A method for identifying faulty cells using a classification tree-based UE diagnosis in LTE

Muñoz et al. EURASIP Journal on Wireless Communications and Networking (2017) 2017:130 DOI 10.1186/s13638-017-0914-3 RESEARCH Open Access A method for identifying faulty cells using a classification tree-based UE diagnosis in LTE P. Muñoz1* , R. Barco1 , E. Cruz2 , A. Gómez-Andrades1 , E. J. Khatib1 and N. Faour2 Abstract The latest advances in wireless technologies have led to a proliferation of data mobile devices and services. As a consequence, mobile networks have experienced a significant increase in data traffic, while voice traffic has shown nearly no growth. It is therefore essential for operators to understand the data traffic behavior at the user level in order to ensure a good customer experience. In the radio access networks (RANs), traditional solutions based on cell-level measurements are not adequate to analyze performance of individual users. Instead, novel alternatives such as the use of call traces and the definition of new user-centric indicators will provide detailed and valuable information for each connection. One of the key measurements related to data services is the user throughput. In this work, the user throughput is adopted as the main attribute to conduct diagnosis in the RAN, which has typically been the bottleneck for data services. To that end, a binary classification tree is proposed to determine the root cause of poor throughput in user-level data sessions. Then, this information is aggregated at the cell level in order to provide effective diagnosis of degraded cells. In particular, a correlation-based analysis of the cell status is proposed in order to identify abnormal cell behaviors in an automatic way. Evaluation has been carried out with datasets from live cellular networks. Results show that the proposed diagnosis approach is an effective means to identify the main factors that limit the user throughput in network cells. Keywords: Self-healing, Fault diagnosis, Long-Term Evolution, Correlation, Self-Organizing Networks 1 Introduction During the last years, the wireless data services have become the dominant traffic source in cellular networks. Behind this, there is an expansion of new mobile applications and a rapid growth in the number of subscribers, both motivated by the advances in cellular communication technologies and the development of user-friendly smartphones. According to a large network vendor [1], global mobile data traffic grew 69% in 2014 while the average smartphone usage grew 45% in the same year. This enormous increase in data traffic has forced operators not only to invest large amounts of money in new infrastructure but also to reduce operational expenditures (OPEX) in order to maintain the levels of user satisfaction. To produce significant cost savings, one of the adopted solutions by standardization bodies was the creation of *Correspondence: Communications Engineering Dept., University of Málaga, Málaga, Spain Full list of author information is available at the end of the article 1 the Self-Organizing Networks (SONs) [2], which provide a new concept of network management where the maintenance and optimization tasks are carried out mostly in an automated way. Typically, technical experts in these fields have to deal with hundreds of traffic measurements and performance indicators every day [3, 4]. The vast diversity and quantity of these metrics makes the operational work very complex. Thus, the use of automated techniques for cellular traffic data analysis is essential to reduce human effort while expertise can be focused on new areas, bringing additional value to the operator [5]. Traditionally, mobile operators paid their attention in providing a good quality of the voice service, since it was the main offered service. To ensure this Quality-ofService (QoS), troubleshooting experts mainly monitored the call blocking and dropping rates at the cell level to measure the levels of accesibility and retainability, respectively, in the network. However, with the explosion of Internet services, the QoS of multimedia and data applications is given by the data rates experienced by the users, © The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Muñoz et al. EURASIP Journal on Wireless Communications and Networking (2017) 2017:130 where integrity metrics such as throughput and latency are essential traffic measurements [4]. The problem of throughput performance indicators is that they are often difficult to interpret because of their dependence on many factors. In particular, there are some aspects beyond the typical variables related to the radio environment (e.g., distance to base station, cell loading, user speed, etc.) that should be considered. First, unlike in traditional voice services, the mobile network is only one segment of the endto-end connection in an IP world. For example, a router in the IP cloud that suffers congestion may influence the user perceived data rate. Second, the recent radio access technologies (RATs) such as Long-Term-Evolution (LTE) have included a class-based QoS model as a mechanism to differentiate between services, establishing various levels of service to the users. Third, the traffic pattern of new data services clearly impacts throughput measurements. Due to the increasing popularity of web navigation, streaming video, social networking, file sharing, online gaming, and other data services, there are significant differences in traffic patterns [6]. As a consequence, operators are investing a large amount of money to investigate traffic modeling and classification through packet inspection in order to better understand the characteristic of today’s cellular data traffic. In addition, sophisticated traffic data filtering, processing, and correlation with other network metrics are also important features to identify root causes of any detected anomaly and increase the reliability of the network [7, 8]. The increasing complexity of network infrastructure and services has also led operators to be interested in managing performance at the user level, instead of the cell- or network level, with the aim of maintaining their competitiveness levels. Today’s solutions based on percell performance counters are insufficient to perform adequate root-cause analysis. For this reason, the standardization bodies have proposed the use of user-centric indicators and call traces to support the optimization and troubleshooting processes [9]. With the Minimization of Drive Tests (MDT) described in [10], the collection of traffic measurements can be done in an autonomous manner. In other words, each device that is active in (...truncated)