WO2021160270A1

WO2021160270A1 - Technique for determining an impact of a failure in a mobile communication network

Info

Publication number: WO2021160270A1
Application number: PCT/EP2020/053789
Authority: WO
Inventors: Attila BÁDER; László KOVÁCS; Gábor MAGYAR; Norbert PURGER
Original assignee: Telefonaktiebolaget LM Ericsson AB
Current assignee: Telefonaktiebolaget LM Ericsson AB
Priority date: 2020-02-13
Filing date: 2020-02-13
Publication date: 2021-08-19
Anticipated expiration: 2022-08-13

Abstract

An aspect of the present disclosure is directed to a device for determining an impact of a failure in a mobile communication network. The device comprises at least one processor and memory comprising instructions executable by the at least one processor. The device is operable to receive fault management, FM, data informing of the failure in the mobile communication network. The fault management data comprises information on a time of the failure and information on a failed cell of the mobile communication network. The device is further operable to determine a set of possibly impacted cells, based on the information on the failed cell, obtain subscriber activity data for subscribers in the set of possibly impacted cells, calculate, based on the subscriber activity data, a network level impact value indicating an impact of the failure to the mobile communication network, and store the network level impact value in a fault register, in association with the failure. Further related aspects are directed to a method, a computer program, and a computer-readable medium.

Description

Technique for determining an impact of a failure in a mobile communication network

Technical Field

The present disclosure generally relates to a technique for determining an impact of a failure in a mobile communication network. In particular, the present disclosure is in the field of Fault Management (FM).

Background

A typical approach for analyzing failure impacts in communication networks is to provide a designated function for a designated purpose, focusing on error cases. Examples related to mobile communications for analyzing handover failures are described in US 10,051,532. Aiming to be more generic, CN 207743973 U describes an architecture including an intelligent terminal and a component configured to analyze failure information.

In terms of impact analysis, previous attempts as in US 7,092,707 aim to describe generic methods, processes to relate services with more services and sub^¬ components to have a dependency model, and to describe how alerts could be handled in such cases. However, such generic handling methods might fail to identify alarms and fail in properly quantifying cases when - due to the domain specific interworking - both positive and negative effects might come into play in a scenario.

More specific to mobile communication network service impacts, US 5,872,911 focuses on assessing network failures in terms of an absolute impact on a type of aggregated service indicator, like total call traffic over the network. However, if the system, like the mobile communication network, is designed to provide various levels of redundancies, such indicators alone might not even change - while the faults still cause degradation in other related metrics. Generally speaking, it is known that FM systems receive node and network alarms from the different network elements, e.g., network nodes of a mobile communication network. Alarms are related to node failures and not directly related to service usage and, therefore, the operator has no information about the impact on services and the number and type of impacted subscribers.

According to current techniques, obtaining additional information on the impact of a failure requires at least one of additional tools, expert knowledge, and manual investigation, which is time consuming. Therefore, this analysis is usually not performed, or takes several hours or even days to be completed.

Lacking this analysis, FM alarms are prioritized without properly considering the service or subscriber impact. Therefore, technical resources working on the solution of tickets generated by FM alarms are not allocated in an optimum way. It can happen that alarms which have no or little impact on services or subscribers are solved with high priority, and alarms having large impacts are down-prioritized.

The number of raised alarms in an FM system is high, usually much higher than the number the operator can analyze or solve. Alarms are prioritized based on rules and best technical knowledge. The operator does not have exact information which alarms can be neglected. There is a high risk that an operator neglects or down- prioritizes important alarms.

Especially, in case of radio node or cell failures, it is difficult to estimate the impact, since the effect of the outage or failure depends highly on the actual location, radio coverage, load, etc. of the area. It can occur that a node failure has no impact on services or subscribers because overlapping cells take over the actual sessions/calls without any problem. However, it can occur that most of the subscribers served by the failed cell(s) are impacted. It can also occur that the traffic taken over impacts the subscribers or services in the neighbor cells or area. The redundancy strongly depends on the actual traffic situation and on the used services as well. Current solutions do not consider the redundancy of the radio cell plan, the traffic load and used services. Summary

Accordingly, there is a need for a technique which solves at least one of the above problems or other related problems of prior art techniques. Specifically, and without limitation, there is a need for an improved technique for determining an impact of a failure in a mobile communication network.

According to a first aspect, a device for determining an impact of a failure in a mobile communication network is provided. The device comprises at least one processor and memory comprising instructions executable by the at least one processor. The device is operable to receive fault management, FM, data informing of the failure in the mobile communication network. The fault management data comprises information on a time of the failure and information on a failed cell of the mobile communication network. The device is further operable to determine a set of possibly impacted cells, based on the information on the failed cell, obtain subscriber activity data for subscribers in the set of possibly impacted cells, calculate, based on the subscriber activity data, a network level impact value indicating an impact of the failure to the mobile communication network, and store the network level impact value in a fault register, in association with the failure.

The following description may apply to all aspects described in this disclosure. In particular, the following description concerning the device aspects may not only apply to the device aspects but also to the method aspects described below, where applicable.

The mobile communication network may operate according to a 3rd Generation Partnership Project (3GPP) standard, such as a 4G standard (e.g., LTE) or a 5G standard. However, the mobile communication network is not limited to 3GPP mobile communication standards. The present technique may also be employed in other mobile communication networks such as cdma2000 systems standardized by 3GPP2. Further, the mobile communication network is not limited to a particular radio access network (RAN) and may comprise a corresponding core network. Further, the mobile communication network may comprise a plurality of different radio access networks (RANs), such as LTE RAN and 5G RAN.

The failure in the mobile communication network may refer to any failure in any node of the above-described mobile communication network, which might have an impact on other nodes and/or devices of the mobile communication network. In particular, the failure might have an impact on mobile devices (e.g., user equipment, UE) connected to a base station (e.g., eNodeB or 5GNB) of a corresponding radio access network. For example, a failure in a particular base station (e.g., eNodeB) may result in a loss of connection of a plurality of mobile devices that were in a cell served by the failed base station.

The device may be integrated in a computer system of the mobile communication network. For example, the device may be part of an operations support system (OSS). In any case, the device is capable of receiving FM data that is indicative of a fault having occurred in the mobile communication network. For this purpose, the device may either be part of the mobile communication network or be communicatively coupled to the mobile communication network.

The device may be implemented in a cloud computing environment. For example, the at least one processor may comprise a plurality of processors at different sites.

The FM data may trigger an FM alarm. In addition to the information on the time of failure and the information on the failed cell of the mobile communication network, the FM data may comprise further details of the failure, such as a type of the failure (planned or unplanned). The information on the time of failure may be represented by a timestamp of the failure (indicating, e.g., date and time of the failure). The information on the failed cell may be represented by an ID of a failed network node or a failed cell, such as a cell ID.

The failed cell may be part of the set of possibly impacted cells. The possibly impacted cells may be regarded as neighboring cells of the failed cell. The possibly impacted cells may use the same Radio Access Technology (RAT) or a different RAT with regard to that of the failed cell.

The subscriber activity data may describe details and/or statistics of a subscriber activity for the subscribers in the set of possibly impacted cells. The subscriber activity may refer to activity before and/or after the time of the failure (indicated, e.g., by a timestamp). For example, for each of the subscribers of the mobile communication network, an activity may be recorded and saved as subscriber activity data together with a timestamp of the activity. In one example, the subscriber activity data may indicate, which subscribers were active in a particular cell at a particular time. In this case, the subscriber activity data may comprise entries for a plurality of times (e.g., at fixed intervals). The network level impact value is calculated based on the subscriber activity data.

For example, a larger impact value may be calculated in a case where more subscribers are affected by the fault, whereas a smaller impact value may be calculated in a case where less subscribers are affected by the fault. For example, a larger impact value may be calculated in a case where subscribers using services that are more important are affected by the fault, whereas a smaller impact value may be calculated in a case where subscribers using services that are less important are affected by the fault.

The network level impact value may be stored in the fault register, in association with the failure, such that an impact of the failure may be put into context with other failures that have occurred or will occur. In other words, based on the network level impact value, a significance or importance of the fault can be derived, such that the fault can be correctly prioritized, if necessary. In other words, based on the information stored in the fault register, a list of faults can be output, ordered by their network level impact value (i.e., ordered by their priority).

Hence, it is possible to decide which of the faults have to be treated with higher priority and which of the faults can be treated with lower priority or may be even neglected.

The device may be further configured to determine the set of possibly impacted cells based on the information on the failed cell and based on handover relation data indicative of previous handover processes between the failed cell and neighboring cells.

Handover relation data may include information on previous handover processes between cells of the mobile communication network. When handover processes involving the failed cell are considered, it is possible to identify neighboring cells. In other words, neighboring cells may be identified as cells from and/or to which previous handover processes have been performed involving the failed cell. These neighboring cells may be identified as the possibly impacted cells.

The device may be further configured to determine the set of possibly impacted cells based on the information on the failed cell and based on geo-location data indicative of geographic locations of a plurality of cells including the failed cell. In addition or alternative to handover relation data, the device may consider geolocation data. The geo-location data may be part of Configuration Management (CM) data. For example, the geo-location data may indicate geographical locations (e.g., latitude and longitude coordinates) for each cell of the mobile communication network. Based on this information, with regard to the failed cell, neighboring cells or cells within a predefined distance from the failed cell can be identified as belonging to the group of possibly impacted cells. Further, the handover relation data may be used to make a pre-selection of neighboring cells and, in a second step, the geolocation data may be used for selecting the possibly impacted cells out of this pre^¬ selection, wherein only cells within a predefined distance from the failed cell are considered.

The subscriber activity data may comprise subscriber activity data for subscribers in the failed cell.

The subscriber activity data may comprise at least one data set for each subscriber present in the failed cell and in the possibly impacted cells, the data set defining a type and/or identification number of the respective subscriber, wherein the subscriber activity data is indicative of a number of the subscribers present in the failed cell and in the possibly impacted cells.

The type of the respective subscriber may indicate a model type of a mobile device used by the subscriber. The identification number of the respective subscriber may be an international mobile subscriber identity (IMSI) or an international mobile equipment identity (IMEI). Based on the plurality of data sets that are part of the subscriber activity data, the number of subscribers present in the failed cell and the possibly impacted cells can be derived. Further, this number may also be explicitly indicated in the subscriber activity data (e.g., number of active subscribers for each cell, for each time).

At least one of the data sets may comprise a communication service used by the respective subscriber and a key performance indicator, KPI, value for the communication service.

The communication service may refer to a particular type of communication service used by the corresponding subscriber at a particular time. Examples of communication services are, e.g., video streaming, audio streaming, data download, telephony, video telephony, etc. Further, the corresponding KPI may be a KPI suitable for describing a performance of the corresponding communication service. For example, "stall ratio" may be used for "video" communication service and "throughput" may be used for "data download" communication service. The communication service may be at least one of the list including non-encrypted TCP, encrypted TCP, VoLTE, non-encrypted video, and non-encrypted web. Suitable KPIs for the communication services encrypted TCP and non-encrypted TCP may be at least one of create session, create PDP, LTE Attach, Packet loss (all directions and network segments), RTT / delay, TCP throughput, and rtt_term_sum_avg. Suitable KPIs for the communication service VoLTE may be at least one of IMS Session Setup Time & Session Setup Attempts, Session Establishments, Audio MOS and voice quality issues (muting, garbling, etc.), IMS Call Drops, Call Duration, Mean opinion score, soft drop, call setup success, packet loss. Suitable KPIs for the communication service non-encrypted video may be at least one of video stall ratio, init time, and mean opinion score (MOS). Suitable KPIs for the communication service non- encrypted web may be at least one of webpage access time, download time, and success ratio.

Each of the data sets may comprise a time stamp and the subscriber activity data may comprise data sets with time stamps indicating a time before the time of the failure and data sets with time stamps indicating a time after the time of the failure.

Subscriber activity data may be recorded and stored in a subscriber activity data storage, from which it is retrieved for each received FM data. A time of recording the respective activity data may be stored in association with the corresponding data set in the form of a timestamp. For example, in case the subscriber activity data is continuously stored (i.e., at fixed or flexible time intervals), data sets exist having a timestamp indicating a time before the time of the failure and data sets exist having a timestamp indicating a time after the time of the failure. These data sets can be analyzed in order to calculate the network level impact value.

The device may be further configured to calculate the network level impact value based on a final impact value for subscribers in the set of possibly impacted cells that are active after the time of the failure.

The final impact value may also be regarded a final impact value for all active users (as compared to users that have gone lost because of the failure. If not indicated otherwise, the expressions "users" and "subscribers" are used as synonyms herein. The final impact value may indicate an estimated KPI deterioration for the users that are active also after the fault (indicated by the FM data).

Calculating the final impact value may comprise determining, based on the subscriber activity data, a set of active subscribers that were active in the possibly impacted cells before and after the time of the failure. Calculating the final impact value may further comprise, for each subscriber of the set of active subscribers, for each communication service, and for each KPI type defined for the given communication service, computing a difference of the KPI values observed for the subscriber before the time of the failure and after the time of the failure, and weighting the differences according to a predefined KPI weight per communication service, to obtain a service impact value for the subscriber per each defined communication service. Calculating the final impact value may further comprise weighting the individual service impact values for each subscriber with a predefined service weight value and summing the weighted service impact values to obtain a total subscriber impact value and summing the total subscriber impact values for each of the active subscribers to obtain the final impact value.

The KPI weights for the different KPIs may be stored in a table within a memory of the device. Similarly, the service weight values for the different communication services may be stored in a table within a memory of the device.

The device may be further configured to calculate the network level impact value based on a number of subscribers that are lost in the mobile communication network due to the failure.

The subscribers that are lost due to the failure may correspond, e.g., to subscribers that lose a wireless connection to the failed cell due to the failure. In case the number of subscribers that are lost is higher, the network level impact value may be higher and in case the number of subscribers that are lost is smaller, also the network level impact value may be smaller.

The device may further be configured to calculate the number of subscribers that are lost based on learnt cell traffic data, the learnt cell traffic data comprising an average estimate of a number of subscribers in a specific cell for a specific time.

The learnt cell traffic data may be derived from the subscriber activity data. The learnt cell traffic data may represent a typical condition in the mobile communication network (i.e., without a fault). Thus, the learnt cell traffic data may indicate, for a particular time of day, an average estimate of a number of subscribers in a specific cell. Further, the data may be specific for a particular weekday (e.g., typical condition on Tuesdays, 9:00 am). Averaging may also be carried out on the basis of other aspects (other than the weekday), e.g., a particular month, a holiday/no holiday, etc.). Further, it may be averaged over a longer period (e.g., half a day, a full day, etc.).

The device may further be configured to calculate the number of subscribers that are lost based on a number of subscribers active before the time of the fault, derived from the subscriber activity data, a number of subscribers active after the time of the fault, derived from the subscriber activity data, an average estimated number of subscribers active before the time of the fault, derived from the learnt cell traffic data, and an average estimated number of subscribers active after the time of the fault, derived from the learnt cell traffic data.

The number of subscribers that are lost may correspond to a difference between the number of subscribers active before the time of the fault and the number of subscribers active after the time of the fault, wherein a typical percentage loss/increase is considered based on the average estimated number of subscribers active before/after the time of the fault. For example, when the average estimated number of subscribers active before/after the time of the fault indicates a typical loss at that time of 10%, then this loss is considered during the determination of the number of subscribers that are lost due to the fault. In other words, it is possible to consider only the users that are lost actually due to the fault and to ignore those who would have entered/exited the cell in any case (in the considered period). In order to keep an error as small as possible, the number of subscribers active before the time of the fault may be derived from data recorded as close to a timestamp of the fault as possible before the fault and the number of subscribers active after the time of the fault may be derived from data recorded as close to a timestamp of the fault as possible after the fault.

The device may further be configured to calculate the network level impact value based on an average KPI impact value, wherein the average KPI impact value is calculated based on historical average KPI levels for each defined communication service, and based on the KPI values of the subscriber activity data after the time of the impact, using predefined KPI weights per communication service and predefined service weight values. The KPI weights and the predefined service weight values may correspond to those mentioned above. The average KPI impact value enables to determine an impact on KPIs for users that have no corresponding entry in the subscriber activity data in the possibly impacted cells before the time of the fault.

The device may further be configured to calculate the network level impact value based on the final impact value, based on the number of subscribers that are lost, and based on the average KPI impact value.

Each of the parameters final impact value, number of subscribers that are lost, and average KPI impact value may be weighted with a predefined weight value. Further, the parameters may be normalized such that a single number may be output as network level impact value (e.g., a number between 0 and 10), depending on the impact.

For example, the network level impact value NLIV may be calculated as:

NLIV = a*FIV + b*IMOSTAL + c*AKIV, wherein FIV is the final impact value, NOSTAL is the number of subscribers that are lost, and AKIV is the average KPI impact value. The parameters a, b, and c are predefined weighting and normalization factors.

Further, the individual values for FIV, NOSTAL, and AKIV may also be stored in the fault register, in association with the failure. Thereby, a more detailed analysis regarding the impact of the failure can be performed. In other words, it may be possible to prioritize the individual faults on the basis of different criteria.

The device may further be configured to update the network level impact value stored in the fault register, in association with the failure.

The network level impact value may be updated in case at least one of the parameters, on the basis of which the previous network level impact value has been calculated, has changed. For example, the network level impact value may be updated in case the subscriber activity data has changed. More precisely, an update may be carried out when a timestamp T_after in the subscriber activity data of a relevant entry is newer than a timestamp TJastupdate of the network level impact value stored in the fault register. In order to update the network level impact value, a new network level impact value may be calculated by considering the newer subscriber activity data.

According to a second aspect, a method for determining an impact of a failure in a mobile communication network is provided. The method comprises receiving fault management, FM, data informing of the failure in the mobile communication network. The fault management data comprises information on a time of the failure and information on a failed cell of the mobile communication network. The method further comprises determining a set of possibly impacted cells, based on the information on the failed cell, obtaining subscriber activity data for subscribers in the set of possibly impacted cells, calculating, based on the subscriber activity data, a network level impact value indicating an impact of the failure to the mobile communication network, and storing the network level impact value in a fault register, in association with the failure.

Each of the above details concerning the device of the first aspect may also apply to the method of the second aspect. In other words, the device of the first aspect may be configured to perform the method of the second aspect.

Determining the set of possibly impacted cells based on the information on the failed cell may further be carried out based on handover relation data indicative of previous handover processes between the failed cell and neighboring cells.

Determining the set of possibly impacted cells based on the information on the failed cell may further carried out based on geo-location data indicative of geographic locations of a plurality of cells including the failed cell.

The subscriber activity data may comprise at least one data set for each subscriber present in the failed cell and in the possibly impacted cells, the data set defining a type and/or identification number of the respective subscriber, wherein the subscriber activity data is indicative of a number of the subscribers present in the failed cell and in the possibly impacted cells. At least one of the data sets may comprise a communication service used by the respective subscriber and a key performance indicator, KPI, value for the communication service.

Calculating the network level impact value may be based on a final impact value for subscribers in the set of possibly impacted cells that are active after the time of the failure.

The method may further comprise calculating the final impact value, wherein calculating the final impact value comprises determining, based on the subscriber activity data, a set of active subscribers that were active in the possibly impacted cells before and after the time of the failure, for each subscriber of the set of active subscribers, for each communication service, and for each KPI type defined for the given communication service, computing a difference of the KPI values observed for the subscriber before the time of the failure and after the time of the failure, and weighting the differences according to a predefined KPI weight per communication service, to obtain a service impact value for the subscriber per each defined communication service, weighting the individual service impact values for each subscriber with a predefined service weight value and summing the weighted service impact values to obtain a total subscriber impact value, and summing the total subscriber impact values for each of the active subscribers to obtain the final impact value.

Calculating the network level impact value may be based on a number of subscribers that are lost in the mobile communication network due to the failure.

The method may further comprise calculating the number of subscribers that are lost, wherein calculating the number of subscribers that are lost is based on learnt cell traffic data, the learnt cell traffic data comprising an average estimate of a number of subscribers in a specific cell for a specific time.

The method may further comprise calculating the number of subscribers that are lost, wherein calculating the number of subscribers that are lost is based on a number of subscribers active before the time of the fault, derived from the subscriber activity data, a number of subscribers active after the time of the fault, derived from the subscriber activity data, an average estimated number of subscribers active before the time of the fault, derived from the learnt cell traffic data, and an average estimated number of subscribers active after the time of the fault, derived from the learnt cell traffic data.

Calculating the network level impact value may be based on an average KPI impact value, wherein the average KPI impact value is calculated based on historical average KPI levels for each defined communication service, and based on the KPI values of the subscriber activity data after the time of the impact, using predefined KPI weights per communication service and predefined service weight values.

Calculating the network level impact value may be based on the final impact value, based on the number of subscribers that are lost, and based on the average KPI impact value.

The method may further comprise updating the network level impact value stored in the fault register, in association with the failure.

According to a third aspect, a computer program is provided. The computer program comprises instructions which, when the program is executed by a computer, cause the computer to carry out the method of the second aspect.

According to a fourth aspect, a computer-readable medium is provided. The computer-readable medium comprises instructions which, when executed by a computer, cause the computer to carry out the method of the second aspect.

The computer-readable medium may be a tangible or non-tangible computer- readable medium. Examples of a computer-readable medium are, e.g., an optical recording medium, a magnetic recording medium, a solid-state recording medium, etc.

Brief Description of the Drawings

Further details of embodiments of the technique are described with reference to the enclosed drawings, wherein:

Fig. 1 shows the impact a failure in a failed cell has on different mobile devices; Fig. 2 shows an arrangement of a device according to the present disclosure in the architecture of a mobile communication network;

Fig. 3 shows a schematic representation of a device for determining an impact of a failure in a mobile communication network according to the present disclosure;

Fig. 4 shows a schematic representation of a method for determining an impact of a failure in a mobile communication network according to the present disclosure;

Fig. 5 shows a further schematic representation of a method for determining an impact of a failure in a mobile communication network according to the present disclosure; and

Fig. 6 shows a further schematic representation of a technique for determining an impact of a failure in a mobile communication network according to the present disclosure.

Detailed Description

In the following description, for purposes of explanation and not limitation, specific details are set forth, such as a specific network environment in order to provide a thorough understanding of the technique disclosed herein. It will be apparent to one skilled in the art that the technique may be practiced in other embodiments that depart from these specific details.

Moreover, when in this disclosure the expression base station is used, it generally refers to a base station of a mobile communication network, said base station being configured to provide a cell. As an example in 4G networks, the base station may be an evolved NodeB (eNodeB or eNB) and in 5G networks, the base station may be one of the base stations defined in the 5G standard (e.g., gNB, ng-eNB, herein summarized as 5GNB). When the expression mobile device is used, it refers to a mobile device of a subscriber (or user) connected to a particular cell (e.g., a smartphone, a portable computer, a tablet, etc.). The mobile device may also be referred to as UE. Moreover, those skilled in the art will appreciate that the services, functions, steps and units explained herein may be implemented using software functioning in conjunction with a programmed microprocessor, an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), a Digital Signal Processor (DSP) or a general purpose computer, e.g., including an Advanced RISC Machine (ARM). It will also be appreciated that, while the following embodiments are primarily described in context with methods and devices, the embodiments may also be embodied in a computer program product as well as in a system comprising a computer processor and memory coupled to the processor, wherein the memory is encoded with one or more programs that may perform the services, functions, steps and implement the units disclosed herein.

Further, in the following, specific devices (in particular a device for determining an impact of a failure in a mobile communication network) are described, which perform or are configured to perform certain steps of a method. However, it will be appreciated by those skilled in the art that those steps do not necessarily have to be performed by one single device but may be performed by different devices that are communicatively coupled with each other. For example, more than one device may be provided and/or more than one processor may be provided, wherein the steps are distributed among the devices and/or processors. Further, a cloud computing environment may be used for performing the steps of one or more of the methods described herein.

Fig. 1 shows a cell 10, in which a cell outage has occurred. A technical fault may have occurred in the base station hosting the cell 10, which has led to the outage. In the following, the cell 10 will be referred to as failed cell or faulty cell. The failed cell 10 is surrounded by (exemplarily 6) neighbor cells 12. Further, exemplary UEs are shown in Fig. 1 (as squares and triangles), which are located either within the failed cell 10 or within one of the neighbor cells 12. As shown in Fig. 1, the cell outage may have a different impact on each of the UEs. The impact is not necessarily predetermined only by the fact whether the respective UE is located within the failed cell 10 or not. For example, for some of the UEs within the failed cell 10, it is possible that a neighbor cell 12 takes over the connection to the respective UE. Hence, as shown in Fig. 1, the following possible impacts to the UEs in the failed cell 10 and in a neighbor cell 12 are possible: Lost in faulty cell, impacted in faulty cell, impacted in neighbor cell, no impact in faulty cell, and no impact in neighbor cell. Since the above-described various types of impacts on UEs exist, the impact of a cell failure is a) difficult to predict and b) requires a consideration of the full impact (e.g., on each failed cell and in other possibly impacted cells). It is therefore an object of the present disclosure to provide a technique that considers situations like that shown in Fig. 1 and which outputs a reliable network level impact value with regard to a particular failure.

Fig. 2 shows a device for determining an impact of a failure in a mobile communication network, wherein said device is referred to as customer experience manager (CEM) 20. The CEM 20 also has additional functions apart from those described below, wherein those additional functions are not relevant for the present invention and will therefore not be described herein. As shown in Fig. 2, the CEM 20 is part of an operations support system (OSS) 22. The CEM 20 comprises a network and subscriber analyzer (impact analysis) 24 and a correlator 26. The correlator 26 receives data from various devices of the mobile communication network and correlates this data in order to be further processed by the network and subscriber analyzer 24. Further, a fault management module FM, a performance management module PM, and a configuration management module CM are connected to the network and subscriber analyzer 24.

The OSS 22 monitors a core network, an LTE radio network, and a 5G radio network. A plurality of UEs is wirelessly connected to each of the radio networks (LTE and 5G). More precisely, a plurality of UEs is connected to a corresponding base station (eNodeB and 5GNB) of the respective radio network. The core network comprises a session management function (SMF), an access and mobility management function (AMF), and a user plane function (UPF). As indicated by the dotted arrows in Fig. 2, the aforementioned elements of the radio networks and the core network each are configured to transmit data to the correlator 26 of the CEM 20.

Details regarding the further processing of the data will be described below in the following detailed description of the embodiments.

Fig. 3 shows a schematic structure of a device 30 for determining an impact of a failure in a mobile communication network, according to the present disclosure. The device 30 comprises a network interface 32 that is adapted to communicatively couple the device 30 to the mobile communication network. The device 30 further comprises a processor 34 and a memory 36 containing instructions executable by the processor 34 to cause the device 30 to perform the method shown in Fig. 4. In particular, the interface 32 may be configured to receive fault management (FM) data informing of a failure in the mobile communication network. Further, it may be possible to store a network level impact value in an external fault register via the interface 32. However, the fault register may also be part of the memory 36. The device 30 may be the device 20 shown in Fig. 2.

Fig. 4 shows a flowchart of a method for determining an impact of a failure in a mobile communication network. The method starts with a step of receiving 40 fault management (FM) data informing of the failure in the mobile communication network, the fault management data comprising information on a time of the failure and information on a failed cell of the mobile communication network. In the next step, a set of possibly impacted cells is determined 42, based on the information on the failed cell. In the next step, subscriber activity data is obtained 44 for subscribers in the set of possibly impacted cells. Next, based on the subscriber activity data, a network level impact value indicating an impact of the failure to the mobile communication network is calculated 46. The network level impact value is stored 48 in a fault register, in association with the failure.

Fig. 5 shows a different representation of a method for determining an impact of a failure in a mobile communication network according to the present disclosure. The method described with regard to Fig. 4 may be considered a more generic representation of the method of Fig. 5.

In a first step 51, an FM alarm is received. The FM alarm may comprise a plurality of additional information, e.g., information on the failed cell (e.g., cell ID) and a timestamp of a time the failure was detected. In a second step 52, geo-location data and neighbor cell relation of cells is obtained/refreshed. This information helps to identify the impacted cells in the following step 53. In step 54, subscriber activity for the impacted cells before and after the failure is obtained. The subscriber activity may be recorded as a continuous process, e.g., at fixed time intervals, wherein each entry of subscriber activity data is provided with a corresponding timestamp. In step 55, the obtained data is correlated. In step 56, activity patterns are learned. In other words, the activity patterns may help to obtain a typical activity in the failed cell and the possibly impacted cells, i.e., a hypothetical activity in those cells for the case that the failure would not have occurred. In the next step 57, an impact analysis is performed on the basis of the data received and processed in the previous steps. As a result, a network level impact value is calculated and output. In step 58, an impact of the alarm (in the form of the network level impact value) is sent to a fault management (FM) system.

In the following, the technique of the present disclosure will be explained in more detail. The following description may apply both to the device for determining an impact of a failure in a mobile communication network according to Fig. 3 and the method for determining an impact of a failure in a mobile communication network according to Fig. 4.

Embodiments of the present disclosure correlate the customer experience information (service metrics), the FM/CM (fault management/configuration management) data, and the cell level reference data including geo-location and/or handover relation information. Then the impact of the cell outage is analyzed and a report is created. The report comprises at least one of:

The list of impacted (serving and neighboring) cells Per cell reports, i.e. a set of well-defined KPIs (RAN, Traffic and Service metrics, Handover failure increase) computed for before and after the incident List of subscribers categorized into at least one of the following categories: o Lost in the faulty cell o Moved from the faulty cell to a neighboring cell o Connected to a neighboring cell but impacted by the outage of the faulty cell (by the temporary load increase because of the outage) o Connected to a neighboring cell and not impacted by the outage List of impacted services for each subscriber and cell Calculated total impact of the outage based on the individual service impacts including spillover effects to neighboring cells (that helps alarm prioritization)

For estimating the impact of the cell failure, learned historical patterns may be used, which include cell traffic and handover statistics. According to one or more embodiments of the present disclosure, service quality, expressed as service related KPIs and their weighted combination, is used to quantify the failure and monitor the recovery of the system.

1.1 Solution overview

The device for determining an impact of a failure in a mobile communication network (and, more precisely, an impact analysis module of said device) may receive and process a plurality of data elements in order to output information on impacted cells and services and/or a subscriber impact report. One or more of the following data elements may be processed by the device: KPI pre-processing, correlation (QoE measurements), see item 1.2.2 below; fault management (time, cell identifier), see item 1.2.1 below; configuration management (neighbor information), see item 1.2.5 below; cell reference data (name, location, etc.), see item 1.2.4 below; and customer information (group info, ARPU, etc.), see item 1.2.3 below.

1.2 Input data that may be used for the analysis

According to an embodiment of the present disclosure, the analytics system of the device automatically collects the following information from external systems in near real-time. However, according to other embodiments, only a subset of the following information is collected and used by the device. In other words, the input data used for the analysis may comprise at least one of the following items:

1.2.1 Fault Management (FM) data

The analytics system subscribes for radio node and cell failure alarms in the network management FM systems. An FM ticket contains - among other data - one or more of the following key fields:

Failed network node / cell ID The timestamp of the failure (Type of the failure: planned / unplanned)

Details of failure

1.2.2 Customer Experience Information

Real-time E2E (end-to-end) customer experience information is obtained from a customer assurance system about active subscribers, including one or more of the following:

IMSI (International Mobile Subscriber Identity)

IMEI (International Mobile Equipment identity)

Time of activity

Used service(s), i.e., communication service (e.g. Web browsing, Video streaming, etc.)

Visited cell(s) within the activity

Subscriber's Quality of Experience (through a set of KPIs relevant for the used service(s)) 1.2.3 Customer Reference Information

The impact analysis obtains one or more of the following customer related information from customer reference data:

IMSI (International Mobile Subscriber Identity)

IMEI (International Mobile Equipment identity)

Customer type (VIP, etc.)

Customer service plan

ARPU (average revenue per user) history

1.2.4 Cell and radio base station reference information Base station, at least one of:

Geographical location, i.e. latitude, longitude coordinates Number of sectors

Cell, at least one of:

Radio access technologies (RAT, e.g. 3G, LTE, 5G)

Carrier frequency Beam direction Range

1.2.5 Configuration management (CM data)

From the CM system the following information is periodically obtained:

Neighbor cell relations

1.3 Learnt patterns

A learnt activity pattern may be considered by the analytics system of the device. In addition to the received data discussed above, the analytics system may consider at least one of the following learnt patterns:

1.3.1 Cell traffic

For each cell, the system maintains and stores historical traffic and QoE (Quality of Experience) data, at least one of: • Day of week and time of day

• Average number of active subscribers

• Average traffic load

• Average number of calls, active user sessions (per each service like web, video etc)

• Type of subscribers active

• Average KPI value levels per each used service

• Deviations for each average value listed above

1.3.2 Handover statistics

For each neighbor relation, the system collects and maintains handover statistics regarding the question what are the most frequent neighbors for each cell

• Having same RAT (Radio Access Technology)

• Having different RAT

1.4 Impact analysis process

In the following, an impact analysis process performed by the device of the present disclosure is described. As mentioned above, the impact analysis process may receive and process at least one of the above data and/or learnt patterns.

The following description is given with reference to Fig. 6, which shows an overview of a method according to an embodiment of the present disclosure.

The impact analysis process is triggered by a new incoming FM alarm. At first, a geographical area for which the impact is calculated must be determined. This is done by determining a set of radio network cells in the network that are located near the faulty cell contained in the FM alarm (so-called set of possibly impacted cells). After this, the correlation phase creates all the necessary data from the above input data.

1.4.1 Possibly impacted cell set determination (see "Impacted cells detection" in Fig. 6)

The determination of which cells to include in the impact analysis is based on at least one of the following data: Handover relations and geo-location data. One of these data types may be sufficient for determining the set of possibly impacted cells. Procedure: Starting from the cell (or base station) represented in the incoming FM alarm (i.e., the failed cell), a set of cells are determined. According to a first approach, the frequent neighbors of the given cell are selected to the set, by utilizing the handover relation data. According to a second approach, the cells within a given radius from the problematic cell are selected, by utilizing the geo-location information. Finally, according to a third approach, a mixture of approach one and two are applied, e.g. selecting cells from the neighborhood within a given distance. The process may be repeated for the different radio access technologies, by utilizing the geo-location of the failed cell. For example, in case of a 4G cell alarm, first the neighboring 4G cells are included in the set of possibly impacted cells, then further 3G cells are added to the set, those that are at the same geo-location as the problematic 4G ceil and then neighbors of those.

1.4.2 Data Correlation (see "Data Correlation" in Fig. 6)

At first, the FM alarm is enriched with all the details about the number and type of subscribers present in the actual cell (failed cell) and the previously determined potentially impacted cells, their activity types together with their perceived QoS (Quality of Service) per used communication service, one-by-one. The data is generated both for the latest timestamp available in the E2E monitoring system before the alarm timestamp, then for subsequent timestamps after the alarm timestamp.

The resulting data set of this correlation and enrichment is the input to the impact analysis, see the Impact Analysis section 1.4.3. The resulting data set may also be referred to as subscriber activity data.

According to an embodiment, the resulting correlated data set is structured as follows, by example:

Table 1:

Each line of the above subscriber activity data may be regarded as one data set. As shown above, a timestamp is stored as part of each data set. Further, the data sets contain data with regard to at least one of cell-ID, RAT, location (expressed, e.g., in latitude and longitude coordinates), subscriber ID (IMSI), communication service, and UE device type. Further, at least one KPI type and KPI value is assigned to each data set. In the above example, the KPI "stall ratio" of 0.1 % is stored as part of the first data set. Further, a historical KPI value is derived from the learned patterns, which indicates a "typical" KPI value to be expected at the time of the respective data set (in the above example, stall ratio = 0.001 %).

1.4.3 Impact Analysis (see "Impact Analysis" in Fig. 6)

A goal of the technique proposed by the present disclosure is to properly assess the identified faults (coming as input from external FM system), namely estimate the total impact (or pain) each reported fault represents for the subscribers in the operator's network. The estimation may include the spillover effects of faults in a cell to neighboring network nodes / cells. The larger the impact estimate, the more severe the problem is.

The initial result of the impact estimate (i.e., the network level impact value) is recorded in the central fault register as a new entry. The faults stored in the central fault register are then periodically updated with the actual impact enabling the continuous ranking of all the problems present at a time in the operator's network. When the original FM alarm is cleared, the fault entry is finally removed from the central fault register.

In the following, it will be focused on the initial calculation of the network level impact value. A procedure for the updates of an existing fault entry already present in the central fault register will be described below at item 1.4.3.2.

1.4.3.1 Procedure for the initial impact calculation for a new incoming FM alarm

Step 1. (New alarm) There is a new incoming FM alarm for cell C, with timestamp T. Current time is T_current.

Step 2. (Cell set) Determine the possibly impacted cell set (C_all) as described above (item 1.4.1). Step 3. (T_before) Let T_before be the time for which there exists Customer Experience data (see item 1.2.2) [depending on the resolution of the customer experience data collection process] and T_before<T and (T-T_before) is minimal. In other words, T_before is the closest time before the alarm for which there exist Customer Experience data.

Step 4. (T_after) Let T_after be the time T_after>T and T_after<=T_current for which there exists Customer Experience data (see item 1.2.2) [depending on the resolution of the customer experience data collection process] and (T_current- T_after) is minimal. In other words, T_after is the closest time to the current time T for which customer experience data exist and it is already past the FM alarm time.

Step 4a. If T_after cannot be determined (i.e., no such customer data exist that fulfils the time conditions), wait/sleep for a given period (a few seconds or minutes, depending on the resolution of the customer experience data collection process), then repeat Step 4.

Step 5. (User sets) Based on the customer experience data for time T_before and T_after, determine the following values and sets:

Let Userset_before be the set containing all the active users in the cell set C_all at time T_before. Let U_before be the cardinality of Userset_before.

Let Userset_after be the set containing all the active users in the cell set C_all at time T_after. Let U_after be the cardinality of Userset_after,

Step 6. (Historical estimate) Based on the learnt traffic patterns for cells and traffic (see item 1.3.1), determine the following values:

Let U_„before_history be the historical average estimate of the number of users in the cell set C_all for the time T_before.

Let U_after_history be the historical average estimate of the number of users in the cell set C_all for the time T_after.

Step 7. (Missing users) Estimate the number of users that are lost in the network due to the fault based on the values U_before, U_after, U_before_history and U_after_history. For example, the number NL of users that are lost (i.e., the number of subscribers that are lost) can be calculated as

NL = (U_before * (U_after_history / U_before_history)) - U_after In the above equation, the relation U_after_history / U_before_history indicates a typical relative loss/gain in the considered period. For example, in case it is known that the number of subscribers increases from 20 to 40 in the considered period, the above relation equals 2. This relation is considered as a factor in the above equation, such that it can be estimated, how many subscribers would be active after the fault in case the fault had not occurred. From this estimated number, the real number U_after is subtracted in order to estimate the numbers of subscribers that are lost.

Step 8. (Impact for active users) Estimate the possible KPI deterioration for the users that are active also after the FM alarm (i.e., they exists in Userset_before and Userset_after as well):

For each user, for each communication service included in the impact model, for each selected KPI type defined for the given service, compute the difference of the KPI values observed for the user at time T_before and T_after, and weight them according to predefined KPI weights per service (see item 1.7 below), to obtain the impact for the user per each defined service; then

Sum the individual service impacts for each user according to the defined service weights to a total user impact value; then Sum up the computed total user impact values for all active users to obtain the final impact value for all active users (see item 1.7 for details on weights, KPIs, services).

Step 9. (Average KPI impact) Based on the difference between the historical average KPI levels for each defined communication service in the impact model and the KPI levels at T_after, using the KPI weights and service weights (see item 1.7), estimate the generic KPI impact for cells in C_all. This estimate will enable the impact calculation extension to users who were not present in C_all at T_before.

Step 10. (Impact calculation) Based on the estimated number of missing users (Step 7), the computed final impact value for all active users (Step 8), and the average KPI impact (Step 9), compute the networkjeveljmpact (= network level impact value) for the FM alarm, and store it into the central fault register. Store T,C, C_all,T_lastupdate and the entry (T_after, networkjeveljmpact) as the first computed <time,impact> value for the alarm, where TJastupdate equals to T_after.

For storing the network level impact value NLIV as one single value considering each of the values calculated in steps 7, 8, and 9, NLIV may be calculated as: NLIV = a*FIV + b*NOSTAL + c*AKIV, wherein FIV is the final impact value, NOSTAL is the number of subscribers that are lost, and AKIV is the average KPI impact value. The parameters a, b, and c are predefined weighting and normalization factors.

It is noted that it may be sufficient to calculate NLIV only based on one or two of the parameters FIV, NOSTAL, and AKIV. In this case, the other two or one parameters are either not considered or are not even determined. For example, it may be sufficient to calculate the network level impact value NLIV only based on the final impact value FIV and the numbers of subscribers that are lost NOSTAL:

NLIV = a* FIV + b*NOSTAL.

In another exemplary embodiment, the NLIV equals the number of subscribers that are lost:

NLIV = NOSTAL.

1.4.3.2 Procedure for periodic update of the impact of an existing fault entry in the central fault register

According to an embodiment, the FM alarm entries stored in the central fault register are periodically updated. The update frequency depends on the availability of the new customer experience data set (the resolution of this data). The following procedure is performed, when a given entry (C,T,C_all,TJastupdate) is updated and a new <T_after,networkJevel_impact> entry is calculated for the existing alarm for the latest impact estimation: T_after is the closest time to the current time for which customer experience data exists and it is already past the time when the FM alarm impact was last updated (TJastupdate).

Step 4a. If T_after cannot be determined (i.e., no such customer data exist that fulfils the time conditions), wait/sleep for a given period (a few seconds or minutes, depending on the resolution of the customer experience data collection process), then repeat Step 1. Step 2. (User numbers) Based on the customer experience data for time T_after, and the learned cell and traffic patterns, determine the following values:

Let U_after be the number of the active users in the cell set C_all at time T_after

Step 3. (Missing users) Estimate the number of users that are lost in the network due to the fault based on the values U_after and U_after_history.

Step 4. (Average KPI impact) Based on the difference between the historical average KPI levels for each defined communication service in the impact model and the KPI levels at T_after, using the KPI weights and service weights (1.7), estimate the generic KPI impact for cells in C_all.

Step 5. (Impact calculation) Based on the estimated number of missing users (Step 3) and the average KPI impact (Step 4), compute the networkjeveljmpact for the FM alarm, and update it in the central fault register. Add the entry (T_after, networkjeveljmpact) as the latest computed <time,impact> value for the alarm, and change TJastupdate to be equal to T_after.

Repeat the above process for all the entries in the central fault register periodically.

1.5 Output reports - Impacted cells and services

According to an embodiment, the proposed system generates a report on the impacted cells. The generated report may include (but is not limited to) at least one of:

List of impacted cells. The list can be ordered by cost / services / service impact to support the different type of analyses and workflows at the operators.

Cell level before-after reports around the incident / outage. Reports contain the most important infrastructure, traffic and service related KPIs, and show the degradation around the outage. The report may include (but is not limited to) at least one of the following main categories and KPIs: o RAN KPIs to describe the radio environment and infrastructure (e.g. RSRP, RSRQ, CQI, RAN drops) o Traffic KPIs to describe the cell traffic and load (e.g. data volume, cell throughput, number of active users) o Service KPIs to describe the service experience of the different services MBB (data volume and service usage, TCP level metrics: throughput, packet loss, delay)

Volte (voice quality metrics, call drop, call setup failure, soft drops..)

1.6 Central fault register

Each FM alarm will trigger the impact analysis and beyond the reports above, the results will be stored in the central fault register. According to one or more embodiments, the records in the fault register are continuously updated with a chosen frequency, e.g. every 5-10 minutes. The records in the fault register are deleted whenever the original alarm was cleared.

According to one or more embodiments, the fault register has at least the following fields per entry:

Original timestamp of FM alarm Node/Cell ID

Possibly impacted cell set (computed initially when the FM alarm arrived) (Timestamp, Impact) vector of computed impact values

1.7 Monitored KPIs for services

The procedures described in item 1.4.3 for the impact analysis contain several domain specific details. These include the following:

1. For each communication service (e.g., web browsing, video streaming), o What KPIs to use in the impact calculation for a service o What is the "direction" of each selected KPI (larger the better or lower the better) o What are the relative weights for each KPIs selected

2. What are the relative weights among each service that is used in the model

3. What are the relative weights among different customer types (based on ref data)

One embodiment is as follows, regarding these important aspects of the calculation model: Table 2:

Example for KPI weights: Table 3:

The above tables are stored in the device and may be adapted, if necessary. Similar to the above KPI weights, service weights are assigned to the individual communication services.

The technique for determining an impact of a failure in a mobile communication network described herein may have one or more of the following advantages (depending on the details of the considered embodiment).

The operator receives prompt feedback about the number of affected subscribers and services related to a radio node or radio nodes failure in case of overlapping cells. The system may provide detailed information on which services and which subscribers are affected and to what extent. This information is primarily useful for prioritization of fault management alarm tickets. The operator can focus on solving issues which highly impact users and services. This information can also be used for customer care to improve customer relation management processes, e.g. offer compensation, or send notification and acknowledge that there was a problem. The solution also identifies if services and subscribers were not impacted by a radio node failure or outage. The solution can identify alarms that can be cleared/ignored, namely reduce the alarms the operator should handle.

The solution may provide information about the robustness, redundancy of the radio network, may identify the affected neighbor cells, may point out areas, where neighbor cells can take over traffic without any major problem; and may identify areas, where outage can cause coverage or capacity issues. The solution and system may identify the critical areas where network improvements, capacity should be increased.

The system may distinguish two different failure types: planned outage when traffic is taken over in a planned way and sudden failure when fast processes try to solve the issue.

The system may determine the impacted subscribers and services lost in the faulty cell, the subscribers and traffic taken over by neighboring cells, the subscribers and services influenced by taking over subscribers and traffic from the faulty cell within a few minutes of the failure.

The system also may provide information in a longer time scale (e.g., a few hours) about the key performance indicators of the affected cells and area. Using the change of the neighbor relation and the handover KPIs among the neighbor cells the system may estimate the remaining coverage hole due to cell or radio node failure, considering the redundancy of the radio network in the affected area.

The solution provides a generic way to estimate the impact of each radio network failure to support the network operator to focus on the most severe issues first, especially including the spillover effects of radio failures to neighboring cells in the estimation.

Many advantages of the present disclosure will be fully understood from the foregoing description, and it will be apparent that various changes may be made in the form, construction and arrangement of the units and devices without departing from the scope of the present disclosure and/or without sacrificing all of its advantages. Since the embodiments can be varied in many ways, it will be recognized that the present disclosure should be limited only by the scope of the following embodiments.

Claims

1. A device (30) for determining an impact of a failure in a mobile communication network, the device comprising: at least one processor (34); and memory (36) comprising instructions executable by the at least one processor (34), wherein the device (30) is operable to: receive fault management, FM, data informing of the failure in the mobile communication network, the fault management data comprising information on a time of the failure and information on a failed cell of the mobile communication network; determine (42) a set of possibly impacted cells, based on the information on the failed cell; obtain (44) subscriber activity data for subscribers in the set of possibly impacted cells; calculate (46), based on the subscriber activity data, a network level impact value indicating an impact of the failure to the mobile communication network; and store (48) the network level impact value in a fault register, in association with the failure.

2. The device of claim 1, further configured to determine the set of possibly impacted cells based on the information on the failed cell and based on handover relation data indicative of previous handover processes between the failed cell and neighboring cells.

3. The device of claim 1 or 2, further configured to determine the set of possibly impacted cells based on the information on the failed cell and based on geo-location data indicative of geographic locations of a plurality of cells including the failed cell.

4. The device of any of claims 1 to 3, wherein the subscriber activity data comprises subscriber activity data for subscribers in the failed cell.

5. The device of claim 4, wherein the subscriber activity data comprises at least one data set for each subscriber present in the failed cell and in the possibly impacted cells, the data set defining a type and/or identification number of the respective subscriber, wherein the subscriber activity data is indicative of a number of the subscribers present in the failed cell and in the possibly impacted cells.

6. The device of claim 5, wherein at least one of the data sets comprises a communication service used by the respective subscriber and a key performance indicator, KPI, value for the communication service.

7. The device of claim 5 or 6, wherein each of the data sets comprises a time stamp and wherein the subscriber activity data comprises data sets with time stamps indicating a time before the time of the failure and data sets with time stamps indicating a time after the time of the failure.

8. The device of claim 7, further configured to calculate the network level impact value based on a final impact value for subscribers in the set of possibly impacted cells that are active after the time of the failure.

9. The device of claim 8, wherein calculating the final impact value comprises: determining, based on the subscriber activity data, a set of active subscribers that were active in the possibly impacted cells before and after the time of the failure; for each subscriber of the set of active subscribers, for each communication service, and for each KPI type defined for the given communication service, computing a difference of the KPI values observed for the subscriber before the time of the failure and after the time of the failure, and weighting the differences according to a predefined KPI weight per communication service, to obtain a service impact value for the subscriber per each defined communication service; weighting the individual service impact values for each subscriber with a predefined service weight value and summing the weighted service impact values to obtain a total subscriber impact value; and summing the total subscriber impact values for each of the active subscribers to obtain the final impact value.

10. The device of any one of claims 1 to 9, further configured to calculate the network level impact value based on a number of subscribers that are lost in the mobile communication network due to the failure.

11. The device of claim 10, further configured to calculate the number of subscribers that are lost based on learnt cell traffic data, the learnt cell traffic data comprising an average estimate of a number of subscribers in a specific cell for a specific time.

12. The device of claim 10, further configured to calculate the number of subscribers that are lost based on: a number of subscribers active before the time of the fault, derived from the subscriber activity data; a number of subscribers active after the time of the fault, derived from the subscriber activity data; an average estimated number of subscribers active before the time of the fault, derived from the learnt cell traffic data; and an average estimated number of subscribers active after the time of the fault, derived from the learnt cell traffic data.

13. The device of claim 7, further configured to calculate the network level impact value based on an average KPI impact value, wherein the average KPI impact value is calculated based on historical average KPI levels for each defined communication service, and based on the KPI values of the subscriber activity data after the time of the impact, using predefined KPI weights per communication service and predefined service weight values.

14. The device of claims 8, 10, and 13, further configured to calculate the network level impact value based on the final impact value, based on the number of subscribers that are lost, and based on the average KPI impact value.

15. The device of any of claims 1 to 14, further configured to update the network level impact value stored in the fault register, in association with the failure.

16. A method for determining an impact of a failure in a mobile communication network, the method comprising: receiving (40) fault management, FM, data informing of the failure in the mobile communication network, the fault management data comprising information on a time of the failure and information on a failed cell of the mobile communication network; determining (42) a set of possibly impacted cells, based on the information on the failed cell; obtaining (44) subscriber activity data for subscribers in the set of possibly impacted cells; calculating (46), based on the subscriber activity data, a network level impact value indicating an impact of the failure to the mobile communication network; and storing (48) the network level impact value in a fault register, in association with the failure.

17. The method of claim 16, wherein determining the set of possibly impacted cells based on the information on the failed cell is further carried out based on handover relation data indicative of previous handover processes between the failed cel! and neighboring cells.

18. The method of claim 16 or 17, wherein determining the set of possibly impacted cells based on the information on the failed cell is further carried out based on geo-location data indicative of geographic locations of a plurality of cells including the failed cell.

19. The method of any of claims 16 to 18, wherein the subscriber activity data comprises subscriber activity data for subscribers in the failed cell.

20. The method of claim 19, wherein the subscriber activity data comprises at least one data set for each subscriber present in the failed cell and in the possibly impacted cells, the data set defining a type and/or identification number of the respective subscriber, wherein the subscriber activity data is indicative of a number of the subscribers present in the failed cell and in the possibly impacted cells.

21. The method of claim 20, wherein at least one of the data sets comprises a communication service used by the respective subscriber and a key performance indicator, KPI, value for the communication service.

22. The method of claim 20 or 21, wherein each of the data sets comprises a time stamp and wherein the subscriber activity data comprises data sets with time stamps indicating a time before the time of the failure and data sets with time stamps indicating a time after the time of the failure.

23. The method of claim 22, wherein calculating the network level impact value is based on a final impact value for subscribers in the set of possibly impacted cells that are active after the time of the failure.

24. The method of claim 23, further comprising calculating the final impact value, wherein calculating the final impact value comprises: determining, based on the subscriber activity data, a set of active subscribers that were active in the possibly impacted cells before and after the time of the failure; for each subscriber of the set of active subscribers, for each communication service, and for each KPI type defined for the given communication service, computing a difference of the KPI values observed for the subscriber before the time of the failure and after the time of the failure, and weighting the differences according to a predefined KPI weight per communication service, to obtain a service impact value for the subscriber per each defined communication service; weighting the individual service impact values for each subscriber with a predefined service weight value and summing the weighted service impact values to obtain a total subscriber impact value; and summing the total subscriber impact values for each of the active subscribers to obtain the final impact value.

25. The method of any one of claims 16 to 24, wherein calculating the network level impact value is based on a number of subscribers that are lost in the mobile communication network due to the failure.

26. The method of claim 25, further comprising calculating the number of subscribers that are lost, wherein calculating the number of subscribers that are lost is based on learnt cell traffic data, the learnt cell traffic data comprising an average estimate of a number of subscribers in a specific cell for a specific time.

27. The method of claim 25, further comprising calculating the number of subscribers that are lost, wherein calculating the number of subscribers that are lost is based on: a number of subscribers active before the time of the fault, derived from the subscriber activity data; a number of subscribers active after the time of the fault, derived from the subscriber activity data; an average estimated number of subscribers active before the time of the fault, derived from the learnt cell traffic data; and an average estimated number of subscribers active after the time of the fault, derived from the learnt cell traffic data.

28. The method of claim 22, wherein calculating the network level impact value is based on an average KPI impact value, wherein the average KPI impact value is calculated based on historical average KPI levels for each defined communication service, and based on the KPI values of the subscriber activity data after the time of the impact, using predefined KPI weights per communication service and predefined service weight values.

29. The method of claims 23, 25, and 28, wherein calculating the network level impact value is based on the final impact value, based on the number of subscribers that are lost, and based on the average KPI impact value.

30. The method of any of claims 16 to 29, further comprising updating the network level impact value stored in the fault register, in association with the failure.

31. A computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method of any of claims 16 to 30.

32. A computer-readable medium comprising instructions which, when executed by a computer, cause the computer to carry out the method of any of claims 16 to 30.