Back to search

IKTPLUSS-IKT og digital innovasjon

DOMINOS: Dissecting and Modeling Interdependencies in Communication Networks

Alternative title: DOMINOS: dissekere og modellering av avhengigheter i Kommunikasjonsinfrastruktur

Awarded: NOK 7.0 mill.

Our society is becoming increasingly reliant on the Internet and cellular networks for both critical and leisure services. At the same time, these networks are becoming increasingly complex. Network operators depend on each other to extend their reach and to increase their capacity. Furthermore, the interaction between these networks and their surrounding eco-system, encompassing everything from the physical infrastructure they run on to the policies regulating their operations, are far from being understood. Understanding these interdependencies is crucial for avoiding large-scale failures. Regulators, network operators, and customers are all interested in understanding these dependencies.This understanding will help regulators in devising better policies to ensure the stability of these critical infrastructures. Operators will leverage such understanding to improve their services, and customers will be able to make informed decisions when buying communication services. DOMINOS focused on investigating interdependencies in complex systems that are organized as a network of networks with emphasis on communication networks. It followed an experimental approach by measuring real systems. Over the past few years, we have developed empirical approaches form inferring interdependencies in measurements data. These approaches were applied to measurements of failures in Norwegian mobile networks to quantify the interdependence between operators. Our findings indicate that most outages affect only one network operator, however, shared infrastructure remains a major vulnerability. Our results are communicated back to relevant stakeholders and will hopefully help informing the building of the next generation emergency networks. We have also modelled cascading failures due to attacks that overload network components, which results in redistributing network traffic triggering further failures. We have also investigated methods for monitoring a complex multi-tenant, multi-service network. An example of such a network is the upcoming fifth generation mobile networks. These networks are designed to allow services with diverse requirements to co-exist over the same physical infrastructure. To this end, we leverage advances in machine learning and data analytics. The initial results are promising and can be extended further to realizing networks and infrastructures that are autonomous and to a large extent self-driving.

DOMINOS has contributed to expand our understanding about how to empirically characterising correlations within a complex system. These results will help us monitoring interdependencies in the upcoming multi-service multi-tenant networks that cater for even more critical services e.g. 5G networks. Going forward, this may help conceptualising new projects that aim to monitor and control interdependencies in real-time. DOMINOS results have been communicated to relevant stakeholders, including telecom operators, regulators and use case owners like emergency networks. We plan to continue sharing our insights and engage in active dialogue with these stakeholders about the building of the next generation emergency networks on commercial networks. The approaches, that we have developed in DOMINOS, for identifying correlations can be applied to other systems, which we envision to help us in building new interdisciplinary collaborations.

In today's society, communication infrastructures are evolving to be the single most important point of failure. At the same time, the complexity of these networks is increasing at an unfollowable pace. The interaction between these networks and their surrounding eco-system, encompassing everything from the physical infrastructure they run on to the policies regulating their operations, are far from being understood. Understanding these interdependencies is crucial for avoiding large-scale failures. This proposal aims to investigate interdependencies in complex systems that are organized as a network of networks with emphasis on communication networks. We will follow an experimental approach by measuring real systems. Then develop agent-based computational model that adequately resembles real networks and use to answer intricate what-if questions. We believe that this work can help both expanding our understanding of interdependencies in communication networks as well as serve as an input to other disciplines that study complex interdependent networks.

Funding scheme:

IKTPLUSS-IKT og digital innovasjon