Four Dimensions of HTTP Performance Optimization

When a product’s user base grows rapidly, the need to optimize the HTTP protocol becomes inevitable. This article, originally presented by TVP expert Tao Hui at the 2019 GOPS Global Operations Conference in Shanghai, summarizes four new dimensions for extreme HTTP performance optimization.

Author Bio: Tao Hui is a Tencent Cloud Valuable Professional (TVP), CTO and co‑founder of Hangzhou Zhilianda Data Co., Ltd., and former engineer at Alibaba Cloud, Tencent, Huawei, and Cisco. He authored the bestseller Deep Understanding of Nginx: Module Development and Architecture Analysis and co‑created popular courses on Web protocols and Nginx.

1. Encoding Efficiency Optimization – Reduce the size of the transmitted character stream. HTTP/1.1 uses whitespace‑delimited encoding, which is inefficient. Binary protocols such as SPDY (the predecessor of HTTP/2) and later HTTP/2 improve encoding efficiency. Compression algorithms (gzip, brotli) are compared, showing brotli’s superior compression ratio and speed. Header compression via HPACK (Huffman coding, static and dynamic tables) is also explained.

2. Channel Utilization Optimization – Improve the utilization of the underlying transport channel. This includes multiplexing (multiple HTTP/2 streams over a single TCP connection), fast error detection and recovery (TCP retransmission, SACK), and fair resource allocation (leaky bucket, HTTP/2 stream priority). The discussion also covers congestion control evolution from CUBIC to Google’s BBR and the impact of queue buildup on latency.

3. Transmission Path Optimization – Optimize the end‑to‑end delivery path. Topics cover caching (including stale‑cache validation with 304 responses), TCP slow‑start tuning (initial congestion window adjustments), and the shift from pull‑mode to push‑mode delivery (parallel fetching of resources in HTTP/2). The limitations of TCP head‑of‑line blocking and the move toward UDP‑based protocols like GQUIC/HTTP/3 are highlighted.

4. Information Security Optimization – Strengthen security while improving performance. The evolution from SSL 3.0 to TLS 1.3 is described, noting the removal of insecure key‑exchange algorithms, reduced handshake round‑trips, and the performance benefits of modern cipher suites.

Conclusion – By addressing encoding efficiency, channel utilization, transmission path, and security, developers can build a comprehensive knowledge tree that covers the majority of HTTP optimization techniques, leading to lower latency, higher concurrency, and a better user experience.