QUIC is not Quick Enough over Fast Internet (2310.09423v2)

Abstract: QUIC is expected to be a game-changer in improving web application performance. In this paper, we conduct a systematic examination of QUIC's performance over high-speed networks. We find that over fast Internet, the UDP+QUIC+HTTP/3 stack suffers a data rate reduction of up to 45.2% compared to the TCP+TLS+HTTP/2 counterpart. Moreover, the performance gap between QUIC and HTTP/2 grows as the underlying bandwidth increases. We observe this issue on lightweight data transfer clients and major web browsers (Chrome, Edge, Firefox, Opera), on different hosts (desktop, mobile), and over diverse networks (wired broadband, cellular). It affects not only file transfers, but also various applications such as video streaming (up to 9.8% video bitrate reduction) and web browsing. Through rigorous packet trace analysis and kernel- and user-space profiling, we identify the root cause to be high receiver-side processing overhead, in particular, excessive data packets and QUIC's user-space ACKs. We make concrete recommendations for mitigating the observed performance issues.

• The paper presents a detailed performance analysis showing up to a 45.2% data rate reduction for QUIC compared to HTTP/2.
• It attributes QUIC's underperformance to high CPU usage from receiver-side packet processing and user-space ACK handling.
• Recommendations include adopting UDP GRO, delayed ACKs, and multi-threading to optimize QUIC performance.

QUIC is not Quick Enough over Fast Internet: An Analysis

This paper presents a comprehensive evaluation of the QUIC protocol's performance over high-speed networks, highlighting its limitations when compared to its TCP+TLS+HTTP/2 counterpart. The authors conduct a meticulous examination using a range of clients, network conditions, and platforms to identify the causes of performance degradation, particularly emphasizing the receiver-side processing overhead.

Key Findings

The paper reveals that in high-speed network environments, the QUIC protocol underperforms relative to HTTP/2. The authors observe a significant data rate reduction, up to 45.2%, when using QUIC over fast Internet connections. This performance gap becomes more pronounced as bandwidth increases, affecting not only file transfers but also video streaming and web browsing activities. The authors attribute this discrepancy to the high receiver-side processing overhead caused by excessive data packet handling and QUIC’s implementation of user-space acknowledgments (ACKs).

Performance Analysis

The performance analysis involved a variety of experimental setups, utilizing different browsers such as Chrome, Edge, Firefox, and Opera across platforms like desktops and mobile devices. The degradation in performance was consistent across the board, with more pronounced effects on less powerful clients such as smartphones. Notably, CPU usage was substantially higher for QUIC than for HTTP/2 during packet processing, further indicating an inefficient handling mechanism in high-bandwidth scenarios.

Application Impact

The implications of QUIC’s limitations extend to video streaming and web browsing. In high-speed environments, QUIC results in up to a 9.8% reduction in video bitrate. This quality of experience (QoE) degradation is accentuated in dynamic network conditions, such as 5G, which are becoming increasingly prevalent. Furthermore, QUIC increased page load times by 3.0% on average, introducing further variability in web-based applications.

Root Cause and Recommendations

The authors identify the root cause of the performance gap as the receiver-side inefficiencies tied to QUIC’s design. The analysis highlights two main issues: the in-kernel UDP stack manages a higher number of packet reads compared to TCP, and the user-space handling of QUIC ACKs incurs additional processing overhead. These factors contribute to increased CPU usage and elongated packet processing times.

To mitigate these challenges, the authors propose several technical recommendations:

• Deploying UDP Generic Receive Offload (GRO) to manage packet handling more efficiently.
• Optimizing receiver logic in QUIC, potentially through delayed ACK implementations and enhanced packet handling.
• Utilizing multi-threading techniques to distribute the processing load across multiple CPU cores.

Implications and Future Work

While QUIC holds promise for advancing Internet protocols, its current implementation poses challenges in high-speed scenarios. This paper highlights the importance of refining protocol architectures to better accommodate advances in network capabilities. The authors' findings suggest a need for continued collaborative efforts in developing and optimizing QUIC and similar protocols.

Given the rapidly evolving nature of Internet technologies, this work underscores the critical need for ongoing research into protocol efficiency, especially in the context of emerging network environments. As the Internet ecosystem transitions towards HTTP/3, it becomes imperative to address these performance bottlenecks to fully harness the potential of high-speed connectivity.

In conclusion, while QUIC offers theoretical benefits over traditional protocols, its practical deployment over fast Internet runs into considerable challenges. The paper's insights contribute to a deeper understanding of QUIC’s architectural and operational limitations, paving the way for improvements in protocol design and implementation for future network infrastructures.