Performance Comparison of HTTP/3 and HTTP/2 with Proxy Integration

TL;DR

This study compares HTTP/3 and HTTP/2 performance in proxy environments, revealing that proxies significantly boost HTTP/2 performance while HTTP/3 remains consistently robust across various network conditions.

Contribution

It provides a comprehensive evaluation of HTTP/3 and HTTP/2 with proxies, highlighting the robustness of HTTP/3's internal mechanisms in diverse network scenarios.

Findings

Proxies improve HTTP/2 download speeds by up to 90% under severe impairments.

HTTP/3 maintains stable performance regardless of proxies or congestion control.

HTTP/3 outperforms HTTP/2 in high-loss and high-latency environments.

Abstract

This paper systematically evaluates the performance of QUIC/HTTP3 (H3) and TCP/HTTP2 (H2) in proxy-enhanced environments. H3 integrates UDP-based flow-controlled streams, built-in TLS, multiplexing, and connection migration to better support modern web communication. While prior studies show that H3 can outperform or underperform H2 depending on network conditions, the role of proxies and connection migration remains underexplored. We assess a variety of H2 and H3 client implementations across, particularly in lossy networks and proxy setups. Our findings show that proxies can significantly enhance H2 performance, yielding a 90% improvement in single-stream downloads under severe impairments when used with the BBR congestion control algorithm. In contrast, proxies have minimal impact on H3, which maintains consistent performance due to its internal mechanisms. H3 excels under high-loss and high-latency conditions, leveraging connection migration and multiplexing to deliver up to 88.36% improvement in migration scenarios and 81.5% in extreme loss cases. While optimized H2 can match H3 in some settings, H3 is more robust overall, showing less sensitivity to proxies, impairments, and congestion control variations.