Cognitive Management Systems
Cognitive Management Systems Fraunhofer FOKUS

Cognitive Management Systems

The 5G environment is foreseen to become even more complex, beyond human administration capabilities, due to the integration of a new 5G access network, a high diversification of the radio-access within the same technology, convergence with other wireless access technologies and backhaul, massive parallelization, separate administration on top of a single shared physical infrastructure and the localization and distribution of applications. In this context, we are able to provide research in the area of:
  • Increasing trust into automatic management features;
  • Harmonized/convergent management plane – for multi-domain (radio, core, backhaul, virtualisation, etc.), for multi-tenant (sharing the infrastructure) and for multi-operator (roaming, infrastructure sharing, etc.);
  • Prediction and learning algorithms, minimizing human intervention (leveraging the internal information);
  • Understanding and adapting based on the human situational context (leveraging the human information);
  • Rapid responses to exceptional cases (failure, congestion, etc.);
  • Direct application to the network components of self-configuration, self-optimization, self-healing and self-protection mechanisms. Their integration;
  • Cost-benefit analysis in a comprehensive system;
  • Definition of the appropriate integrated KPIs for automation;
  • Methodology mechanism for assessing possible actions in exceptional cases;
  • Reducing the specialization required for administrating a complex system.

Gaining trust

One very large step, delaying up to completely stopping the adoption of the new management functionality is the lack of trust in such technologies from the Telco operators. In order to gain this trust, vendors and research have to work together in defining the appropriate KPIs for such systems as well as to practically prove that the benefits of such automated solutions are highly overcoming the costs while ensuring the same service robustness as the current network architectures.
For this, there is a stringent need for a methodology and a benchmarking mechanism supporting the most exceptional cases (e.g. overload, massive failures, etc.) through which the technology can be assessed for capacity and high availability. However, the current system was designed based on silo-ed managed domains. For being able to define this methodology, first a critical overview of the required live system parameters has to be made, simplifying, through convergences, the management system (e.g. which are the KPIs of the overall 5G system, which are the system parameters influencing these KPIs and in which proportion) and then taking a further look on how to realize the most appropriate and robust adaptation decisions.
It is important to mention that a certain human interaction will still be needed to overlook the network and for setting the governance policies and taking care of the major exception cases, especially without requiring very specialized, highly educated experts (e.g. you don’t need to be able to repair the automatic transmission functions of a car to get a drivers’ license). The most appropriate parameters have to be exposed (even by considering multiple administrative perspectives), in order to be able to enable the current network administrator to manage more complex and more automated networks.