Software Defined Networks/Network Functions Virtualization (SDN/NFV) market is growing massively but users struggle to have complete visibility and control from their virtualized network. Usual problems in this area are: (i) Large number of Service Level Agreement (SLA) violations caused by poor SLA definition and management in complex multi-vendor, multi-technology systems; (ii) Capex impacted negatively because SDN service provisioning ignores SLAs, (iii) High downtime experienced by subscribers because of high latency between service disruption and healing.
In this project, we will execute research activities to allow the creation of SLA generators and recommendation based on historical and real-time data to reduce SLA violations, SLA-aware provisioning systems based on learned behavior & predictive insight, and automated real-time configuration update and self-healing based on continuous feedback. To do so, we plan to bring knowledge both from the networking domain and from the machine learning area. Particularly, we plan to explore the use of novel deep learning techniques to understand the effect of different network topologies on key network metrics and automatically suggest network configuration and topology changes.