Improving Transport Protocols for better flexibility
PhD Thesis · 2024
Abstract
Transport protocols are an integral part of the Internet infrastructure. They are used to realise the many Internet services and applications that exist today. Yet, the Internet infrastructure evolves at a slow pace, with care for its continued operating nature and constraints stemming from its initial assumptions dating back to the 1980s. More recently, the QUIC transport protocol has been standardised and adopted by many large companies. While QUIC is now a competitor to the TCP/TLS stack, pressuring it to evolve further, it still shares some limiting assumptions. This thesis makes transport protocols more flexible with five contributions. First, we study the QUIC protocol by testing and analysing its first implementations during its standardisation phase. Then, we revisit the strict layering assumption between the application and QUIC to enable applications to tune the QUIC protocol to their need with secure and portable executable code. Drawing from our experience in the QUIC protocol, we propose a new version of TCPLS, a tightly-integrated protocol enabling modern transport services for TCP and TLS. The thesis also considers the impact of selecting IPv4 or IPv6 for latency-sensitive applications when reaching services over the Internet and proposes an adaptive selection technique. Finally, the thesis concludes by reconsidering the end-to-end principle and bringing more flexibility to the QUIC protocol from this perspective. All contributions were implemented, tested and analysed over simulated, emulated or real networks with several Internet applications.
People
Cite (BibTeX)
@phdthesis{piraux2024improving,
title={Improving Transport Protocols for better flexibility},
author={Piraux, Maxime},
year={2024},
school={UCLouvain}
}