Page 51 - Proceedings of the 2018 ITU Kaleidoscope
P. 51

SELF-HEALING AND RESILIENCE IN FUTURE 5G COGNITIVE AUTONOMOUS
                                                     NETWORKS




                                                                    2
                                        1
                                                     2
                          Janne Ali-Tolppa ; Szilard Kocsis ; Benedek Schultz ; Levente Bodrog ; Marton Kajo 3
                                                                                  2
                                              1 Nokia Bell Labs, Munich, Germany
                                             2 Nokia Bell Labs, Budapest, Hungary
                                       3 Technische Universität München, Munich, Germany

                                                              Mobility Load Balancing (MLB) -- optimize the network
                              ABSTRACT                        function configuration so that specified objectives are
                                                              fulfilled. [2]
           In the Self-Organizing Networks (SON) concept, self-healing
           functions are used to detect, diagnose and correct degraded   While self-optimization  functions optimize a set of
           states in the managed network functions or other resources.   configuration parameters  to improve the  network
           Such methods are increasingly important in future network   performance  from a  given  network state, the self-healing
           deployments, since ultra-high reliability is one of the key   functions  focus on ensuring that the  system can  fulfill its
           requirements for the future 5G mobile networks, e.g. in   purpose and serve its customers even in case of unexpected
           critical machine-type communication. In this paper, we   changes or events that risk a degradation in the network
           discuss the considerations for improving the resiliency of   performance. In other words, their objective is to make the
           future cognitive autonomous mobile networks. In particular,   network more resilient [3, 4, 5].
           we present an automated anomaly detection and diagnosis
           function for SON self-healing based on multi-dimensional   In this paper, we discuss the  methods for improving
           statistical methods, case-based reasoning and  active   resiliency in future 5G mobile networks and present a SON
           learning techniques. Insights from both the  human expert   self-healing function applying automatic anomaly detection
           and sophisticated machine learning methods are combined   and diagnosis that enables detection of unexpected changes
           in an iterative way. Additionally, we present how a more   and events and allows  fast  reaction to them. Unlike in
           holistic view on mobile network self-healing can improve its   optimization, the  self-healing functions can  typically only
           performance.                                       monitor the symptoms of a fault and the causes are not a
                                                              priori known. Due to this and the less-constrained problem
            Keywords – SON, cognitive network management, self-  space, the self-healing process is often more complex than
                 healing, anomaly detection, machine learning   the optimization control loops.  This complexity is
                                                              accentuated by the distributed and heterogenous  nature of
                         1   INTRODUCTION                     mobile  networks. Finally, diverse  fault  states often occur
                                                              only in very rare cases, which makes it impossible to collect
           Mobile networks are becoming ever more complex, while   statistically  meaningful data for each case. The lack of
           the pressure for reducing the operational expenses and cost   statistical samples  makes the reliable root-cause analysis
           per bit is increasing. Network management automation is a   extremely difficult. Therefore, mobile network self-healing
           key enabler for the increased dynamicity of the future 5G   functions typically  require more  sophisticated machine
           networks. Furthermore, the dynamic and complex  future   learning and augmented intelligence  methods, as  well as
           networks require automation that can autonomously adapt to   knowledge-sharing between domains. In our concept we’ve
           changes in the context and in the environment,  which   applied a holistic view on the network to detect the diverse
           necessitates the use of  machine learning, analytics and   problem states that may occur, as well as considered active
           artificial intelligence. [1]                       and transfer learning methods.
           Traditionally, the  Self-Organizing Networks (SON)  use   The rest of the paper is structured as follows. In section 2,
           cases are divided into three  categories:  self-configuration,   we present the main principles of resilient systems and how
           self-optimization and self-healing. Self-configuration refers   they map to the future cognitive mobile networks. Section 3
           to automated, “plug and play” deployment of new network   focuses on SON self-healing as an enabler for resiliency and
           functions, which are implemented by SON functions such as   gives an overview of the self-healing process. In sections 4
           the Automatic Neighbor Relationship (ANR) detection and   and 5, we look more closely to the most important phases of
           Physical Cell Identifier (PCI) allocation. Self-optimization   the self-healing process, namely the anomaly detection and
           functions -- for example Mobility Robustness Optimization   diagnosis, respectively. Section 6 outlines how a  more
           (MRO), Capacity and Coverage Optimization (CCO) and   holistic  view on the complete network, over several





          ‎ 978-92-61-26921-0/CFP1868P-ART‎                – 35 –                                     Kaleidoscope
   46   47   48   49   50   51   52   53   54   55   56