Reinforcement Learning for Effective Service Function Chain Provision in Overlay Networks

PROJECT TITLE :

Efficient Provision of Service Function Chains in Overlay Networks using Reinforcement Learning

ABSTRACT:

The technologies of Software-Defined NetWorking (SDN) and Network Functions Virtualization (NFV) make it easier to deploy Service Function Chains (SFCs) in clouds with greater flexibility and efficiency. However, it is still difficult to chain Virtualized Network Functions (VNFs) in overlay networks efficiently without having knowledge of the configurations of the underlying networks. Although there are many different deterministic methods for the placement and chaining of VNFs, these methods are notoriously difficult to implement and require detailed knowledge of the substrate networks. Fortunately, Reinforcement Learning (RL) presents opportunities to alleviate this challenge because it can learn to make appropriate decisions without prior knowledge. This opens up a world of possibilities. As a result, we propose in this article an RL approach for efficient SFC provision in overlay networks, which are networks in which the same VNFs are provided by multiple vendors, each of which has a different level of performance. To be more specific, we begin by transforming the problem into a model that uses integer linear programming (ILP) for benchmarking purposes. Then, we propose a policy-gradient-based solution that corresponds to the online SFC path selection that we present as a Markov Decision Process (MDP). In the end, we evaluate our proposed method by running extensive simulations using randomly generated SFC requests and a dataset based on real-world video streaming. Additionally, we implement an emulation system for the purpose of verifying the system's feasibility. The results that are related show that the performance of our approach is comparable to that of the ILP-based method, but it is superior to the performance of the deep Q-learning, random, and load-least-greedy methods.

Did you like this research project?

To get this research project Guidelines, Training and Code... Click Here