Wirewiki网站开发者公开了其域名自动完成功能的技术方案，通过优化架构和客户端预取策略，将P99延迟目标控制在121毫秒内1。该系统支持240万个域名的查询1，大多数API请求在2毫秒内完成响应，在1.6k req/s负载下Nginx与API能在15毫秒内处理99%的请求1。
实现这一性能目标的核心策略包括充分利用键盘按键间隔时间1。开发者在keyDown事件时触发预取建议，随后在keyUp事件时渲染结果，由于最短按键间隔可达8.33毫秒（基于60Hz显示器），查询结果在用户释放按键前即可准备就绪1。系统架构采用字符前缀树存储热门域名（占用内存），辅以内存映射块索引支撑其他gTLD域名查询1，总数据占用约2.5GB磁盘空间1。数据源来自Tranco前100万热门域名和CZDS的其他gTLD域名1。
受限于目前仅在欧洲单一服务器部署，远程用户的网络延迟会显著影响体验1。以美国用户为例，网络延迟会额外增加100-200毫秒1。
A Wirewiki developer has shared technical details of an autocomplete system achieving P99 latency across 240 million domain names 1. The implementation leverages client-side prefetching, trie data structures, and memory-mapped block indexing to deliver search suggestions with minimal delay.
The architecture combines two complementary storage layers: a character trie stored in memory handles the most common prefixes, while a 27 MB indexed directory backed by SSD provides fallback lookups for less frequent queries 1. The system draws its domain dataset from the top one million entries in the Tranco rankings plus additional generic top-level domains from CZDS, consuming approximately 2.5 GB of disk space 1. Most API requests complete within 2 milliseconds, with 99 percent of queries resolved in 15 milliseconds under a load of 1.6 thousand requests per second 1.
The optimization strategy exploits the natural timing gap between keyboard presses. Suggestions are prefetched on key-down events—leveraging the minimum 8.33 millisecond interval between keypresses on 60 Hz displays—and rendered on key-up, allowing results to be ready before users release their keys 1. This timing-aware approach targets a P99 latency of 121 milliseconds based on measured typing patterns 1. However, the single European server deployment introduces geographic overhead; users in North America experience an additional 100 to 200 milliseconds of latency, pushing remote response times beyond the original target 1.
评论
还没有评论，欢迎留下第一条。