Towards a Smart, Adaptable, Dynamic Networking Design using Virtual Slicing in Software Defined Networks


Steven Schmitt, M.Sc

Caption text goes here

The importance of reliable and adaptable networks has become increasingly relevant with the escalation of connectivity in our lives. The growth of streaming of entertainment and development of always online softw- are has created an environment of large data flows that need to be handled efficiently. Historically this problem has been solved with hardware-based load balancers. Although these hardware-based load balancers provide a solution, they often times are expensive and lack flex- ibility and scalability. With the use of Software-Defined Networking (SDN), a more dynamic solution can be created to meet network load balancing needs. We in this work propose a smart, dynamic and adaptable net- working design seeking to utilize network resources more efficiently by identifying traffic patterns and ana- lyzing network metric to dynamically build virtual slices. Our results show that with this approach, we were able to solve the issues with current load balancing techniques by minimizing packet loss, maximizing network link utilization, and efficiently reduce the load on the controller.



Research Dialogue 2018 


The emergence of Software-Defined Networking (SDN) has brought along a wave of new technologies and dev- elopments in the field of networking with hopes of dealing with network resources more efficiently. SDN allows for both flexibility and adaptability by separating the control and data planes in a network environment by virtualizing network hardware.

Through the programmability features of SDN, handling data transfer between hosts become more efficient. The idea is to connect hosts through the use of virtual switches and virtual controllers that can be used to aut- omate network functions programmatically. Our resea- rch seeks to develop an SDN system that uses metric- based analysis to virtually and automatically slice our network. These virtual slices will group hosts based on prior network activity allowing for dynamic load balan- cing. By dynamically slicing the network based on metrics such as link utilization and packet drop rate we can achieve an adaptable network design that outper- forms traditional load-balancing techniques.