Network Slicing and Quality of Service in Programmable Data Planes

5G has capabilities which exceed what has been seen for the previous generations of mobile networks: it can handle very heterogeneous traffic, each one with its own requirements in terms of resources needed for normal operativity. There is the need to support Ultra-Reliable Low-Latency Communications along with Enhanced Mobile Broadband and Massive Machine-Type Communications traffic, and every flow is competing for resources with each other. Therefore, solutions to handle such strict requirements emerged, and got standardized in what is the Fifth Generation (5G) of mobile networks: in particular, Network Slicing enables the scalability needed by modern communications to operate and allows for efficient management of network resources, isolating them according to various strategies. Many implementations of it are available for the 5G Core and the Radio Access Network, but there is not a widespread solution for the Transport layer, which is a crucial part of the whole infrastructure. Thus, this thesis proposes an implementation of a Network Slicing solution in the Transport layer with Quality of Service delivery, using the Programming Protocol-Independent Packet Processors (P4) language. This latter tool promises very high performances, both in terms of throughput and latency, as it provides the possibility to deploy custom network configurations and behaviors both in software and in hardware, specialized, targets. The framework is then deployed in a real testbed and evaluated, also against more widespread and properly configured alternatives, to assess the performance of our solution. An orchestration module for the Control Plane will be then proposed, to allow for a seamless and unified management of multiple, different, P4-enabled devices, each one with its own capabilities and exposed metadata. This module will lay the foundations for a real-world deployment, as industries will be able to gradually introduce P4-enabled devices in their physical infrastructure.