A Web-Based Platform for Experimenting with Virtual Networks Adaptations via Reinforcement Learning over the GENI Testbed

Network emulators and simulation environments traditionally support computer networking and distributed system research. The continued use of multiple approaches highlights both the value and inadequacy of each approach. To this end, several large-scale virtual network testbeds such as GENI have emerged, allowing testing of a networked system in controlled yet realistic environments. Nevertheless, setting up those experiments first, and integrate machine learning models later to these deployments is a challenge. In this thesis, we propose the design and implementation of a web and command line tool that integrate Reinforcement Learning (RL) with a virtual network experiment using resources acquired within the GENI testbed. After some configuration setup, users draw the network topology of their experiment and then reserve the GENI resources with a button push. A reinforcement learning algorithm is then launched to learn and steer traffic dynamically based on emulated traffic network conditions. The back end includes a Software-Defined Network (SDN) controller that reprograms the routing table of Open Virtual Switches after reading the output of the RL model. While in this thesis the system focuses on the deployment of experiments for virtual network adaptation, the platform can be easily extended to other network management mechanisms and machine learning algorithms.