KR102521361B1 - Apparatus and method for generating training data training network failure analysis model - Google Patents

Abstract

As a learning data generation device that generates learning data for learning a network failure analysis model, it stores a unit information conversion table in which a plurality of manufacturer unit information is mapped for each standard unit information, and a plurality of manufacturer alarm information is mapped for each standard alarm information. A conversion table storage unit for storing an alarm information conversion table, an alarm data collection unit for receiving alarm data generated when a failure occurs from at least one of an Element Management System (EMS) or a Network Management System (NMS), and in the alarm data Extracting manufacturer unit information and manufacturer alarm information, using the unit information conversion table to determine standard unit information corresponding to the extracted manufacturer unit information, and using the alarm information conversion table to extract manufacturer alarm information corresponding to the extracted manufacturer and a learning data generation unit that determines standard alarm information and generates learning data including the determined standard unit information and the determined standard alarm information.

Description

Apparatus and method for generating training data for training network failure analysis model

The present invention relates to an apparatus and method for generating learning data for training a network failure analysis model.

Recently, AI (Artificial Intelligence) technology is applied to various IT fields and is rapidly growing. In addition, in the network field, attempts to introduce AI technology to predict failures that will occur in the network in advance and analyze the causes of the failures are rapidly increasing. Specifically, an attempt to create a network failure analysis model using data received from a network or network equipment as learning data when a failure occurs is increasing.

Meanwhile, in order to manage a network and equipment constituting the network, an Element Management System (EMS) and a Network Management System (NMS) are typically used. EMS, as a network equipment management system, provides alarm data including information on the failure of equipment constituting the network, and NMS, as a network management system, provides alarms including information on network configuration, performance, failure, and security. Provide data.

However, the alarm data provided from the EMS and NMS includes a large number of unstructured data, and even if the data is partly structured, the information provided for the same failure is different for each manufacturer operating the EMS and NMS. Therefore, when the alert data provided from the EMS and NMS is used as training data to generate a network failure analysis model, there is a problem in that desired performance is not achieved.

An object of the present invention is to provide an apparatus and method for generating learning data for generating a network failure analysis model by processing alert data collected from EMS and NMS.

According to an embodiment of the present invention, a learning data generation apparatus for generating learning data for learning a network failure analysis model stores a unit information conversion table in which a plurality of manufacturer unit information is mapped for each standard unit information, and each standard alarm information Alarm data collection for receiving alarm data generated when a failure occurs from at least one of a conversion table storage unit storing an alarm information conversion table in which a plurality of manufacturer alarm information is mapped, an EMS (Element Management System) or a NMS (Network Management System) and manufacturer unit information and manufacturer alarm information are extracted from the alarm data, standard unit information corresponding to the extracted manufacturer unit information is determined using the unit information conversion table, and standard unit information is determined using the alarm information conversion table. and a learning data generation unit that determines standard alarm information corresponding to the extracted manufacturer alarm information and generates learning data including the determined standard unit information and the determined standard alarm information.

The learning data includes at least one of self information, slot information, port information, and standard signal hierarchy information, and the learning data generator generates the self information based on at least one of system name information and location information included in the alert data. , At least one of the slot information and the port information is determined, and the standard signal hierarchy information is determined using signal hierarchy information included in the alert data.

The learning data further includes reception number information, and the learning data generation unit determines the number of times the alert data has been received as the reception number information.

According to the present invention, since learning data is generated using alert data collected from EMS and NMS, AI technology can be introduced into the network field by using the existing system as it is.

In addition, according to the present invention, since unstructured data collected from EMS and NMS is processed into learning data that is easy to learn, the application rate of AI technology to the network field can be increased.

1 is a diagram illustrating an environment in which an apparatus for generating learning data according to an embodiment of the present invention is implemented.
2 is a structural diagram of an apparatus for generating learning data according to an embodiment of the present invention.
3 to 5 are diagrams illustrating a method of determining information for constructing a conversion table.
6 is a diagram illustrating a method of generating learning data using alert data received by the learning data generating apparatus according to an embodiment of the present invention.
7 is a diagram for explaining a method of updating a unit information conversion table and an alarm information conversion table by using the alert data received by the learning data generating device according to an embodiment of the present invention.

Hereinafter, with reference to the accompanying drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily carry out the present invention. However, the present invention may be embodied in many different forms and is not limited to the embodiments described herein. And in order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification.

Throughout the specification, when a certain component is said to "include", it means that it may further include other components without excluding other components unless otherwise stated. In addition, terms such as “… unit”, “… unit”, and “module” described in the specification mean a unit that processes at least one function or operation, which may be implemented as hardware or software or a combination of hardware and software. there is.

1 is a diagram illustrating an environment in which an apparatus for generating learning data according to an embodiment of the present invention is implemented.

Referring to FIG. 1 , an environment in which a learning data generating device is implemented includes an Element Management System (EMS) 100 , a Network Management System (NMS) 200 and a learning data generating device 300 .

The EMS 100 is a network equipment management system, and when a failure occurs in a network equipment (hereinafter, a unit) constituting a network, it receives failure information from the corresponding unit and provides alarm data.

The NMS 200, as a network management system, provides alarm data on network configuration, performance, failure, and security.

The items included in the alarm data provided by the EMS 100 and the NMS 200 are the same as defined in the standard (eg, ITU-T Rec. G.709/Y.1331 (06/2016)), Information on each item is different depending on the manufacturer that manufactured the unit even when the same failure occurs in a unit performing the same function.

For example, the first network and the second network are networks composed of ROADM (Re-configurable Optical Add-Drop Multiplexer) technology built by the first manufacturer and the second manufacturer, respectively, and the system communication unit of the first network and the second network 2 When a failure occurs in the system communication unit of the network in which the optical signal does not enter at the same time, the first alarm data provided by the first EMS 110 interlocked with the first network is the same as row 2 in Table 1 and interlocked with the second network The second alarm data provided by the second EMS 120 is shown in row 3 in Table 1.

System name information time information status information Manufacturer unit information signal hierarchy information Manufacturer alarm information location information No. 1
network 2018/3/31
-02:04:46 CR SSC Client R_LOS 1-1-1-1-1 2nd
network 2018/3/31
-02:04:46 CR WCU Client (C) STM-LOS 2-1-2-1-2

Although the first alarm data and the second alarm data include items defined in the standard in common, even if the same failure occurs in a unit performing the same function, since each network is built by different manufacturers, the first alarm data It has manufacturer unit information SSC and manufacturer alarm information R_LOS, and the second alarm data has manufacturer unit information WCU and manufacturer alarm information (C) STM-LOS to have different manufacturer unit information and manufacturer alarm information. Thus, the same function Even if the same failure occurs in the unit that performs the training data for learning the network failure analysis model without processing the alarm data provided from the EMS (100) and the NMS (200), as each manufacturer provides different alarm data. There are difficulties in learning when used as

The learning data generating device 300 receives alert data from at least one of the EMS 100 and the NMS 200, and generates learning data by processing the received alert data using a pre-stored conversion table. How the learning data generating device 300 generates learning data will be described with reference to FIG. 2 .

2 is a structural diagram of an apparatus for generating learning data according to an embodiment of the present invention, and FIGS. 3 to 5 are diagrams illustrating a method of determining information for constructing a conversion table.

Referring to FIG. 2 , the learning data generating device 300 includes an alert data collecting unit 310 , a conversion table storage unit 320 and a learning data generating unit 330 .

The alarm data collection unit 310 receives alarm data from at least one of the EMS 100 and the NMS 200 .

Specifically, when a failure occurs in the network or a unit constituting the network, the alarm data collection unit 310 collects alarm data from at least one of the EMS 100 interworking with the unit with failure or the NMS 200 interworking with the network where failure occurs. receive

The conversion table storage unit 320 stores a conversion table for converting information included in the received alert data. Specifically, the conversion table storage unit 320 stores a unit information conversion table and an alarm information conversion table.

3 to 5 are diagrams illustrating a method of determining information for constructing a conversion table.

Referring to FIG. 3 , it is possible to determine manufacturer unit information for each manufacturer for units performing the same function using information included in each manufacturer's manual. In this case, the location where each unit is implemented can be additionally determined.

Referring to FIG. 4 , manufacturer alarm information for each manufacturer for the same failure information may be determined using information included in each manufacturer's manual.

Referring to FIG. 5 , standard unit information for a unit performing a specific function and standard alarm information for specific failure information may be determined using information included in a standard document.

Standard unit information Manufacturer unit information CONT SSC, WCU MUX M40V, WMDUe-40

Standard alarm information Manufacturer alarm information REM REM_SF, (D) STM-LOS LOS R_LOS, (C) STM-LOS

Table 2 is an example of a unit information conversion table, and Table 3 is an example of an alarm information conversion table. By combining the information determined in FIGS. 3 to 5, a unit information conversion table and an alarm information conversion table can be constructed.

For example, for the system communication unit in FIG. 3, the manufacturer unit information is SCC for the first manufacturer and WCU for the second manufacturer, and since the standard unit information for the same unit in FIG. 5 is CONT, the conversion table storage unit ( 320) may map manufacturer unit information SSC and WCU to standard unit information CONT and store them in a unit information conversion table. In the same way, the conversion table storage unit 320 may map the standard unit information MUX to the manufacturer unit information M40V and WDUe-40 for the optical multiplexing unit and store them in the unit information conversion table. That is, the conversion table storage unit 320 stores a table mapping a plurality of manufacturer unit information for each standard unit information.

Also, for example, in FIG. 4, the manufacturer alarm information is REM_SF in the case of the first manufacturer and (D) STM-LOS in the case of the second manufacturer for the failure in which the optical signal of the other party does not come in. Since the standard alarm information is REM, the conversion table storage unit 320 may map manufacturer alarm information REM_SF and (D) STM-LOS to the standard alarm information REM and store them in the alarm information conversion table. In the same way, the conversion table storage unit 320 may map standard alarm information LOS with manufacturer alarm information R_LOS and (C) STM-LOS for failures in which the optical signal does not enter and store them in the alarm information conversion table. That is, the conversion table storage unit 320 stores a table mapping a plurality of manufacturer alarm information for each standard alarm information.

The learning data generation unit 330 generates learning data by processing information included in the collected alarm data using the unit information conversion table and the alarm information conversion table stored in the conversion table storage unit 320 . In this case, the learning data includes self information, slot information, port information, standard unit information, signal hierarchy information, standard alarm information, and reception count information.

Specifically, the learning data generator 330 extracts manufacturer unit information and manufacturer alarm information from the collected alarm data, determines standard unit information corresponding to the extracted manufacturer unit information using a unit information conversion table, and determines alarm information Standard alarm information corresponding to the manufacturer alarm information extracted using the conversion table is determined.

Also, the learning data generator 330 determines at least one of self information, slot information, and port information based on at least one of system name information and location information included in the collected alert data. Self information, slot information, and port information are information representing the physical location of a unit that has transmitted alarm data in each network.

In addition, the learning data generation unit 330 determines the signal hierarchy information included in the collected alert data as the signal hierarchy information of the learning data. In this case, the signal hierarchy information is divided into User, Client, OPU, ODU, OTU, OCH, OTS, and OSC. It means the signal hierarchy related to the connection. OPU, ODU, and OTU belong to the digital signal layer of OTN (Optical Transport Network) technology, and OCH and OTS belong to the optical signal layer of OTN technology. OSC is defined separately from OTN technology as a signaling hierarchy related to optical monitoring channels.

In addition, the learning data generation unit 330 counts the number of times that the alarm data determined by the same self information, slot information, port information, standard unit information, signal hierarchy information, and standard alarm information is received, and converts the received number into reception number information to decide

The learning data generation unit 330 generates learning data by mapping self information, slot information, port information, standard unit information, standard signal hierarchy information, standard alarm information, and reception number information determined through the above process with system name information. . For example, the learning data generation unit 330 may generate learning data as shown in Table 4 with respect to the alert data in Table 1. Although the first alarm data and the second alarm data in Table 1 are provided from a network built by a separate manufacturer, the corresponding unit is the same as the system communication unit, and the generated failure is the same as a failure in which the optical signal does not enter. Standard unit information and standard alarm information of the first network and the second network are the same.

System name information 1st network 2nd network self information 2 3 slot information 2 4 port information One One Standard unit information CONT CONT Standard Signal Hierarchy Information Client Client Standard alarm information LOS LOS Reception count information One One

6 is a diagram illustrating a method of generating learning data using alert data received by a learning data generating apparatus according to an embodiment of the present invention. In FIG. 6, the descriptions of FIGS. 1 to 5 omit

Referring to FIG. 6 , the learning data generation device 300 stores a unit information conversion table mapping a plurality of manufacturer unit information for each standard unit information and an alarm information conversion table mapping a plurality of manufacturer alarm information for each standard alarm information. (S100).

The learning data generating device 300 receives alert data from at least one of the EMS 100 and the NMS 200 (S110).

The learning data generating device 300 extracts manufacturer unit information and manufacturer alarm information from the received alarm data, and determines standard unit information and standard alarm information respectively corresponding to the extracted manufacturer unit information and manufacturer alarm information (S120).

The learning data generating device 300 determines self information, slot information, and port information based on at least one of system name information and location information included in the collected alert data (S130).

The learning data generating device 300 determines reception count information based on the number of times the same alert data is received (S140).

The learning data generating device 300 learns the received alarm data using the determined self information, slot information, port information, standard unit information, signal hierarchy information included in the collected alarm data, standard alarm information, and reception count information. Data is generated (S150).

7 is a diagram for explaining a method of updating a unit information conversion table and an alarm information conversion table by using the alert data received by the learning data generating device according to an embodiment of the present invention.

Referring to FIG. 7 , the learning data generating device 300 receives alert data from at least one of the EMS 100 and the NMS 200 (S200).

The learning data generating device 300 extracts manufacturer unit information from the received alert data (S210).

The learning data generating device 300 checks whether the extracted manufacturer unit information exists in the previously stored unit information conversion table (S220).

If the extracted manufacturer unit information does not exist in the unit information conversion table, the learning data generating device 300 generates new standard unit information (S221) and converts the manufacturer unit information extracted by mapping to the new standard unit information into unit information. It is stored in the table (S223). In this case, new standard unit information may be set by randomly extracting a character string included in manufacturer unit information set or extracted by the user. Thereafter, the data generating device extracts manufacturer alarm information from the received alarm data (S230).

If the extracted manufacturer unit information exists in the unit information conversion table, the learning data generating device 300 extracts manufacturer alarm information from the received alarm data (S230).

The learning data generating device 300 checks whether the extracted manufacturer alarm information exists in a pre-stored alarm information conversion table (S240).

If the extracted manufacturer alarm information does not exist in the alarm information conversion table, the learning data generating device 300 generates new standard alarm information (S241), maps the extracted manufacturer alarm information to the new standard alarm information, and converts the extracted manufacturer alarm information into alarm information. It is stored in the table (S243). In this case, the new standard alarm information may be set by randomly extracting a character string included in manufacturer alarm information set or extracted by the user. Then, the data generating device 300 ends updating the unit information conversion table and the alarm information conversion table (S250).

If the extracted manufacturer alarm information exists in the alarm information conversion table, the data generating device 300 ends updating the unit information conversion table and the alarm information conversion table (S250).

Although the unit information conversion table is first updated and the alarm information conversion table is then updated in FIG. 7 , it goes without saying that the alarm information conversion table may be updated first and then the unit information conversion table may be updated according to an embodiment.

According to the present invention, since learning data is generated using alert data collected from EMS and NMS, AI technology can be introduced into the network field by using the existing system as it is.

In addition, according to the present invention, since unstructured data collected from EMS and NMS is processed into learning data that is easy to learn, the application rate of AI technology to the network field can be increased.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements made by those skilled in the art using the basic concept of the present invention defined in the following claims are also included in the scope of the present invention. that fall within the scope of the right.

Claims (3)

As a learning data generating device,
A conversion table storage unit for storing a unit information conversion table mapping a plurality of manufacturer unit information for each standard unit information and an alarm information conversion table mapping a plurality of manufacturer alarm information for each standard alarm information;
An alarm data collection unit that receives alarm data generated when a failure occurs from at least one of an EMS (Element Management System) and an NMS (Network Management System), and
Manufacturer unit information and manufacturer alarm information are extracted from the alarm data, the extracted manufacturer unit information is converted into standard unit information using the unit information conversion table, and the manufacturer alarm information is extracted using the alarm information conversion table. After converting the standard alarm information into standard alarm information, a learning data generator for generating standardized learning data for learning a network failure analysis model from the standard unit information and the standard alarm information;
The learning data generating unit
If standard information corresponding to the manufacturer unit information extracted from the alarm data or the manufacturer alarm information is not present in the unit information conversion table or the alarm information conversion table, new information is generated based on the manufacturer unit information or characters extracted from the manufacturer alarm information. A learning data generating device that generates standard unit information or new standard alarm information, and stores the new standard unit information or new standard alarm information in the unit information conversion table or the alarm information conversion table. In paragraph 1,
The learning data is
including at least one of self information, slot information, port information, or standard signaling hierarchy information;
The learning data generating unit
At least one of the self information, the slot information, and the port information is determined based on at least one of system name information and location information included in the alert data, and the standard information is determined using signal hierarchy information included in the alert data. A learning data generating device that determines signal hierarchy information. In paragraph 1,
The learning data is
Further includes information on the number of receptions,
The learning data generating unit
Learning data generating device for determining the number of times the alert data has been received as the number of times of receiving information.

Images