Performance Comparison of HTTP/3 and HTTP/2: Proxy vs. Non-Proxy Environments

📅 2024-09-24
🏛️ arXiv.org
📈 Citations: 0
Influential: 0
📄 PDF
🤖 AI Summary
This work systematically evaluates performance differences between QUIC/HTTP/3 and TCP/HTTP/2 under proxy-enhanced and high-loss, high-latency network conditions, focusing on the impact of proxy deployment and connection migration. Method: We construct a comparative experimental platform based on the QUIC protocol stack and TCP with BBR congestion control, incorporating realistic network impairments—including packet loss rates up to 10% and RTTs up to 500 ms—and conduct end-to-end latency and throughput measurements. Contribution/Results: Our study provides the first quantitative evidence that proxies improve HTTP/2 download performance by up to 90%, whereas their effect on HTTP/3 is negligible. Conversely, HTTP/3 achieves up to 88.36% higher throughput than HTTP/2 under high-loss, high-latency conditions, retaining an 81.5% advantage even under extreme packet loss. These results demonstrate HTTP/3’s superior network robustness—its performance gains are intrinsic and do not rely on proxy optimization or specific congestion control mechanisms—thereby offering empirical support for next-generation internet protocol deployment.

Technology Category

Application Category

📝 Abstract
This paper systematically evaluates the performance of QUIC/HTTP3 (H3) and TCP/HTTP2 (H2) protocols in proxy-enhanced environments. H3 leverages features such as UDP-based flow-controlled streams, integrated TLS, multiplexed connections, and connection migration, offering the potential for improved web communication. Despite extensive research, the impact of proxy integration and connection migration remains underexplored. This study addresses this gap by evaluating H3 and H2 across various scenarios, particularly in noisy networks and proxy setups. Our findings show that H3 excels under high loss and high latency conditions, significantly benefiting from its connection migration and multiplexing features, with improvements of up to 88.36% under high-loss and high-latency conditions, and 81.5% under extreme loss conditions, respectively. H3's connection migration remains robust, maintaining stable performance even in proxy-enhanced environments, ensuring seamless network transitions. The proxy has a more neutral impact on H3, while it significantly enhances H2 performance, particularly when paired with BBR, resulting in a 90% improvement in the single-stream file download experiment under severe network impairments. Any improvements observed in H3 under a proxy are minor and do not fundamentally alter H3's performance as they do for H2. Importantly, while H2 with the right congestion control algorithm (CCA) can achieve performance comparable to H3, H3's performance is more robust, as it is less impacted by network conditions, proxy settings, and CCA variations.
Problem

Research questions and friction points this paper is trying to address.

Compare HTTP/3 and HTTP/2 performance with proxy integration
Evaluate impact of proxies on H2 and H3 in lossy networks
Assess H3 robustness in high-loss and high-latency conditions
Innovation

Methods, ideas, or system contributions that make the work stand out.

UDP-based flow-controlled streams with built-in TLS
Connection migration and multiplexing for robustness
Proxy-enhanced performance with BBR congestion control
🔎 Similar Papers
No similar papers found.