Quality-of-Experience for Multimedia

Chapter 4

User-Driven Routing Algorithm Application for CDN Flow

In this chapter, we will present a new QoE-based routing algorithm, called QoE QLearning-based Adaptive Routing, which is based on a bio-inspired mechanism and uses the Q-learning approach, an algorithm of reinforcement learning. First, this chapter will briefly present the knowledge base of the Reinforcement Learning and Q-routing. Subsequently, the chapter will focus on the routing algorithm proposal.

4.1. Introduction

Over the past few decades Quality of Service (QoS) played an important role in the QoS framework of the Internet. The QoS routing strategy determines the suitable paths for different types of data traffic sent by various applications. The goal is to maximize the utilization of network resources while satisfying the traffic requirements in the network. To achieve this objective, it is necessary to develop a routing algorithm that determines multi-constrained paths taking into account the network state and the traffic needs (e.g. delay, loss rate and bandwidth). In fact, the core of any QoS routing strategy is the path computation algorithm. The algorithm has to select several alternative paths, which can satisfy a set of constraints like end-to-end delay and bandwidth requirements. However, the algorithms to solve such a problem are generally met with high computational complexity. In fact, Wang et al. [WAN 96] proved that an end-to-end QoS model with more than two non-correlated criteria is NP-completeness. ...