Gilson Miranda

Computer networks are becoming increasingly complex, with several interconnected devices, protocols, control knobs, and applications sharing a physical infrastructure. Different applications can have different Quality of Service (QoS) requirements to operate correctly. The user-perceived experience, i.e., Quality of Experience (QoE), when consuming network applications is also a relevant indicator that must be accounted for, to minimize customer churn and offer a satisfactory experience to the users. Such complexity of the networks and applications makes it difficult for network operators to define a one-fits-all configuration or manually set device parameters tailored to each application. Thus, automating the configuration of network functions and enabling the allocation of the appropriate resources for each application is crucial to support the scaling of future networks. To realize this automation, several components are required for monitoring network performance, acting to control the behavior of network flows, and managing the available resources to accommodate the different applications sharing the infrastructure. Network slicing is one of the main technologies enabling such tailored configurations, but other supporting enablers must still be integrated and combined into an overarching solution. This thesis presents solutions tackling the different activities required for automating the management of network slices including real-time monitoring of their performances, estimating user experience when applicable, exerting control over network flows, and automatically adjusting resources in the face of issues. Therefore, we build a complete solution to realize network slicing using both QoS and QoE measurements as feedback signals, implemented in real devices with a strong experimental component. Each component presented in this thesis was experimentally evaluated using real testbeds including wired and wireless segments, highlighting the feasibility of our solutions.

Deterministic performance and reliable operation are vital for many applications with industrial-grade requirements. Such applications rely on Time-Sensitive Networking (TSN) to enable time-critical deterministic communication. While standardization efforts were focused mainly on TSN features for wired domains, recent advances in wireless technologies (e.g., Wi-Fi 6/7) are extending time-sensitive communication towards wireless networks as well. However, achieving multi-domain LAN/wlan end-to-end TSN communication requires addressing challenges on end-to-end time synchronization, multi-domain control plane interoperability, run-time end-to-end scheduling, and fine-grained monitoring. Because state-of-the-art TSN controllers’ scope lays far below these new required capabilities, in this work we present a novel, fully-programmable controller for end-to-end TSN-enabled networks. Our controller is based on a modular architecture to be adaptable to challenges arising when shifting the standard TSN scope towards WLAN domain. We deploy a proof-of-concept in a cloud-wired environment to evaluate its key performance indicators when handling increasing numbers of nodes and simultaneous requests. Further, we run experiments on real TSN-enabled networks comprising Ethernet and Wi-Fi technologies, demonstrating the effectiveness of the controller in performing seamless fine-grained traffic control in both domains.

With the increasing traffic of Video on Demand (VoD), network providers are seeking to deliver high Quality of Experience (QoE) for their users. Many methods have been proposed to assess VoD-related QoE. Some of them rely on client instrumentation and reporting QoE information to network elements, such as Server and Network Assisted DASH, others are based on statistical methods that make QoE inferences using monitored network conditions, such as throughput and delays. In this article, we present a practical method to estimate QoE for VoD using the widely supported Internet Control Message Protocol (ICMP) probes. Measured network conditions are used as input to a Machine Learning (ML) model that estimates QoE in terms of Mean Opinion Score (MOS), based on the ITU-T P.1203 Recommendation. The estimation encompasses video quality switches and playback stalls. We estimate MOS with an average Root Mean Square Error (RMSE) of 1.05 for a catalog of 25 different videos, training a model with sessions of the shortest video, and evaluating the generalization to the full catalog. We performed experiments using a virtualized setup as well as in a Wide Area Network.