Data Centric / Scheduled Networks
The current network ecosystem is passing through a very large paradigm shift from protocol enhancements towards network behaviour optimization. This was triggered by:
- getting closer to the end of the available spectrum – without having cost-effective spectrum where to deploy new networks, the current allocated spectrum will have to be scheduled in a subscriber and network aware manner to fit the usage requirements (see RAN-core convergence).
- the decoupling between the software and the hardware and the possibility to configure dynamically the network topology and the network functions brought a new level of flexibility in the optimization of the systems (see Organic Networks).
- the current protocol level optimizations are rather minimal, a shift towards easy programmability and reachability of more developers being considered more important.
- with the development of virtual networks and cloud deployments a more holistic approach towards networks was advocated where complete networks are configured and deployed based on centrally defined configurations (see Organic Networks).
To progress beyond this level, network behavior optimization is required where the scheduling of different network resources (infrastructure, functionality, etc.) according to the momentary system needs. Using this data, new algorithms are defined which account for the specificity and on the same time build new handling processes:
- Network performance – providing optimized services within the available resources
- Service quality – providing the best service possible to the users within the momentary conditions
- Reliability – providing seamless service in face of perturbations and failures
- Security – privacy and system protection against attacks
- Energy efficiency – reduced energy consumption while maintaining the same network quality
This can be achieved only by having a holistic perspective of the system, knowing the impact of any scheduling on the rest of the system and making coordinated, conflict-free decisions. For this, we implement aggregation control entities for topology control and user distribution making decisions based on the algorithmic indications to increase the overall momentary network utility under specific SLA constraints.