Network Monitoring with Software Defined Networking

Abstract

Network monitoring is becoming more and more important as more Internet Service Providers and Enterprise networks deploy real-time services, like voice and video. Network operators need to have an up-to-date view of the network and to measure network performance with metrics like link usage, packet loss and delay, in order to assure the quality of service for such applications. Obtaining accurate and meaningful network statistics helps the service providers to estimate the "health" of their network, to find service degradation due to congestion and could even use them for routing optimization. Finally, a more accurate picture of the nature of the Internet traffic is important for continued research and innovation. In this thesis, Software Defined Networking is used as a unified solution to measure link utilization, packet loss and delay. Currently, there is no single solution capable to measure all the mentioned metrics, but a collection of multiple different methods. They all need separate infrastructure which needs additional installations and expenses. Furthermore, those methods are not capable to meet all the requirements for a monitoring solution, some are not accurate or granular enough, others are adding additional network load or lack scalability. Software Defined Networking is still in an early development stage. Unfortunately, obtaining network statistics is a problem that has not been discussed very much. Nevertheless, OpenFlow protocol is already gaining a lot of popularity and it could be used as a unified monitoring solution. Its built-in features are already sufficient to provide accurate measurements. This thesis explores what are the current monitoring methods with and without Software Defined Networking and how monitoring could be archived in an OpenFlow enabled topology. It proposes methods to measure link usage, packet loss and delay. In terms of link usage statistics several approaches are proposed to reduce the generated overhead. Furthermore, the thesis also suggests how to measure how packet loss in networks that use OpenFlow.