Quality of Service enhancement with Resilient routing and Machine learning

By now, the Internet has become a topmost critical infrastructure. Current technology trends point toward a new era of dependable Internet, which includes the concepts of the Internet of Everything, Network Intelligentization, Vehicle-to-everything communication, Augmented and Virtual Reality applications, and many others. Telesurgery, autonomous driving, Industry 4.0 and the stock market are just a few examples of the emerging mission-critical services that involve telecommunication and on which both people and governments increasingly rely. To address the rising challenges of these new mission-critical application-based concepts, ultra-reliable and low-latency communication channels have to be established. Despite this reliance on the Internet, nowadays, it often falls behind the needs and expectations. To enable the advent of a truly dependable Internet, the main goal of QoSeRM is to drastically improve the most important Quality of Service (QoS) metric, the end-to-end availability, that measures the fraction of time in which two endpoints are able to communicate. To create such a reliable Internet, we have to step further from protecting only the classical single network equipment failures and ensure that telecommunication remains operational in the presence of large-scale regional failures caused by natural or man-made disasters too. To this end, QoSeRM forges two major building blocks into an overarching framework, namely: 1) enhanced time series prediction techniques for predicting future spare link bandwidths, and 2) scalable resilient routing algorithms for utilising the spare link capacities for sending redundant data between the communicating endpoints. Thus, my findings will combine the strengths of artificial intelligence and combinatorial optimisation, which are two fields with very different philosophies: while the former says that ‘data is better than algorithms’, the creed of the latter is contrary: `algorithms are better than data’.