QUALITY oF SERVlcE (QoS) OVER VPN NETWORKS

Abstract

Virtual Private Networks (VPNs) are private communications networks usually used by a single company, or by several different companies or organizations, to communicate over a public network. VPN message traffic is carried on public networking infrastructure (e.g. the Internet) using standard (often insecure) protocols, or over a service provider's network providing VPN service guarded by well defined Service Level Agreement (SLA) between the VPN customer and the VPN service provider. Informally, we could say that a VPN is a set of sites that can communicate with each other. More formally, a VPN is defined by a set of administrative policies that control both connectivity and Quality of Service (QoS) among sites. Two years ago, in the Kingdom of Saudi Arabia, most service providers offered IP connectivity for their customers through MPLS/VPN networks. This kind of service was built based on shared resources for two important reasons: reduce the cost for the customer and network scalability for the service provider. However, the provision of QoS has become a nightmare for network engineers. As a service provider, it is not that easy to solve this issue. In our thesis, we propose a methodology for the network provider to choose the QoS technique that fits his special network design and traffic statistics. The main goal of this thesis is to increase the efficiency of VPN networks for the transport of various types of multimedia traffic by tuning the QoS parameters. We have setup a methodology for VPN-SPs which will give a way to find the best QoS parameters based on VPN-SP’s network and traffic. To implement this methodology, we selected one of the largest VPN network in Saudi Arabia as a study case and determined the best QoS mechanism that should be be used. A simulation model for their network was built. Then based on real network traffic measurements, traffic models were built and used in the simulation. In our simulation model, six queuing mechanisms were considered: FIFO, FQ, PQ, two types of CQ, and LLQ. After the analysis of the obtained results, LLQ was selected as being the best queuing policy for VPN-SP. Furthermore, we studied the effects of changing some network parameters, such as the number of voice calls and the bandwidth of the last mile in the network, and studied their effects on each one of the previous scenarios.