Advanced Peering with a Software Defined Knowledge Plane (2014 – 2016). Funded by Ericsson.
The main contributions envisioned for this project are the research, design and prototypical validation of a Knowledge Plane suitable for Software Defined Networks (SDN) that shall allow novel networking features and applications. The architecture will be exercised in the context of Internet Exchange Points (IXP) scenarios where participants implement SDN technologies inside their domains and wish to collaborate beyond the peering capabilities of the traditional BGP routing protocol. Suitable peering agreements and inter-SDN protocols will be investigated. A broad approach to SDN will be adopted where OpenFlow is regarded as one of the protocols used to bidirectionally and programmatically communicate with network elements. The proposed broad approach to SDN retains a clean separation of control and data planes but considers and hybrid approach where the logically centralized SDN control plane logic and actions are meant to augment existing control plane technologies, allowing partial and gradual deployability. The Knowledge Plane architecture will be conceived by leveraging the separation between the forwarding and control planes to build a rich base of information among multiple dimensions (multiple layers, actors, constraints) and time windows (long/mid term and real-time stats) upon which Big Data analytics can be applied to generate the “knowledge plane” that will expose programmatic interfaces (APIs) to applications in order to deliver usable services based on the interactions with the Knowledge Plane. Intra-domain and single-site deployments as well as multi-site deployments of the Knowledge Plane instantiations will be investigated. The experimental validation includes software prototypes based on best of breed projects such as the OpenDaylight controller. NoSQL graph-oriented databases like Neo4j or Titan, and Hadoop-like infrastructures.