Exploring Proxying QUIC and HTTP/3 for Satellite Communication

TL;DR

This study compares QUIC and HTTP/3 performance over satellite links using a testbed, revealing that PEPs significantly enhance web performance, especially for HTTP/3, suggesting potential benefits of deploying PEPs with QUIC.

Contribution

The paper introduces a satellite emulation testbed to evaluate QUIC and HTTP/3 with and without PEPs, providing new insights into their performance in satellite communication environments.

Findings

QUIC PEP reaches slow start threshold up to 2s faster than non-PEP.

HTTP/3-PEP outperforms HTTP/1.1-PEP in multiple scenarios.

HTTP/3-PEP improves Page Load Time by over 7 seconds in edge cases.

Abstract

Low-Earth Orbit satellites have gained momentum to provide Internet connectivity, augmenting those in the long-established geostationary orbits. At the same time, QUIC has been developed as the new transport protocol for the web. While QUIC traffic is fully encrypted, intermediaries such as performance enhancing proxies (PEPs) - in the past essential for Internet over satellite performance - can no longer tamper with and optimize transport connections. In this paper, we present a satellite emulation testbed and use it to compare QUIC and TCP as well as HTTP/3 and HTTP/1.1 with and without minimal PEP functionality. Evaluating goodput over time, we find that the slow start threshold is reached up to 2s faster for QUIC PEP in comparison to QUIC Non-PEP. Moreover, we find that HTTP/3 and HTTP/3-PEP outperform HTTP/1.1 and HTTP/1.1-PEP in multiple web performance scenarios, where HTTP/3-PEP improves over HTTP/3 for Page Load Time by over 7s in edge cases. Hence, our findings hint that these performance gains may warrant exploring PEPs for QUIC.