Reliance on distributed systems and virtualization technologies by many data processing platforms is on the increase within networking infrastructure.

Virtualization is frequently being evaluated as a replacement for the dedicated network systems associated with process-intensive functions. Software Defined Networking architectures appear to offer a more effective way to manage network connectivity for services in that they dynamically adapt connectivity to ever-changing requirements.

However, these new technologies pose several challenges. Systems and control researchers and their colleagues in Nokia Bell Labs are addressing these challenges, including more effective ways for managing network and computing resources, orchestrating distributed software components, improving the scalability and reliability of distributed systems and the management of distributed state.

Lab configuration to evaluate optimization performance of ALTO-based topology abstraction and network management of video services
Lab configuration to evaluate optimization performance of ALTO-based topology abstraction and network management of video services.

Fang Hao, T.V. Lakshman and Sarit Mukherjee have been designing novel ways to manage distributed state across controllers in Software Defined Networks that meet scalability and robustness needs. That same team, along with Murali Kodialam, has been exploring multi-constraint algorithms for locating components in a distributed system in order to meet performance requirements and optimize networking and compute resource utilization.

Exposing network state and topology is essential to the orchestration of services and networks. Network abstraction algorithms developed by Michael Scharf, Thomas Voith, Manuel Stein and Volker Hilt combine graph and policy language processing to generate simple but effective network views for applications. Several Bell Labs researchers have been active in the Internet Engineering Task Force on the Application Layer Traffic Optimization (ALTO) protocol for providing a network/application interface for exchanging topology and status information.

Current virtualization methods introduce timing performance issues that render the resulting systems unusable for latency-sensitive real-time applications. Martin Havemann is exploring alternative approaches for virtualization that will enable its use in one such application related to the air link in 4G-LTE systems.