Aspects of Quality : Using Quality Measurements to Improve Computer Network Performance

Sammanfattning: As computer networks grow in complexity, measuring the performance becomes a challenge. Intrinsic properties of the networks capability to transport data, such as bandwidth, latency, loss, and error rate are not always enough to provide a clear picture of how well the network can satisfy the users' expectations. In these situations, the relationship between the users' expectations and the quality provided by the network is of interest. Today multiple approaches to quantifying quality exists, such as Quality of Service (QoS), Quality of Experience (QoE) and Quality of Information (QoI). In this thesis, we explore how such quality measurements can be used as means to improve network performance. In the first part, we examine the Quality of Service in shared experiment networks, with a focus on the FEDERICA network. We present a method for statistical analysis of metadata, based on clustering. We show that, using this method, it is possible to improve the reliability of experiments in shared experiment networks. In the second part, we take a Quality of Experience viewpoint, while doing experimental development. The goal is to improve the mobility performance of an implementation of the Stream Control Transfer Protocol (SCTP) on mobile devices. We present optimisations of the SCTP implementation along with a mobility framework, which simplifies the introduction of mobility functionality in an existing SCTP application. We show that by using this combination, it is possible to perform seamless vertical handover between WiFi and 3G cellular networks. In the third part, we focus on Quality of Information in Delay Tolerant Wireless Sensor Networks (DT-WSN). We study how it is possible to improve the quality of the measurements obtained under conditions where bandwidth and storage capacity are limited, forcing the network to discard a significant fraction of the data. We introduce the SmartGap algorithm, a buffer management algorithm for DT-WSNs, and demonstrate that this algorithm can provide significantly improved QoI over a wide range of network configurations.