Knowledge Defined Networking

Network management is the original research focus area of Walton Institute. Core telecoms and data centre networks have critical requirements that need to be met. Given ongoing cost and capacity constraints, and growing complexity, network management is still an active research topic. Recent developments in softwarisation – especially SDN and NFV – enable techniques like network slicing that make more efficient use of available resources. However, software-defined infrastructures require more capable management systems to reap the benefits. Knowledge of how the network will be used is essential for predictive management using AI and hence better outcomes.

Key areas of expertise:

• Specification of management objectives for networks and services is an essential first step in any management system. Such objectives are specified as intents, e.g., ensuring connectivity between nodes in separate subnets, prioritising desirable data flows, or enforcing service-level requirements. Intents need to be transformed to deployable configuration settings in the network, hence Intent-Based Networking.

• Scaling Virtual Function Chains in response to demand is a challenge for network operators. Virtual Network Functions (VNFs) are software implementations of network functions like Layer-2 switches, firewalls, etc., deployed on commodity hardware.  VNFs need to scale outwards to meet service level requirements. Graph Neural Network models show promise here.

• Resource management of Cloud Radio Access Networks is a critical requirement for mobile network operators. Mobile devices connect wirelessly to the nearest cell and the signal is routed through the operator’s network, terminating in a Baseband Unit (BBU). Increasingly, BBUs are virtualised and implemented in the cloud. Demand prediction using machine learning helps to make better scaling decisions.