The primary objective of the AI-NET research program is to Accelerate digital transformation in Europe by Intelligent NETwork automation at each network segment, i.e., edge, metro, core and data centers.
Complete network automation is a clear prerequisite for the efficient use of highly integrated and flexible edge infrastructures which are programmable across all its components, from basic connectivity setup to fully virtualized network functions and application components.
AI-NET will explore a number of use cases spanning the technology challenges of services being deployed and operated at the network edge, in order to distill the various scenarios and deployments of each use case, and thereby the technical requirements as well as the values.
AI-NET will research and develop technologies specific for an edge infrastructure, which is characterized by a large number of edge locations, heterogeneous hardware and site configurations, resource constrained compute environments, a mix of base technologies for virtualization platforms and transport networks, and finally supporting critical services in customized network slices.
To manage the resulting complexity, we need to take advantage of artificial intelligence (AI) and machine learning (ML) to complement or replace traditional optimization and prediction algorithms. The edge platform will be an efficient way to support a learning infrastructure.
Finally, a deep edge network will be deployed at locations which have not been prepared for the power requirements of data centers, even small ones. AI-NET will analyze requirements on deployment and power supply for the edge infrastructure and develop methods to minimize power consumption.
The benefit for participating partnering industries will be the access to new knowledge demonstrated in proof of concepts running in test-beds and presented in scientific papers. The PhD candidates will also find their way to the industry R&D departments improving the skill set and capacity of European industry. New IPRs, products and service concepts will be developed and enhance the business performance of the participants. The project will also strengthen the European leadership in mobile communication and challenge the global competition in cloud, edge and artificial intelligence technologies.
GWDG will introduce an automation mechanism of legacy networks with the help of OpenFlow agents. An OpenFlow agent plays the role of a mediator between an SDN controller and a non-SDN data plane. It receives the OpenFlow instructions coming from the controller, translates them into data plane specific language and pushes them towards the data plane forwarding agents. Furthermore, GWDG aims to contribute with an advanced monitoring system with the use cases of traffic accounting and security monitoring of the live network in mind. The proposed system entails the collecting, processing and storing of individual flow records for the said purpose.
In the context of Security, GWDG envisages an effort to enhance and improve existing ML/AI solutions for detecting/mitigating different kinds of network attacks. We take advantage of the fact that a centralized controller is the best place to monitor the flows for detecting any abnormality in the network and therefore, together with the programmability the mitigation process will be deployed much quicker and more efficiently. The SDN controller enables the security applications to decide when and how to monitor the network dynamically, instead of running an intrusion detector at a few critical points.
Project Coordinator: ADVA Optical Networking SE, Germany
A complete list of project partners can be seen here: https://www.celticnext.eu/project-ai-net-protect/