between the saturation period T1 and the abnormal period T3 is equal to or greater than the threshold (S19: Y), the server 10 records the status data to be determined in the storage unit 12 as a candidate for status data related to the abnormality (S20). The server 10 determines whether all status data have been determined (S21). If it is determined in S21 that there is status data that has not yet been determined (S21: N), the process returns to S12, and processing is performed on the next status data to be determined.
[0097] If it is determined in S21 that all status data have been determined (S21: Y), the server 10 determines, from the status data recorded as candidates in S20, status data related to the abnormality (S22), and this process ends. In S22, the server 10 calculates a feature value indicating the degree of relationship between the status data and the abnormality based on the amount of change in the saturation period T1 and the reference period T2 of each of the status data recorded as candidates in S20, and determines a predetermined number of status data in descending order of the feature value as status data related to the abnormality. Note that the process of S22 may be skipped and the status data recorded as candidates in S20 may be determined as status data related to the abnormality as is.
[0098] [1-5. Summary of the embodiment] The monitoring system 1 of this embodiment determines whether the state of the communication device 20 is saturated based on status data relating to time-series status changes of the communication device 20. Based on the determination result, the monitoring system 1 determines whether the status data is related to an abnormality in the communication service. This allows the monitoring system 1 to determine whether the status data is related to an abnormality in the communication service. As a result, for example, the monitoring system 1 can speed up recovery from the abnormality. In addition, for example, the monitoring system 1 can quickly detect an abnormality and analyze the cause of the abnormality.
[0099] Furthermore, when an abnormality is detected, the monitoring system 1 acquires status data, determines whether the status of the communication device 20 is saturated, and determines whether the status data is related to the abnormality. When an abnormality is detected, the monitoring system 1 executes recovery processing based on the status data determined to be related to the abnormality. This allows the monitoring system 1 to speed up recovery when an abnormality is detected.
[0100] Furthermore, when an abnormality is detected, the monitoring system 1 sets a period related to the status data based on the abnormality occurrence timing t3 at which the abnormality occurred, and acquires status data related to time-series changes in the status of the communication device 20 during that period. This allows the monitoring system 1 to determine whether the status data is related to the abnormality based on the status data for the period related to the abnormality, thereby improving the accuracy of the determination. By accurately identifying the status data related to the abnormality, the monitoring system 1 can speed up recovery from the abnormality.
[0101] Furthermore, the monitoring system 1 determines whether the state is saturated based on the normalized state data. As a result, even if the range of change in communication volume varies for each communication device 20, the monitoring system 1 can absorb that range through normalization, thereby improving the accuracy of determining whether the state data is related to an abnormality.
[0102] Furthermore, the monitoring system 1 determines whether the status data is related to the abnormality based on a saturation period T1 during which the status is determined to be saturated among periods during which a change is indicated in the status data. This allows the monitoring system 1 to determine whether the status data is related to the abnormality by using a period that is likely to be directly linked to the abnormality, such as the saturation period T1, thereby improving the accuracy of determining whether the status data is related to the abnormality.
[0103] Furthermore, the monitoring system 1 determines whether the status data is related to the abnormality based on the highest status value in a period prior to the abnormal period T3 in which the abnormality occurred, among the periods in which the status data showed a change. This allows the monitoring system 1 to determine whether the status data is related to the abnormality by taking into consideration, for example, whether the communication volume steadily increases from when no abnormality occurs, thereby improving the accuracy of determining whether the status data is related to the abnormality.
[0104] Furthermore, the monitoring system 1 determines whether the status data is related to the abnormality based on the highest status value in another period before the abnormality period T3 and the highest status value in a period before that other period. This allows the monitoring system 1 to determine whether the status data is related to the abnormality by taking into consideration, for example, whether the communication volume steadily increases from when no abnormality occurs, thereby improving the accuracy of determining whether the status data is related to the abnormality.
[0105] Furthermore, the monitoring system 1 determines whether the status data is related to the abnormality based on the highest value of the state indicated by the status data during the abnormal period T3 during which the abnormality occurred, among the periods during which the status data showed a change. Even if the state indicated by the status data is saturated, if the value is not too high, the state may not be related to the abnormality. Therefore, the monitoring system 1 can improve the accuracy of determining whether the status data is related to the abnormality by determining whether the status data is related to the abnormality by taking into account the highest value of the state indicated by the status data during the abnormal period T3.
[0106] Furthermore, the monitoring system 1 determines whether the status data is related to the abnormality based on the degree of coincidence between the abnormal period T3, in which an abnormality occurs, and the saturation period T1, in which the status is determined to be saturated, among the periods in which a change is indicated in the status data. Since the higher the degree of coincidence between the saturation period T1 and the abnormal period T3, the higher the degree to which the status data is related to the abnormality, the monitoring system 1 can improve the accuracy of determining whether the status data is related to the abnormality by using the degree of coincidence.
[0107] Furthermore, the monitoring system 1 determines whether the status data is related to the abnormality based on the amount of change in the status during each of the saturation period T1 in which the status is saturated and the abnormality period T3 in which the abnormality occurs, among the periods in which the status data indicates a change. This allows the monitoring system 1 to determine whether the status data is related to the abnormality, even if the status indicated by certain status data is saturated, by taking into consideration whether the state is likely to be steadily saturated, thereby improving the accuracy of determining whether the status data is related to the abnormality.
[0108] [1-6. Modifications] The present disclosure is not limited to the above-described embodiments, and can be modified as appropriate without departing from the spirit of the present disclosure.
[0109] [1-6-1. Modification 1-1] FIG. 9 is a diagram illustrating an example of functions implemented by a monitoring system 1 according to Modification 1-1. The monitoring system 1 according to Modification 1-1 includes a detection process execution unit 107. The detection process execution unit 107 is implemented by the control unit 11. For example, in the embodiment, a case has been described in which a series of processes is executed when an abnormality is detected, but the series of processes may be executed before the abnormality is detected. Modification 1-1 describes a case in which the determination result of the saturation determination unit 104 is used to detect an abnormality. Note that in Modification 1-1, the period of the status data used in the series of processes (the period corresponding to the time window W in FIG. 7 ) is set to the period from a predetermined period (e.g., approximately 15 minutes) before the current time to the current time.
[0110] The status data acquisition unit 101 in Modification 1-1 acquires status data before an abnormality is detected. The method of acquiring status data by the status data acquisition unit 101 is the same as in the embodiment. The status data acquisition unit 101 acquires status data in real time while the communication service is in operation. The status data acquisition unit 101 acquires status data regardless of whether an abnormality has occurred. The saturation determination unit 104 in Modification 1-1 determines whether the status is saturated based on the status data before an abnormality is detected. The method of determining saturation by the saturation determination unit 104 is also the same as in the embodiment. The saturation determination unit 104 determines whether the status indicated by the status data is saturated regardless of whether an abnormality has occurred.
[0111] The relationship determination unit 105 in Modification 1-1 determines whether or not status data is related to the abnormality before the abnormality is detected. The relationship determination unit 105 determines whether or not the status indicated by the status data is related to the abnormality, regardless of whether or not the abnormality has occurred. For example, if the status indicated by the status data is determined to be not saturated, the relationship determination unit 105 determines that the status is not related to the abnormality, and if the status indicated by the status data is determined to be saturated, the relationship determination unit 105 determines that the status is related to the abnormality.
[0112] The detection process execution unit 107 executes a detection process for detecting an abnormality based on the status data determined to be related to an abnormality. The detection process may be any process for detecting an abnormality that has occurred. In Modification 1-1, the detection process corresponds to a process for displaying the occurrence of an abnormality on the administrator screen SC. For example, when an abnormality is detected, the detection process execution unit 107 displays an image indicating that an abnormality has been detected on the administrator screen SC. The detection process execution unit 107 may execute the detection process by sending a notification to the administrator by means of e-mail or the like.
[0113] The monitoring system 1 of Modification 1-1 acquires status data, determines whether the status of the communication device 20 is saturated, and determines whether the status data is related to the abnormality before detecting an abnormality. The monitoring system 1 executes a detection process for detecting the abnormality based on the status data determined to be related to the abnormality. This allows the monitoring system 1 to quickly detect the abnormality.
[0114] [1-6-2. Modification 1-2] For example, the monitoring system 1 can be applied to services other than communication services. For example, the monitoring system 1 may determine whether an abnormality in another service, such as an e-commerce service, a travel reservation service, a financial service, a payment service, an online flea market service, or a video distribution service, is related to status data of a device in the other service.
[0115] For example, the monitoring system 1 may determine whether an anomaly in each of a plurality of services is related to status data of a device in the service or another service. The monitoring system 1 can be interpreted as a tenant in a cloud infrastructure or cloud platform. The monitoring system 1 is configured by a plurality of offering services selected from a group of offering services including software configurations or hardware configurations.
[0116] The status data acquisition unit 101 in Modification 1-2 acquires status data of a device for each of a plurality of services. The process by which the status data acquisition unit 101 acquires status data of a device for each service may be the same as in the embodiment. The status data acquisition unit 101 communicates with each device for each of the plurality of services and acquires the status data of the device.
[0117] The saturation determination unit 104 in the modification 1-2 determines whether the device is saturated based on the device status data for each of a plurality of services. The process by which the saturation determination unit 104 determines saturation based on the device status data for each service may be the same as that in the embodiment.
[0118] The relationship determination unit 105 in Modification 1-2 determines whether or not the status data is related to an abnormality in each of the plurality of services based on the determination result of the saturation determination unit 104. The process by which the relationship determination unit 105 determines whether or not the status data is related to an abnormality in each service may be the same as in the embodiment.
[0119] The monitoring system 1 of Variation 1-2 acquires device status data for each of a plurality of services. Based on the device status data for each of the plurality of services, the monitoring system 1 determines whether the device is saturated. Based on the determination result, the monitoring system 1 determines whether the status data is related to an abnormality in each of the plurality of services. This allows the monitoring system 1 to respond to an abnormality in each of the plurality of services.
[0120] [1-6-3. Other Modifications] For example, the above modifications may be combined.
[0121] For example, in the embodiment, the main processing is described as being executed on the server 10, but the processing described as being executed on the server 10 may be executed on the administrator terminal 40 or another computer, or may be shared among multiple computers.
[0122] [2. Second Embodiment] A second embodiment, which is another example of a monitoring system 1 according to the present disclosure, will be described. In the second embodiment, an example of processing related to an abnormality in a service will be described. Conventional technology can only identify recovery procedures based on the status of a predetermined device among various devices in the system. An abnormality that occurs in a system may involve various devices. Even if the status data of multiple devices are closely related, conventional technology cannot identify the relationship between these devices. Therefore, the monitoring system 1 of the second embodiment identifies the relationship between status data related to the status of devices to be monitored in a service.
[0123] [2-1. Hardware Configuration of Monitoring System of Second Embodiment] The hardware configuration of the monitoring system 1 of the second embodiment may be the same as that of the first embodiment.
[0124] [2-2. Overview of Monitoring System] In this embodiment, an example is given in which the communication device 20 is a server computer having a containerized network function (CNF). Furthermore, an example is given in which the user terminal 30 is a smartphone. The communication device 20 relays communications between a large number of user terminals 30. For example, when the communication volume that the communication device 20 must process reaches the communication volume that the communication device 20 can process, an abnormality may occur in the communication service. The communication volume can be expressed by a known index such as bps.
[0125] For example, when an abnormality occurs in a communication service, the administrator identifies the communication device 20 involved in the abnormality and performs recovery work. However, in a communication service, it is possible that not only one communication device 20 but multiple communication devices 20 are involved in the abnormality. For example, when an abnormality occurs in a communication service in a specific area, in addition to the communication device 20 identified by the administrator as the cause of the abnormality, other communication devices 20 related to the communication device 20 may also be involved in the abnormality. It is difficult for the administrator to identify all communication devices 20 involved in the abnormality on their own.
[0126] For example, it is conceivable that an abnormality is detected when the communication volume processed by the communication device 20 exceeds a threshold value. However, there are cases where the communication device 20 can process a larger communication volume than the upper limit set by the administrator, or where the communication device 20 can only process a lower communication volume than the upper limit set by the administrator. Furthermore, since the administrator himself may not know the upper limit value of the communication device 20, in this embodiment, no upper limit value for the communication volume used to determine an abnormality is set. Therefore, even if the communication volume of the communication device 20 reaches a certain level, it does not necessarily correspond to an abnormality.
[0127] In this embodiment, the server 10 identifies the communication volume of other communication devices 20 that have characteristics similar to the communication volume of a communication device 20 involved in an anomaly that has occurred in the communication service. For example, the communication device 20 involved in the anomaly is identified through analysis by an administrator. As in a modified example described below, the communication device 20 involved in the anomaly may be identified by saturation of the communication volume. Because the other communication devices 20 may be involved in the anomaly, identifying the other communication devices 20 is useful for detecting and recovering from the anomaly.
[0128] FIG. 10 is a diagram showing an example of a change in communication volume over time for a communication device 20 when an abnormality is detected in a communication service. The horizontal axis of FIG. 10 is the time axis. The vertical axis of FIG. 10 is the axis showing communication volume. In the example of FIG. 10, the change in communication volume for each of five communication devices 20A to 20E located at the facility of a telecommunications carrier is shown. Hereinafter, when there is no need to distinguish between the communication devices 20A to 20E, they will simply be referred to as communication device 20.
[0129] In the example of Figure 10, at a certain point in time, the communication volume of communication device 20A increases suddenly. After that, communication device 20A is no longer able to handle communications above a certain level, and the communication volume becomes saturated. In this case, a user terminal 30 that attempts to connect to communication device 20A in a certain area will be unable to use the communication service. In other words, an abnormality occurs in the communication service in that area. The user may contact the call center of the communication service and inform them that the communication service is unavailable.
[0130] For example, an administrator receives inquiries from multiple users and becomes aware that an abnormality has occurred. At this point, the administrator has not yet identified which communication device 20 is causing the abnormality. The amount of communication traffic on communication device 20A subsequently decreases, and the abnormality spontaneously recovers. Furthermore, the administrator stops receiving inquiries from users and becomes aware that the abnormality has spontaneously recovered. Once the abnormality spontaneously recovers, the user terminal 30 becomes able to use the communication service via communication device 20A.
[0131] For example, after an abnormality occurs in a specific area and is resolved, the administrator analyzes the traffic volume of each of communication devices 20A-20E. The administrator determines that the traffic volume of communication device 20A increased rapidly and reached saturation during the period when the abnormality occurred. The administrator inputs information to administrator terminal 40 indicating that the traffic volume of communication device 20A is related to the abnormality that occurred in the area. Administrator terminal 40 uploads the information to server 10. Server 10 stores the area and the fact that the traffic volume of communication device 20A is related to the abnormality in the area in association with each other.
[0132] For example, the server 10 clusters each of the communication devices 20A to 20E to determine whether or not the status data of each of the communication devices 20B to 20E contains status data that is related to the status data of the communication device 20A. In the example of Fig. 10, the clustering identifies that the characteristics of the communication volume of the communication device 20A and the characteristics of the communication volume of the communication device 20E are similar.
[0133] For example, the server 10 records in the storage unit 12 that an abnormality in a certain area is related not only to the communication device 20A specified by the administrator, but also to the communication device 20E identified by clustering. If an abnormality occurs again in the area, the server 10 transmits a notification to the administrator terminal 40 urging the administrator to check the status of not only the communication device 20A but also the communication device 20E.
[0134] 11 is a diagram showing an example of a notification displayed on the administrator terminal 40. For example, suppose that a call center for a communication service receives an inquiry from a user who is attempting to use the communication service from a specific area X when an abnormality in the communication volume of communication device 20A was previously detected. If the call center currently receives an inquiry from a user who is attempting to use the communication service in the same area X, it detects that an abnormality has occurred in the communication service in area X. In this case, the server 10 transmits, to the administrator terminal 40, display data for an administrator screen SC that prompts the administrator to confirm communication device 20A specified by the administrator and communication device 20E identified by clustering.
[0135] For example, when the administrator terminal 40 receives the display data of the administrator screen SC, a message is displayed prompting the administrator to check the communication volume of communication device 20A and the communication volume of communication device 20E, which is related to the communication volume of communication device 20A. The administrator screen SC may display a graph showing the current communication volume of each of communication devices 20A and 20E, or may display a graph showing the communication volume of each of communication devices 20A and 20E when a similar abnormality occurred in the past. The administrator checks the administrator screen SC and performs maintenance on communication devices 20A and 20E. The maintenance itself of each of communication devices 20A and 20E may be performed by a known method.
[0136] As described above, when an abnormality occurs in a communication service, the monitoring system 1 uses clustering to identify the communication device 20A that is related to the abnormality and the communication device 20E that is related to the abnormality. For example, when a similar abnormality occurs again, the monitoring system 1 makes it easier for the administrator to identify the communication device 20 that caused the abnormality, allowing for rapid recovery work. The monitoring system 1 will be described in detail below.
[0137] 2-3. Functions Realized in the Monitoring System FIG. 12 is a diagram showing an example of functions realized in the monitoring system 1. FIG. 12 shows functions realized in the server 10 and functions realized in the administrator terminal 40. The functions of the communication device 20 and the user terminal 30 are similar to those of known communication services, and therefore are omitted from FIG. 12. For example, the communication device 20 has a function to transmit status data or part of the status data to the server 10, which will be described later. The user terminal 30 has a function that enables the user to use the communication service.
[0138] [2-3-1. Functions Realized by the Server] For example, the server 10 includes a data storage unit 100, a status data acquisition unit 101, a clustering execution unit 108, a related data identification unit 109, and a recovery processing execution unit 106. The data storage unit 100 is realized by the storage unit 12. The status data acquisition unit 101, the clustering execution unit 108, the related data identification unit 109, and the recovery processing execution unit 106 are realized by the control unit 11.
[0139] [Data Storage Unit] The data storage unit 100 stores data related to communication services. For example, the data storage unit 100 stores a status database DB.
[0140] 13 is a diagram showing an example of a status database DB. The status database DB is a database in which status data for each of a plurality of communication devices 20 is stored. For example, the status database DB stores device identification information capable of identifying each of the plurality of communication devices 20, status data indicating the status of the communication device 20, abnormality cause identification information capable of identifying whether or not the communication device 20 corresponds to the abnormality cause data described below, and related identification information capable of identifying whether or not the communication device 20 corresponds to the related data described below. The status database DB may also store other data. For example, the status database DB may store information indicating the location where the communication device 20 is located, etc.
[0141] The identification information of the communication device 20 may be any information, such as an IP address, a device name, or a MAC address. In this embodiment, the status data is data relating to changes in the status of the communication device 20 over time. The status of the communication device 20 can also be referred to as the load on the communication device 20. The status of the communication device 20 may refer to the hardware status or the software status.
[0142] In this embodiment, an example is given in which the communication volume of the communication device 20 corresponds to the state of the communication device 20, but other states of the communication device 20 may also correspond to the state of the communication device 20. Examples of other states may include the amount of resource consumption of the communication device 20, the processing time (response time) for a request for a service, or the number of errors returned in response to a request. The state data may be indicators known as golden signal metrics or other indicators used in known benchmark tests. For example, the state data may be CPU usage, memory usage, power consumption, communication speed, temperature, or a combination thereof.
[0143] The status data may indicate a status related to performance of the communication service or other services. The other services are services that users use via the communication service, such as e-commerce services, travel reservation services, payment services, or financial services. For example, the status data may indicate the number of users currently connected to the communication device 20 (logged in to the communication service or other services), payments made by users via the communication device 20, or orders placed by users via the communication device 20. The greater the number of payments, payment amounts, number of orders, and order amounts, the greater the load on the communication device 20. Therefore, the number of payments, payment amounts, number of orders, and order amounts also correspond to the status of the communication device 20.
[0144] In this embodiment, the status data of each of the multiple communication devices 20 indicates changes in the status of the communication device 20 over time. For example, the status data includes the date and time when the status of the communication device 20 was acquired and a numerical value indicating the status of the communication device 20 at that date and time. The status of the communication device 20 may be expressed in other formats, such as characters, instead of numerical values. In this embodiment, the communication volume of the communication device 20 corresponds to the status of the communication device 20, so the status data includes the date and time when the communication volume of the communication device 20 was acquired and the communication volume of the communication device 20 at that date and time.
[0145] Note that the status data of each of the multiple communication devices 20 may indicate a status at a certain pinpoint (a single point in time) rather than a time-series change in the status of the communication device 20. For example, some status data may indicate a time-series change in the status, and other status data may indicate a status at a certain pinpoint. By collecting a large number of status data indicating a status at a certain pinpoint, a time-series change in the status may be indicated as a whole by the large number of status data.
[0146] Furthermore, the data stored in the data storage unit 100 is not limited to the status database DB. The data storage unit 100 may store any data related to the communication service. For example, the data storage unit 100 may store data necessary for displaying an administrator screen SC on the administrator terminal 40. The data storage unit 100 may store various thresholds used in the processing described below. The thresholds may be specified by the administrator or may be determined based on machine learning techniques or statistical techniques.
[0147] [Status Data Acquisition Unit] The status data acquisition unit 101 acquires status data relating to the status of each of the multiple communication devices 20 that are to be monitored in the communication service. For example, the status data acquisition unit 101 periodically (e.g., every 1 to 30 seconds, or every 1 to 5 minutes) requests the status of each of the multiple communication devices 20. When the communication device 20 receives a request from the status data acquisition unit 101 from the server 10, the communication device 20 transmits its own device identification information and information indicating its own status to the server 10. The status data acquisition unit 101 adds the current date and time and the information received from the communication device 20 to the status data associated with the device identification information of the communication device 20.
[0148] Note that the communication device 20 may accumulate status data for a certain period of time and transmit the status data to the server 10 all at once. In this case, the status data acquisition unit 101 acquires the status data all at once from the communication device 20 and stores it in the status database DB. Alternatively, for example, the status data acquisition unit 101 may acquire the status data from the communication device 20 irregularly rather than periodically. The status data acquisition unit 101 may acquire the status data from the communication device 20 at a timing instructed by an administrator.
[0149] The communication device 20 is an example of a device to be monitored in a communication service. The device to be monitored may be any device and is not limited to the communication device 20. In this embodiment, the description of the communication device 20 can be replaced with another device to be monitored. The other device may be a device used in the communication service, such as the server 10, a computer other than the server 10, components such as a CPU, memory, or network card included in the server 10 or another computer, a battery, an antenna, a communication cable, or a sensor.
[0150] In this embodiment, the status data acquisition unit 101 acquires status data when an abnormality is detected. The status data acquired when an abnormality is detected indicates the status during all or part of the period from when the abnormality occurs to when the abnormality is resolved. Hereinafter, this period will be referred to as the abnormality period. The status data may indicate not only all or part of the abnormality period, but also the status during periods other than the abnormality period. The status data acquisition unit 101 may acquire the status data during the period from when the abnormality occurs to when the abnormality is resolved, or may acquire the status data after the abnormality is resolved.
[0151] In this embodiment, the status data does not indicate the status at a specific pinpoint timing, but indicates the status over a certain period of time. The status data acquisition unit 101 may acquire status data indicating the status over the entire past period, but in this embodiment, it acquires status data indicating the status over a partial past period. For example, the status data acquisition unit 101 acquires status data regarding changes in the status of the device over time over a period including the time when an abnormality was detected.
[0152] [Clustering Execution Unit] The clustering execution unit 108 performs clustering on the state data of each of the multiple communication devices 20. Clustering is a process of identifying state data that have similar characteristics. Clustering is also called grouping. The clustering execution unit 108 performs clustering so that state data that have similar characteristics belong to the same cluster. Each of the multiple state data that belong to a certain cluster exhibits similar characteristics. The number of clusters may be predetermined, or may not be particularly determined.
[0153] The clustering method itself may be a known method. For example, the clustering execution unit 108 may perform clustering based on k-means, hierarchical clustering, density-based clustering, Gaussian mixture model, or spectral clustering. Alternatively, for example, the clustering execution unit 108 may perform clustering based on a machine learning model created using a machine learning method. The machine learning model may also be a model created using a known method. For example, the clustering execution unit 108 may perform clustering by inputting state data of each of the multiple communication devices 20 into a model such as a recurrent neural network (RNN), a long short-term memory (LSTM), or a transformer-based model.
[0154] FIG. 14 is a diagram showing an example of the results of clustering. FIG. 14 shows a vector space in which the state data of each of a large number of communication devices 20 has been converted into feature quantities. The white circles in FIG. 14 represent anomaly relation data, which will be described later. The black circles in FIG. 14 represent state data other than the anomaly relation data. The proximity of the circles in FIG. 14 indicates similar features. In this embodiment, since the state data indicates time-series state changes, the clustering execution unit 108 executes clustering based on the time-series state changes indicated by the state data of each of the multiple communication devices 20. For example, the clustering execution unit 108 calculates feature quantities of the waveforms indicated by the individual state data.
[0155] For example, the clustering execution unit 108 calculates feature quantities of the state data in the time domain. The clustering execution unit 108 calculates the average value, variance, standard deviation, minimum value, maximum value, rising timing, average amount of change, maximum amount of change, minimum amount of change, or a combination thereof, of the time-series state change indicated by certain state data as feature quantities of the state data in the time domain. The clustering execution unit 108 may also calculate the period during which the state indicated by certain state data is saturated, the communication volume at that time, the frequency of saturation, or a combination thereof, as feature quantities of the state data in the time domain. The clustering execution unit 108 may also calculate feature quantities of the state data in the frequency domain instead of the time domain.
[0156] For example, the clustering execution unit 108 performs clustering based on the feature values of each of the plurality of state data so that state data with similar feature values belong to the same cluster. When a machine learning model is used, the feature values of the state data are sometimes called embedded representations. The actual clustering process is performed within the machine learning model. The clustering execution unit 108 performs clustering by inputting each of the plurality of state data into the machine learning model and obtaining output from the machine learning model. For example, the output of the machine learning model is data indicating the relationship between each of the plurality of state data and the cluster to which the state data belongs. An unsupervised learning technique may be used for the clustering. The clustering execution unit 108 may perform clustering based on a technique of capturing waveforms indicated by the state data as images and determining whether the images are similar.
[0157] In this embodiment, when an abnormality is detected, the clustering execution unit 108 executes clustering based on the status data of each of the multiple devices. For example, the clustering execution unit 108 may execute clustering automatically when the server 10 detects an abnormality, or may execute clustering when the administrator of the communication service performs a predetermined operation on the administrator terminal 40. The clustering execution unit 108 may execute clustering after an abnormality is detected. An example in which clustering is executed before an abnormality is detected will be described later in Modification 2-1.
[0158] In the example of FIG. 14 , four pieces of anomaly-related data are specified by the administrator, and the clustering execution unit 108 identifies four clusters C1 to C4 through clustering. For example, the white circle (anomaly-related data) in cluster C1 is the status data of communication device 20A in the examples of FIGS. 10 and 11 . One of the black circles in cluster C1 is the status data of communication device 20E in the examples of FIGS. 10 and 11 . The white circles in each of the other clusters C2 to C4 are the status data of communication devices 20 that caused anomalies in areas other than area X, where communication device 20A caused the anomaly. For the other areas, the status data of other communication devices 20 related to the communication device 20 is also identified through clustering. The number of status data belonging to each of clusters C1 to C4 may be the same or different.
[0159] [Related Data Identification Unit] Based on the results of clustering, the related data identification unit 109 identifies related data, which is other status data related to the abnormality cause data, which is status data related to the cause of an abnormality in the service. In this embodiment, the administrator specifies which communication device 20's status data corresponds to the abnormality cause data. When the server 10 receives, from the administrator terminal 40, information indicating the status data of the communication device 20 designated by the administrator as the abnormality cause data, the server 10 updates the abnormality cause identification information associated with the device identification information of the communication device 20 in the status database DB. In the example of FIG. 13 , the server 10 updates the abnormality cause identification information of the communication device 20A so that the status data of the communication device 20A indicates that the status data is abnormality cause data related to the cause of the abnormality in area X.
[0160] The related data is status data that has been identified by clustering as having similar characteristics to the abnormality factor data. In other words, the related data is status data that belongs to the same cluster as the abnormality factor data. The related data identifying unit 109 identifies at least one piece of related data for one piece of abnormality factor data. If there is no status data that has similar characteristics to the abnormality factor data, the related data identifying unit 109 may not identify any related data. The related data identifying unit 109 updates the related information in the status database DB that is associated with the status data of the communication device 20 that has been identified as the related data. In the example of FIG. 13 , the related data identifying unit 109 identifies the status data of the communication device 20E as the related data, and therefore updates the related information associated with the status data of the communication device 20E.
[0161] The related data identifying unit 109 may identify, as related data, status data that has somewhat similar characteristics even if it does not belong to the same cluster as the abnormality factor data. For example, the related data identifying unit 109 may identify, as related data, status data that does not belong to the same cluster as the abnormality factor data but whose difference in feature amount (in the example of FIG. 14 , the distance in vector space) is less than a threshold. Furthermore, when calculating the difference in feature amount, the related data identifying unit 109 can identify related data that is closer to the more distinctive data by applying a weight (weight parameter) to a data point section with a more distinctive feature. Conversely, the related data identifying unit 109 may not identify, as related data, status data that belongs to the same cluster as the abnormality factor data. For example, the related data identifying unit 109 may identify, as related data, a predetermined number of status data from among multiple status data that belong to the same cluster as the abnormality factor data, in descending order of difference in feature amount from the abnormality factor data.
[0162] In this embodiment, the related data identification unit 109 identifies related data when an abnormality is detected. For example, the related data identification unit 109 may automatically identify related data when the server 10 detects an abnormality, or may identify related data when an administrator of the communication service performs a predetermined operation on the administrator terminal 40. The related data identification unit 109 may identify related data after an abnormality is detected. An example in which related data is identified before an abnormality is detected will be described in Modification 2-1 below.
[0163] [Recovery Processing Execution Unit] When an abnormality is detected, the recovery processing execution unit 106 executes recovery processing related to recovery from the abnormality based on the abnormality cause data and related data. The recovery processing may be any processing related to the abnormality that has occurred. In this embodiment, the recovery processing corresponds to processing for displaying status data related to the abnormality on the administrator screen SC. The recovery processing may be other processing, such as processing for executing a prepared recovery program or processing for sending a notification such as an email to the administrator.
[0164] For example, when an abnormality is detected, the recovery processing execution unit 106 transmits display data for the administrator screen SC to the administrator terminal 40 based on the abnormality cause data and the related data, thereby displaying the administrator screen SC on the administrator terminal 40. When an abnormality is detected, the recovery processing execution unit 106 displays images showing each of the abnormality cause data and the related data on the administrator screen SC.
[0165] 12, when a certain abnormality occurs, the recovery process execution unit 106 displays on the administrator screen SC the state indicated by the state data determined by the related data identification unit 109 when the same or similar abnormality occurred in the past. When a certain abnormality occurs and state data related to the abnormality is identified, the recovery process execution unit 106 executes recovery process based on the state data related to the abnormality if the same abnormality occurs again.
[0166] In the example of the status database DB in Figure 13, when an abnormality occurs in area X, the recovery process execution unit 106 refers to the abnormality cause identification information in the status database DB and identifies that the status data of communication device 20A is abnormality cause data that is the cause of the abnormality. Furthermore, the recovery process execution unit 106 refers to the related identification information in the status database DB and identifies that the status data of communication device 20E is related data. Based on the identified abnormality cause data and related data, the recovery process execution unit 106 generates display data for the administrator screen SC and transmits it to the administrator terminal 40.
[0167] [2-3-2. Functions Realized by the Administrator Terminal] For example, the administrator terminal 40 includes a data storage unit 400, an operation reception unit 401, and a display control unit 402. The data storage unit 400 is realized mainly by the storage unit 42. The operation reception unit 401 and the display control unit 402 are realized mainly by the control unit 41.
[0168] [Data Storage Unit] The data storage unit 400 stores data necessary for the administrator's work. For example, the data storage unit 400 stores data necessary for displaying the administrator screen SC. The data storage unit 400 may also store maintenance tools necessary for the administrator's work. The maintenance tools themselves may be publicly known tools, and may be, for example, tools capable of monitoring the status of at least one of hardware and software.
[0169] [Operation Acceptance Unit] The operation acceptance unit 401 accepts various operations by the administrator. For example, the operation acceptance unit 401 accepts operations on the administrator screen SC.
[0170] [Display Control Unit] The display control unit 402 displays various screens on the display unit 45. For example, the display control unit 402 displays an administrator screen SC on the display unit 45.
[0171] 15 is a diagram showing an example of processing executed in the monitoring system 1. The control units 11 and 41 execute programs stored in the storage units 12 and 42, respectively, to execute the processing in FIG.
[0172] As shown in FIG. 15 , the server 10 communicates with each of the multiple communication devices 20, acquires status data, and stores the acquired data in the status database DB (S100). The process of S100 is executed periodically. When an abnormality occurs, the administrator terminal 40 accepts an abnormality occurrence operation from the administrator (S101). The administrator terminal 40 transmits abnormality occurrence data indicating that an abnormality occurrence operation has been performed to the server 10 (S102). The server 10 receives the abnormality occurrence data from the administrator terminal 40 (S103). The server 10 records the timing at which the abnormality occurrence data is received as the abnormality occurrence timing in the storage unit 12 (S104). The administrator may specify the abnormality occurrence timing. In this case, the server 10 records the abnormality occurrence timing specified by the administrator in the storage unit 12.
[0173] When the abnormality is resolved, the administrator terminal 40 accepts an abnormality resolution operation by the administrator (S105). The administrator terminal 40 transmits abnormality resolution data indicating that an abnormality resolution operation has been performed to the server 10 (S106). The server 10 receives the abnormality resolution data from the administrator terminal 40 (S107). The server 10 records the timing at which the abnormality resolution data is received in the storage unit 12 as the abnormality resolution timing at which the abnormality was resolved (S108). The administrator may also specify the abnormality resolution timing. In this case, the server 10 records the abnormality resolution timing specified by the administrator in the storage unit 12.
[0174] The administrator analyzes the status data of each of the multiple communication devices 20 and identifies the abnormality factor data. When the administrator specifies the abnormality factor data, the administrator terminal 40 transmits abnormality factor identification information capable of identifying the abnormality factor data to the server 10 (S109). In the example of Fig. 10 etc., the administrator terminal 40 transmits abnormality factor identification information to the server 10 indicating that the status data of the communication device 20A is abnormality factor data.
[0175] When the server 10 receives the abnormality cause identification information from the administrator terminal 40 (S110), it updates the status database DB (S111). In the example of FIG. 13, the server 10 updates the abnormality cause identification information associated with the device identification information of the communication device 20A in the status database DB. The server 10 performs clustering based on the status data of each of the multiple communication devices 20 (S112). The server 10 identifies associated data that is related to the abnormality cause data based on the clustering results (S113), and this process ends. If a similar abnormality occurs thereafter, the server 10 performs recovery processing based on the abnormality cause data and the associated data.
[0176] [2-5. Summary of the embodiment] The monitoring system 1 of this embodiment acquires status data relating to the status of each of the multiple communication devices 20. The monitoring system 1 performs clustering on the status data of each of the multiple communication devices 20. The monitoring system 1 identifies related data based on the results of the clustering, and therefore can identify the relevance of the status data of the communication devices 20. For example, even if the administrator is unaware of the existence of related data, the monitoring system 1 can identify the related data through clustering, allowing the administrator to quickly recover from an abnormality. The monitoring system 1 can support the administrator's work.
[0177] Furthermore, when an abnormality is detected, the monitoring system 1 acquires status data for each of the multiple communication devices 20. When an abnormality is detected, the monitoring system 1 performs clustering based on the status data for each of the multiple communication devices 20. When an abnormality is detected, the monitoring system 1 identifies related data. When an abnormality is detected, the monitoring system 1 performs recovery processing based on the abnormality cause data and related data. The recovery processing enables the monitoring system 1 to quickly recover from an abnormality in a communication service. For example, when a similar abnormality occurs again, the monitoring system 1 can have the administrator check the abnormality cause data and related data, thereby effectively supporting the administrator's work.
[0178] Furthermore, the monitoring system 1 performs clustering based on time-series state changes indicated by the state data of each of the multiple communication devices 20. The monitoring system 1 can identify the association of the state data from the time-series changes in the state of the communication devices 20. For example, the monitoring system 1 can identify, as associated data, state data that has similar time-series characteristics to the anomaly cause data, thereby performing more accurate recovery processing based on the time-series changes in the state data.
[0179] [2-6. Modifications] The present disclosure is not limited to the above-described embodiments. The present disclosure can be modified as appropriate without departing from the spirit of the present disclosure.
[0180] 16 is a diagram showing an example of functions realized by a modified monitoring system 1. The modified monitoring system 1 includes a detection process execution unit 110, a saturation determination unit 104, and an abnormality factor data identification unit 111. The detection process execution unit 110, the saturation determination unit 104, and the abnormality factor data identification unit 111 are realized by the control unit 11.
[0181] [2-6-1. Modification 2-1] For example, in the embodiment, a case where a series of processes is executed when an abnormality is detected has been described, but the series of processes may be executed before the abnormality is detected. In Modification 2-1, a case where the related data identified by the related data identification unit 109 is used for detecting the abnormality, rather than for recovering from the abnormality as in the embodiment, will be described.
[0182] The status data acquisition unit 101 in Modification 2-1 acquires status data of each of the multiple communication devices 20 before an abnormality is detected. The method of acquiring status data by the status data acquisition unit 101 is the same as in the embodiment. The status data acquisition unit 101 acquires status data in real time while the communication service is in operation. The status data acquisition unit 101 acquires status data regardless of whether an abnormality has occurred. For example, even if an abnormality has actually occurred, if the abnormality has occurred before it is detected by the monitoring system 1, this corresponds to the above-mentioned "before the abnormality is detected."
[0183] The clustering execution unit 108 of Modification 2-1 executes clustering based on the status data of each of the plurality of communication devices 20 before an abnormality is detected. The method of executing clustering is also the same as in the embodiment. The clustering execution unit 108 executes clustering regardless of whether an abnormality has occurred. The related data identification unit 109 of Modification 2-1 identifies related data before an abnormality is detected. The method of identifying related data is also the same as in the embodiment. The related data identification unit 109 identifies related data regardless of whether an abnormality has occurred.
[0184] The monitoring system 1 includes a detection process execution unit 110. The detection process execution unit 110 executes a detection process for detecting an abnormality based on the abnormality cause data and related data. The detection process may be any process for detecting an abnormality that has occurred. In variant example 2-1, the detection process corresponds to a process for displaying the occurrence of an abnormality on the administrator screen SC. For example, when an abnormality is detected, the detection process execution unit 110 displays an image indicating that an abnormality has been detected on the administrator screen SC. The detection process execution unit 110 may execute the detection process by sending a notification to the administrator by means of email or the like.
[0185] In Modification 2-1, a criterion indicating whether an abnormality has occurred is associated with each of the abnormality cause data and the related data. The criterion is determined based on abnormalities that have occurred in the past. For example, the criterion may be a threshold value, or, as in Modification 2-2 described below, may be whether saturation has occurred. The detection process execution unit 110 determines whether each of the abnormality cause data and the related data satisfies the criterion.
[0186] For example, the detection process execution unit 110 executes the detection process when it determines that at least one of the abnormality factor data and the related data satisfies the criterion. Note that the detection process execution unit 110 may execute the detection process when it determines that both the abnormality factor data and the related data satisfy the criterion. The detection process execution unit 110 may execute the detection process when it determines that only either the abnormality factor data or the related data satisfies the criterion.
[0187] The monitoring system 1 of Variation 2-1 acquires status data for each of the multiple communication devices 20, performs clustering, and identifies related data before detecting an abnormality. The monitoring system 1 executes a detection process for detecting an abnormality based on the abnormality cause data and the related data. This allows the monitoring system 1 to quickly detect an abnormality.
[0188] [2-6-2. Modification 2-2] For example, in the embodiment, an administrator analyzes status data and specifies abnormality cause data. The status data may be analyzed by a monitoring system and classified as abnormality cause data. For example, in a communication service, a large number of communication devices 20 exist and various abnormalities can occur, making it difficult for the administrator to identify the communication device 20 that caused the abnormality. In this case, the communication device 20 that caused the abnormality may be unable to process any more communications, and the communication volume may become saturated.
[0189] Saturation of communication volume means that the communication volume is positive (not zero) and does not change or changes almost unchanged. No change in communication volume means that the change in communication volume is zero. Almost no change in communication volume means that the change in communication volume is less than a threshold. Specific methods for determining saturation will be described later. An upper limit for communication volume may be set, but there are cases where the communication device 20 can handle a communication volume greater than the upper limit set by the administrator, or where the communication device 20 can only handle a communication volume lower than the upper limit set by the administrator, and even where the administrator himself does not know the upper limit for the communication device 20. Therefore, it is assumed that no upper limit for communication volume is set. Therefore, even if the communication volume of the communication device 20 reaches a certain level, an abnormality is not detected.
[0190] In the example of Figure 10, the communication volume of communication device 20A increases sharply, and then the communication volume of communication device 20A reaches saturation. During this period, communication device 20A cannot handle any more communications, so a user terminal 30 attempting to connect to communication device 20A in a specific area may be unable to use the communication service. Note that in variant example 2-2, the period from the saturation start timing to the saturation end timing is treated as an example of a period in which an abnormality has occurred. However, the period from the saturation start timing to the present time when saturation has not yet ended may also be treated as an example of a period in which an abnormality has occurred. For example, the period from the saturation start timing to a time specified by the administrator may also be treated as an example of a period in which an abnormality has occurred.
[0191] The monitoring system 1 of Modification 2-2 includes a saturation determination unit 104 and an abnormality factor data identification unit 111. The saturation determination unit 104 determines whether the state indicated by the state data of each of the multiple communication devices 20 is saturated, based on the state data of the communication devices 20. The saturation determined by the saturation determination unit 104 occurs when the amount of change in the state indicated by the state data falls below a threshold, or when the amount of change remains below the threshold for a predetermined period of time or more. In cases where there is no upper limit, such as with communication volume, as communication volume increases, the amount of change primarily refers to the amount of increase. Therefore, saturation occurs when the total amount of change increases over a certain period of time, and then the amount of increase in the state indicated by the state data falls below a threshold, or when the amount of increase remains below the threshold for a predetermined period of time or more. The amount of change may be an integrated value over a certain period of time, or the amount of change between one point in time and the next. The amount of change may also refer to the amount of decrease in communication volume. In other words, the saturation determination unit 104 determines saturation based on the time-series change in the amount of change itself indicated by the state data.
[0192] For example, the saturation determination unit 104 determines that the time-series state of the communication device 20 is saturated when the increase in the state indicated by the state data changes from a state where the increase is equal to or greater than a threshold to a state where the increase is less than the threshold, or a state where the increase is less than the threshold continues for a predetermined period of time or more. Even if the increase in the state indicated by the state data changes to a state where the increase is less than the threshold, or a state where the increase is less than the threshold continues for a predetermined period of time or more, the saturation determination unit 104 determines that the time-series state of the communication device 20 is not saturated if the increase was less than the threshold before that. For example, the saturation determination unit 104 may determine that the time-series state of the communication device 20 is saturated when the state indicated by the state data is equal to or greater than a threshold and the increase in the state changes to a state where the increase is less than the threshold, or a state where the increase is less than the threshold continues for a predetermined period of time or more. The saturation determination unit 104 determines that the time-series state of the communication device 20 is not saturated if the increase in the state indicated by the state data is less than a threshold value, or if the increase in the state less than the threshold value continues for a predetermined period of time or more.
[0193] For example, the saturation determination unit 104 determines whether the time-series state of a communication device 20 is saturated based on the state data of each of the multiple communication devices 20 and a predetermined threshold. In other words, the saturation determination unit 104 determines whether the time-series state indicated by the state data of each communication device 20 is saturated based on the predetermined threshold. The threshold referenced by the saturation determination unit 104 when determining saturation may be specified by an administrator or may be calculated from past abnormality trends. The threshold may also be determined based on a model using a machine learning technique.
[0194] The saturation determination unit 104 may determine whether the state of the communication device 20 is saturated based on a model that uses a machine learning technique, without using a threshold value. In this case, the model is trained with training data that receives as input time-series state changes indicated by training state data and outputs labels indicating whether the state is saturated. The saturation determination unit 104 may input the state data to be determined to the trained model, and determine whether the state of the communication device 20 is saturated based on the labels output from the model.
[0195] For example, the saturation determination unit 104 determines whether the time-series state of a communication device 20 is saturated by determining whether the amount of change indicated by the state data of the communication device 20 is equal to or greater than a threshold. If the saturation determination unit 104 determines that the amount of change indicated by the state data of the communication device 20 is equal to or greater than the threshold, the saturation determination unit 104 does not determine that the time-series state of the communication device 20 is saturated. If the saturation determination unit 104 determines that the amount of change indicated by the state data of the communication device 20 is less than the threshold, the saturation determination unit 104 determines that the time-series state of the communication device 20 is saturated.
[0196] For example, the saturation determination unit 104 determines whether the time-series state of a communication device 20 is saturated by determining whether a state in which the amount of change indicated by the status data of the communication device 20 is less than a threshold has continued for a predetermined time or more. If the saturation determination unit 104 determines that a state in which the amount of change indicated by the status data of the communication device 20 is less than a threshold has not continued for a predetermined time or more, the saturation determination unit 104 does not determine that the time-series state of the communication device 20 is saturated. If the saturation determination unit 104 determines that a state in which the amount of change indicated by the status data of the communication device 20 is less than a threshold has continued for a predetermined time or more, the saturation determination unit 104 determines that the time-series state of the communication device 20 is saturated.
[0197] For example, when an abnormality is detected, the saturation determination unit 104 determines whether the state of the communication device 20 is saturated based on the state data of the communication device 20. That is, after an abnormality occurs, the saturation determination unit 104 determines whether the state of the communication device 20 is saturated based on the state data of the communication device 20. For example, the saturation determination unit 104 determines whether the state of the communication device 20 is saturated based on the state data on which normalization has been performed. The saturation determination unit 104 determines whether the state of the communication device 20 is saturated based on the state indicated by the normalized state data.
[0198] The saturation determination unit 104 may determine whether a state indicated by the state data is saturated based on a state within a predetermined time window. The saturation determination unit 104 may determine whether a state indicated by the state data is saturated regardless of the time window. For example, the saturation determination unit 104 may determine whether a state indicated by the state data is saturated based on a state within a reference period. The saturation determination unit 104 may determine whether a state indicated by the state data is saturated based on a state within an abnormal period in which an abnormality occurs. The saturation determination unit 104 may determine whether a state indicated by the state data is saturated based on a state throughout the entire period indicated by the state data.
[0199] The abnormality factor data identifying unit 111 identifies, as abnormality factor data, the status data of the communication device 20 that is determined to be in a saturated state, from among the status data of each of the multiple communication devices 20. In the example of Fig. 10, the status of the communication device 20A indicates a tendency toward saturation, so the abnormality factor data identifying unit 111 identifies the status data of the communication device 20A as abnormality factor data. The processing after the abnormality factor data is identified is the same as in the embodiment.
[0200] The monitoring system 1 of Variation 2-2 determines whether the state of a communication device 20 is saturated based on the state data of each of the multiple communication devices 20. The monitoring system 1 identifies, among the state data of each of the multiple communication devices 20, state data determined to be saturated as abnormality cause data. This allows the monitoring system 1 to determine whether the state data is related to an abnormality in the communication service. As a result, for example, the monitoring system 1 can speed up recovery from the abnormality. In addition, for example, the monitoring system 1 can quickly detect an abnormality and analyze the cause of the abnormality.
[0201] [2-6-3. Modification 2-3] For example, the status data of each of the multiple communication devices 20 is generated based on the time managed by the communication device 20. The times managed by each of the multiple communication devices 20 may be different from each other. In this case, the time axes of the status data of each of the multiple communication devices 20 may be different from each other. For this reason, the clustering execution unit 108 may align the time axes of the status data of each of the multiple communication devices 20, and then perform clustering based on the status data with the aligned time axes.
[0202] For example, the clustering execution unit 108 acquires the current time managed by each of the multiple communication devices 20 from that communication device 20. For each communication device 20, the clustering execution unit 108 calculates the difference between the current time managed by the server 10 and the current time managed by that communication device 20. For each communication device 20, the clustering execution unit 108 aligns the time axes of the multiple communication devices 20 by shifting the time axis of that communication device 20 so that the difference is smaller than a threshold value (for example, so that the difference becomes zero).
[0203] For example, suppose the clustering execution unit 108 determines that the time managed by communication device 20A is 15 seconds ahead of the time managed by server 10. In this case, the clustering execution unit 108 changes the time axis indicated by the status data of communication device 20A so that it is 15 seconds behind overall. In another example, suppose the clustering execution unit 108 determines that the time managed by communication device 20B is 20 seconds behind the time managed by server 10. In this case, the clustering execution unit 108 changes the time axis indicated by the status data of communication device 20B so that it is 20 seconds ahead overall.
[0204] Although the example in Modification 2-3 uses the time managed by the server 10 as the reference, the clustering execution unit 108 may use the time managed by any of the multiple communication devices 20 as the reference. For example, the clustering execution unit 108 may use the time managed by communication device 20A as the reference. In this case, if the clustering execution unit 108 determines that the time managed by communication device 20B is 32 seconds ahead of the time managed by communication device 20A, it changes the time axis indicated by the status data of communication device 20B so that it is 32 seconds behind overall. If the clustering execution unit 108 determines that the time managed by communication device 20C is 26 seconds behind the time managed by communication device 20A, it changes the time axis indicated by the status data of communication device 20C so that it is 26 seconds ahead overall. The clustering execution unit 108 does not change the time axis of communication device 20A.
[0205] The monitoring system 1 of Variation 2-3 aligns the time axes of the status data of each of the multiple communication devices 20, and then performs clustering based on the status data with the aligned time axes. If the time axes of the status data of each of the multiple communication devices 20 are misaligned, even if the status data originally had similar characteristics, the misalignment of the time axes of the status data may result in the status data being determined not to have similar characteristics. In this regard, the monitoring system 1 can increase the accuracy of identifying related data by aligning the time axes of the status data of each of the multiple communication devices 20.
[0206] [2-6-4. Modification 2-4] For example, even if there is a correlation between the status data of communication device 20A and the status data of communication device 20B, a time lag may occur between these correlations, such as a sudden increase in the communication volume of communication device 20B 30 seconds after a sudden increase in the communication volume of communication device 20A. In this case, when the communication volumes at the same time are compared by clustering, they may be determined to have dissimilar characteristics. If the clustering execution unit 108 advances the waveform indicated by the status data of communication device 20B by 30 seconds overall and then compares it with the status data of communication device 20A, it becomes easier to determine that the overall waveform characteristics are similar. For this reason, the clustering execution unit 108 may shift the time axis of each of the status data of multiple communication devices 20 and then perform clustering based on the time-shifted status data.
[0207] In Modification 2-4, the shift amount, which indicates how much the time axis is shifted, is assumed to be predetermined. Furthermore, multiple shift amounts are assumed to be defined as the amount of shift on the time axis. For example, six shift amounts are assumed to be defined: 10 seconds, 20 seconds, 30 seconds, 40 seconds, 50 seconds, and 60 seconds. In the above example, if the status data of communication device 20A is abnormality cause data, the clustering execution unit 108 advances the time axis of the status data of communication device 20B by 10 seconds and then performs clustering. Furthermore, the clustering execution unit 108 delays the time axis of the status data of communication device 20B by 10 seconds and then performs clustering.
[0208] Thereafter, the clustering execution unit 108 similarly advances the time axis of the status data of communication device 20B by 20 seconds, 30 seconds, 40 seconds, 50 seconds, and 60 seconds, and then performs clustering. Furthermore, the clustering execution unit 108 delays the time axis of the status data of communication device 20B by 20 seconds, 30 seconds, 40 seconds, 50 seconds, and 60 seconds, and then performs clustering. The clustering execution unit 108 similarly advances or delays the time axis of the status data of other communication devices 20 other than communication device 20C, etc., by a predetermined shift amount, and then performs clustering.
[0209] The related data identifying unit 109 of Modification 2-4 identifies related data based on the results of clustering performed after the time axis is shifted as described above. For example, the related data identifying unit 109 identifies, as related data, status data classified into the same cluster as the abnormality factor data by clustering based on any of the multiple shift amounts. The related data identifying unit 109 may also identify, as related data, status data classified into the same cluster as the abnormality factor data by clustering based on each of the multiple shift amounts, with a number of clusterings equal to or greater than a reference number.
[0210] In the above example, the clustering execution unit 108 determines that the status data of communication device 20A and the status data of communication device 20B are similar by processing when the time axis of the status data of communication device 20B is shifted forward by 30 seconds overall. Therefore, even if there is a time lag as described above, the clustering execution unit 108 can classify the status data of communication device 20A and the status data of communication device 20B into the same cluster. The clustering execution unit 108 classifies into the same cluster the status data that are determined to be similar to each other with respect to at least one of the multiple shift amounts.
[0211] The monitoring system 1 of Modification 2-4 shifts the time axis of the status data of each of the multiple communication devices 20, and then performs clustering based on the status data with the shifted time axis. Even if there is a time lag between the abnormality cause data and the related data that are correlated with each other, the monitoring system 1 can absorb the time lag by deliberately shifting the time axis and identify the related data that is correlated with the abnormality cause data.
[0212] [2-6-5. Modification 2-5] For example, in Modification 2-4, the clustering execution unit 108 may determine the amount of shift on the time axis based on a portion of a period of time-series state changes indicated by the status data of each of the multiple communication devices 20, and may shift the time axis of the status data of each of the multiple communication devices 20 based on the determined amount of shift. In Modification 2-5, the clustering execution unit 108 determines which of the multiple shift amounts to use by using only a portion of the period indicated by the status data.
[0213] For example, the clustering execution unit 108 determines a characteristic period indicated by the anomaly cause data as the partial period. The characteristics serving as the basis for determining the period are assumed to be predetermined. In the example of FIG. 10 , the status data of the communication device 20A corresponds to the anomaly cause data. Furthermore, if the timing at which the communication volume of the communication device 20A suddenly increased indicates a characteristic of an anomaly, the clustering execution unit 108 determines a period including that timing as the partial period. The period may also be manually specified by an administrator.
[0214] For example, the clustering execution unit 108 compares, for each of the multiple shift amounts, the state change in the state data of the communication device 20A for the partial period with the state change in the state data of the other communication devices 20 for the partial period. The clustering execution unit 108 determines, from the multiple shift amounts, the shift amount that results in the smallest difference in state change as the shift amount on the time axis. Based on the determined shift amount, the clustering execution unit 108 executes the process described in Modification 2-4 and performs clustering. Clustering based on the shift amount may be similar to Modification 2-4.
[0215] The monitoring system 1 of Variation 2-5 determines the amount of shift on the time axis based on a partial period of time-series state changes indicated by the status data of each of the multiple communication devices 20, and shifts the time axis of each of the status data of the multiple communication devices 20 based on the determined amount of shift. Even if there is a time lag between the abnormality cause data and the related data that are correlated with each other, the monitoring system 1 can absorb the time lag and identify the related data that is correlated with the abnormality cause data by shifting the time axis by an optimal amount of shift. Because the monitoring system 1 does not need to perform processing based on unnecessary shift amounts, the processing load can be reduced.
[0216] [2-6-6. Modification 2-6] For example, the clustering execution unit 108 may calculate the difference between the values indicated by the status data of each of the multiple communication devices 20, and perform clustering based on the calculated difference. Taking FIG. 10 as an example, the clustering execution unit 108 calculates the difference between the anomaly cause data, which is the status data of the communication device 20A, and the status data of each of the communication devices 20B to 20E. A known index such as the Mahalanobis distance may be used to determine the difference. The clustering execution unit 108 classifies the status data for which the change in the difference is less than a threshold into the same cluster as the anomaly cause data.
[0217] In the example of Figure 10, the difference between the abnormality factor data, which is the status data of communication device 20A, and the status data of communication device 20E continues to fall within a certain range, so the clustering execution unit 108 classifies the abnormality factor data and the status data of communication device 20E into the same cluster. In other words, even if there is a certain degree of difference from the abnormality factor data, the clustering execution unit 108 classifies the status data into the same cluster as the abnormality factor data as long as the difference falls within a certain range. Note that the clustering execution unit 108 may perform clustering based only on the difference, rather than the amount of change. The processing after clustering is performed is as described in the embodiment.
[0218] The monitoring system 1 of Modification 2-6 calculates the difference between the values indicated by the status data of each of the multiple communication devices 20, and performs clustering based on the calculated difference. This allows the monitoring system 1 to perform clustering that focuses on the difference indicated by the status data of each of the multiple communication devices 20, thereby improving the accuracy of the clustering.
[0219] [2-6-7. Modification 2-7] For example, depending on the communication device 20, some of the state data may not be acquired due to a so-called node down such as a forced restart or shutdown. In this case, the state data acquisition unit 101 may acquire state data so that all or some of the communication devices 20 for which some of the state data has not been acquired among the multiple communication devices 20 indicate a predetermined value. The predetermined value may be a value specified in advance by an administrator, or may be dynamically calculated on the spot.
[0220] In the example of FIG. 10 , assume that a portion of the status data of communication device 20B is lost due to a forced restart of communication device 20B. In this case, the status data acquisition unit 101 may fill in the missing portion with the average value of the remaining portion of the status data. The status data acquisition unit 101 may also fill in the missing portion of the status data with a value specified by an administrator. Filling in the missing portion is sometimes called padding. The method of filling in the missing portion itself can be a known padding method.
[0221] The monitoring system 1 of Modification 2-7 acquires status data so that all or part of the status data of communication devices 20 for which some of the status data has not been acquired indicates a predetermined value among the multiple communication devices 20. This allows the monitoring system 1 to perform clustering even if some of the status data is missing.
[0222] [2-6-8. Modification 2-8] For example, the clustering execution unit 108 may perform clustering based on the status data of each of the multiple communication devices 20 and the attributes associated with the status data. The attributes may be any information that allows the status data to be classified from some perspective. In Modification 2-8, an example is given in which the attributes indicate the locations where each of the multiple communication devices 20 is located. The location may be a relatively broad area such as a city, town, or village, or a relatively narrow area such as a building, room, or rack where the communication device 20 is located. The attributes are not limited to the locations where the communication devices 20 are located, but may also be, for example, the performance, role, or type of the communication device 20. The attributes may be the content of the status data or acquisition conditions such as the sampling period of the status data. Data indicating the attributes of the communication devices 20 is assumed to be pre-stored in the data storage unit 100.
[0223] For example, the clustering execution unit 108 performs clustering based on the state data of each of the multiple communication devices 20 and the locations associated with the state data. The clustering execution unit 108 performs clustering based on the state data of each of the multiple communication devices 20 that are located in the same location. Even if the characteristics of the state data of each of the multiple communication devices 20 are similar, the clustering execution unit 108 prevents communication devices 20 that are located in different locations from belonging to the same cluster. The clustering execution unit 108 performs clustering only among the multiple communication devices 20 that are located in the same location.
[0224] Note that clustering when an attribute other than location is used may also be performed in a similar manner to the above. For example, the clustering execution unit 108 performs clustering based on the status data of multiple communication devices 20 associated with the same attribute. Even if the characteristics of the status data of multiple communication devices 20 are similar, the clustering execution unit 108 prevents the status data of communication devices 20 associated with different attributes from belonging to the same cluster. The clustering execution unit 108 performs clustering only among multiple communication devices 20 associated with the status data of the same attribute. The processing after clustering is performed is the same as in the embodiment.
[0225] The monitoring system 1 of Variation 2-8 performs clustering based on the status data of each of the multiple communication devices 20 and the attributes associated with the status data. This allows the monitoring system 1 to perform clustering that takes into account the attributes of the status data, thereby improving the accuracy of the clustering. For example, the monitoring system 1 can accurately identify the associations between the status data by performing clustering among multiple communication devices 20 whose status data are associated with the same attributes.
[0226] Furthermore, the monitoring system 1 performs clustering based on the status data of each of the multiple communication devices 20 and the locations associated with the status data. This allows the monitoring system 1 to perform clustering that takes into account the locations of the communication devices 20, thereby improving the accuracy of the clustering. For example, the monitoring system 1 can accurately identify the associations between the status data of multiple communication devices 20 by performing clustering among the communication devices 20 that are located in the same location.
[0227] [2-6-9. Modification 2-9] For example, in Modification 2-9, the clustering execution unit 108 may average time series data based on the status data of each of the multiple communication devices 20 at predetermined time intervals. For example, if status data is acquired at a frequency of 1 item per second, the quantity of status data will be 3,600 items per 60 minutes. If the status data is averaged at 1-minute intervals, the 3,600 items per 60 minutes of status data will be reduced to 60 items. By averaging the status data in advance, the clustering execution unit 108 of Modification 2-9 can create lightweight time series data that is similar to the time series data before averaging, thereby reducing the processing load of clustering.
[0228] [2-6-10. Other Modifications] For example, the above modifications may be combined.
[0229] For example, the monitoring system 1 may detect an abnormality in a service based on posts on a social networking service (SNS). In this case, the monitoring system 1 acquires post data related to posts on the SNS. The post data may be text, images, videos, hashtags, or a combination thereof. The monitoring system 1 detects an abnormality based on the number of posts including the name of the service being monitored. The monitoring system 1 detects an abnormality when the number of posts exceeds a threshold. The related data identifying unit 109 may identify a correlation between the post data on the SNS and the status data of the communication device 20. For example, when there is a correlation between the timing when the number of posts on the SNS increases and the timing when the load indicated by the status data of the communication device 20 increases, the related data identifying unit 109 may identify a correlation between the post data on the SNS and the status data of the communication device 20.
[0230] For example, the monitoring system 1 may be applied to services other than communication services. For example, the monitoring system 1 may identify associated data related to abnormality cause data that may be a cause of an abnormality in other services such as e-commerce services, travel reservation services, financial services, payment services, online flea market services, or video streaming services. The monitored devices may be devices used in these services and are not limited to communication devices as in the embodiment. For example, the monitored devices may be server computers, personal computers, tablets, power supplies, printers, scanners, or other devices.
[0231] For example, in the embodiment, the main processing is described as being executed on the server 10, but the processing described as being executed on the server 10 may be executed on the administrator terminal 40 or another computer, or may be shared among multiple computers.
[0232] [3. Supplementary Note] For example, the monitoring system can also be configured as follows.
[0233] [3-1. Supplementary notes regarding the first embodiment] (1-1) A monitoring system including: a status data acquisition unit that acquires status data related to time-series status changes of a device to be monitored in a service; a saturation determination unit that determines whether the status is saturated based on the status data; and a relationship determination unit that determines whether the status data is related to an abnormality in the service based on a determination result of the saturation determination unit. (1-2) The status data acquisition unit acquires the status data when the abnormality is detected, the saturation determination unit determines whether the status is saturated based on the status data when the abnormality is detected, and the relationship determination unit determines whether the status data is related to the abnormality when the abnormality is detected, and the monitoring system further includes a recovery process execution unit that executes recovery process for recovering from the abnormality based on the status data determined to be related to the abnormality when the abnormality is detected. (1-3) The monitoring system according to (1-1) or (1-2), wherein the status data acquisition unit acquires the status data before the abnormality is detected, the saturation determination unit determines whether the status is saturated based on the status data before the abnormality is detected, and the relationship determination unit determines whether the status data is related to the abnormality before the abnormality is detected, and the monitoring system further includes a detection process execution unit that executes detection process related to the detection of the abnormality based on the status data determined to be related to the abnormality. (1-4) The monitoring system according to any of (1-1) to (1-3), wherein the status data acquisition unit acquires the status data of the device in each of the plurality of services, the saturation determination unit determines whether the device is saturated based on the status data of the device in each of the plurality of services, and the relationship determination unit determines whether the status data is related to the abnormality in each of the plurality of services based on the determination result of the saturation determination unit.(1-5) The monitoring system according to any of (1-1) to (1-4), further including a period setting unit that sets a period related to the status data based on an abnormality occurrence timing when the abnormality is detected, and the status data acquisition unit acquires the status data related to a time-series change in the status of the device during the period. (1-6) The monitoring system according to any of (1-1) to (1-5), further including a normalization execution unit that performs normalization on the status data, and the saturation determination unit determines whether the status is saturated based on the normalized status data. (1-7) The monitoring system according to any of (1-1) to (1-6), further including a period setting unit that sets a period related to the status data based on an abnormality occurrence timing when the abnormality is detected, and the status data acquisition unit acquires the status data related to a time-series change in the status of the device during the period. (1-8) The monitoring system according to any one of (1-1) to (1-7), wherein the relationship determination unit determines whether the status data is related to the abnormality based on a maximum value of the status in another period prior to an abnormal period in which the abnormality occurs, among the periods in which the change is indicated in the status data. (1-9) The monitoring system according to (1-8), wherein the other period is a period prior to the abnormal period, and the relationship determination unit determines whether the status data is related to the abnormality based on a maximum value of the status in the other period and a maximum value of the status in a period prior to the other period. (1-10) The monitoring system according to any one of (1-1) to (1-9), wherein the relationship determination unit determines whether the status data is related to the abnormality based on a maximum value of the status indicated by the status data in an abnormal period in which the abnormality occurs, among the periods in which the change is indicated in the status data. (1-11) The monitoring system according to any one of (1-1) to (1-10), wherein the relationship determination unit determines whether the status data is related to the abnormality based on the degree of coincidence between an abnormal period in which the abnormality occurs and a saturation period in which the status is determined to be saturated, among the periods in which the change is indicated in the status data.(1-12) The monitoring system according to any one of (1-1) to (1-11), wherein the relationship determination unit determines whether the status data is related to the abnormality based on an amount of change in the status during a saturation period in which the status is saturated and during another period prior to the abnormality period in which the abnormality occurs, among the periods in which the change is indicated in the status data.
[0234] [3-2. Supplementary notes regarding the second embodiment] (2-1) A monitoring system comprising: a status data acquisition unit that acquires status data related to the status of each of a plurality of devices to be monitored in a service; a clustering execution unit that performs clustering on the status data of each of the plurality of devices; and a related data identification unit that identifies, based on a result of the clustering execution, related data, which is other status data that is related to abnormality cause data, which is the status data related to a cause of an abnormality in the service. (2-2) The monitoring system described in (2-1), wherein the status data acquisition unit acquires the status data of each of the plurality of devices when the abnormality is detected, the clustering execution unit performs the clustering based on the status data of each of the plurality of devices when the abnormality is detected, and the related data identification unit identifies the related data when the abnormality is detected, and the monitoring system further includes a recovery process execution unit that performs recovery process for recovering from the abnormality based on the abnormality cause data and the related data when the abnormality is detected. (2-3) The monitoring system according to (2-1) or (2-2), wherein the status data acquisition unit acquires the status data of each of the plurality of devices before the abnormality is detected, the clustering execution unit executes the clustering based on the status data of each of the plurality of devices before the abnormality is detected, the associated data identification unit identifies the associated data before the abnormality is detected, and the monitoring system further includes a detection process execution unit that executes a detection process related to the detection of the abnormality based on the abnormality cause data and the associated data. (2-4) The monitoring system according to any of (2-1) to (2-3), wherein the status data of each of the plurality of devices indicates a change in the status of the device over time, and the clustering execution unit executes the clustering based on the change in the status over time indicated by the status data of each of the plurality of devices.(2-5) The monitoring system according to (2-4), further including: a saturation determination unit that determines whether the state indicated by the status data of each of the plurality of devices is saturated, based on the status data of the plurality of devices; and an abnormality factor data identification unit that identifies, from the status data of each of the plurality of devices, the status data determined to be in a saturated state as the abnormality factor data. (2-6) The monitoring system according to (2-4) or (2-5), wherein the clustering execution unit aligns time axes of the status data of each of the plurality of devices and then performs the clustering based on the status data with the aligned time axes. (2-7) The monitoring system according to any of (2-4) to (2-6), wherein the clustering execution unit shifts time axes of the status data of each of the plurality of devices and then performs the clustering based on the status data with the shifted time axes. (2-8) The monitoring system according to (2-7), wherein the clustering execution unit determines an amount of shift on the time axis based on a partial period of the chronological state changes indicated by the status data of each of the plurality of devices, and shifts the time axis of the status data of each of the plurality of devices based on the determined amount of shift. (2-9) The monitoring system according to any of (2-1) to (2-8), wherein the clustering execution unit calculates a difference between values indicated by the status data of each of the plurality of devices, and performs the clustering based on the calculated difference. (2-10) The monitoring system according to any of (2-1) to (2-9), wherein the status data acquisition unit acquires the status data of some of the devices among the plurality of devices for which some of the status data has not been acquired so that the devices indicate a predetermined value. (2-11) The monitoring system according to any one of (2-1) to (2-10), wherein the clustering execution unit executes the clustering based on the status data of each of the plurality of devices and attributes associated with the status data.(2-12) The monitoring system according to (2-11), wherein the attribute indicates a location where each of the plurality of devices is located, and the clustering execution unit executes the clustering based on the status data of each of the plurality of devices and the location associated with the status data.
Claims
1. A monitoring system comprising: a status data acquisition unit that acquires status data relating to the status of a device to be monitored in a service; and, based on the status data, executes processing relating to an abnormality in the service.
2. The monitoring system of claim 1, wherein the status data acquisition unit acquires the status data relating to changes in the status of the device over time, and the monitoring system further includes: a saturation determination unit that determines whether the status is saturated based on the status data; and a relationship determination unit that determines whether the status data is related to an abnormality in the service based on a determination result of the saturation determination unit.
3. The monitoring system of claim 2, wherein the status data acquisition unit acquires the status data when the abnormality is detected, the saturation determination unit determines whether the status is saturated or not based on the status data when the abnormality is detected, the relationship determination unit determines whether the status data is related to the abnormality when the abnormality is detected, and the monitoring system further includes a recovery processing execution unit that executes recovery processing for recovering from the abnormality based on the status data determined to be related to the abnormality when the abnormality is detected.
4. The monitoring system of claim 2 or 3, wherein the status data acquisition unit acquires the status data before the abnormality is detected, the saturation determination unit determines whether the status is saturated based on the status data before the abnormality is detected, and the relationship determination unit determines whether the status data is related to the abnormality before the abnormality is detected, and the monitoring system further includes a detection process execution unit that executes detection process for detecting the abnormality based on the status data determined to be related to the abnormality.
5. A monitoring system as described in claim 2 or 3, wherein the status data acquisition unit acquires the status data of the device in each of the multiple services, the saturation determination unit determines whether the device is saturated based on the status data of the device in each of the multiple services, and the relationship determination unit determines whether the status data is related to the abnormality in each of the multiple services based on the determination result of the saturation determination unit.
6. The monitoring system according to claim 2 or 3, further comprising a period setting unit which, when the abnormality is detected, sets a period relating to the status data based on the timing at which the abnormality occurred, and the status data acquisition unit acquires the status data relating to the time-series changes in the status of the device during the period.
7. The monitoring system according to claim 2 or 3, further comprising a normalization execution unit that performs normalization on the status data, and the saturation determination unit determines whether or not the status is saturated based on the status data on which the normalization has been performed.
8. A monitoring system as described in claim 2 or 3, wherein the relationship determination unit determines whether or not the status data is related to the abnormality based on a saturation period during which the status is determined to be saturated, within the period during which the change is indicated in the status data.
9. A monitoring system as described in claim 2 or 3, wherein the relationship determination unit determines whether or not the status data is related to the abnormality based on the maximum value of the status in a period prior to the abnormality period in which the abnormality occurs, among the periods in which the change is indicated in the status data.
10. The monitoring system of claim 7, wherein the other period is a period preceding the abnormal period, and the relationship determination unit determines whether the status data is related to the abnormality based on a maximum value of the status during the other period and a maximum value of the status during a period preceding the other period.
11. A monitoring system as described in claim 2 or 3, wherein the relationship determination unit determines whether or not the status data is related to the abnormality based on the maximum value of the status indicated by the status data during an abnormal period in which the abnormality occurs during the period in which the change is indicated in the status data.
12. A monitoring system as described in claim 2 or 3, wherein the relationship determination unit determines whether or not the status data is related to the abnormality based on the degree of coincidence between an abnormality period in which the abnormality occurs and a saturation period in which the status is determined to be saturated, within the period in which the change is indicated in the status data.
13. A monitoring system as described in claim 2 or 3, wherein the relationship determination unit determines whether or not the status data is related to the abnormality based on the amount of change in the status during each of a saturation period in which the status is saturated and another period prior to the abnormality period in which the abnormality occurs, during the period in which the change is indicated in the status data.
14. The monitoring system of claim 1, wherein the status data acquisition unit acquires the status data relating to the status of each of the plurality of devices, and the monitoring system further includes: a clustering execution unit that performs clustering on the status data of each of the plurality of devices; and an associated data identification unit that identifies associated data, which is other status data that is related to abnormality cause data, which is status data relating to the cause of an abnormality in the service, based on a result of the clustering execution.
15. The monitoring system of claim 14, wherein the status data acquisition unit acquires the status data of each of the multiple devices when the abnormality is detected, the clustering execution unit executes the clustering based on the status data of each of the multiple devices when the abnormality is detected, the related data identification unit identifies the related data when the abnormality is detected, and the monitoring system further includes a recovery processing execution unit that executes recovery processing for recovering from the abnormality based on the abnormality cause data and the related data when the abnormality is detected.
16. The monitoring system of claim 14 or 15, wherein the status data acquisition unit acquires the status data of each of the multiple devices before the abnormality is detected, the clustering execution unit executes the clustering based on the status data of each of the multiple devices before the abnormality is detected, the associated data identification unit identifies the associated data before the abnormality is detected, and the monitoring system further includes a detection process execution unit that executes detection process related to detecting the abnormality based on the abnormality cause data and the associated data.
17. A monitoring system as described in claim 14 or 15, wherein the status data of each of the plurality of devices indicates a time-series change in the status of the device, and the clustering execution unit executes the clustering based on the time-series change in the status indicated by the status data of each of the plurality of devices.
18. The monitoring system according to claim 17, further comprising: a saturation determination unit that determines whether or not the state indicated by the status data of each of the plurality of devices is saturated based on the status data of the each of the plurality of devices; and an abnormality factor data identification unit that identifies, among the status data of each of the plurality of devices, the status data that is determined to be saturated as the abnormality factor data.
19. The monitoring system according to claim 17, wherein the clustering execution unit aligns a time axis of the status data of each of the plurality of devices, and then executes the clustering based on the status data whose time axis has been aligned.
20. The monitoring system according to claim 17, wherein the clustering execution unit shifts a time axis of the status data for each of the plurality of devices, and then executes the clustering based on the status data with the time axis shifted.
21. The monitoring system of claim 20, wherein the clustering execution unit determines an amount of shift of the time axis based on a portion of a period of the chronological change in the status indicated by the status data of each of the plurality of devices, and shifts the time axis of the status data of each of the plurality of devices based on the determined amount of shift.
22. The monitoring system according to claim 14 or 15, wherein the clustering execution unit calculates a difference between values indicated by the status data of each of the plurality of devices, and executes the clustering based on the calculated difference.
23. A monitoring system as described in claim 14 or 15, wherein the status data acquisition unit acquires the status data such that all or some of the devices for which some of the status data has not been acquired among the plurality of devices show a predetermined value.
24. The monitoring system according to claim 14 or 15, wherein the clustering execution unit executes the clustering based on the status data of each of the plurality of devices and attributes associated with the status data.
25. The monitoring system of claim 24, wherein the attribute indicates a location where each of the plurality of devices is located, and the clustering execution unit performs the clustering based on the status data of each of the plurality of devices and the location associated with the status data.
26. A monitoring method comprising: a status data acquisition step of acquiring status data relating to the status of a device to be monitored in a service; and executing processing relating to an abnormality in the service based on the status data.
27. A program that causes a computer to function as a status data acquisition unit that acquires status data relating to the status of a device to be monitored in a service, and causes the computer to execute processing relating to an abnormality in the service based on the status data.
Citation Information
Patent Citations
Traffic monitoring system
JP2009017393A
Abnormal state detection device and abnormal state detection method
JP2013011987A
Failure detection device, failure detection method, failure detection program and recording medium
JP2015028700A
Network system, network control method, and control device
JP2016192661A
Clustering program, clustering method, and information processing apparatus
JP2017068748A
