Wirewiki achieves '0ms' perceived autocomplete for 240 million domains
By combining in-memory tries with SSD-backed memory-mapped indexes, Ruurt Jan has eliminated perceived latency for domain searches.
Ruurt Jan, the creator of Wirewiki.com, has implemented an autocomplete system for 240 million domain names that achieves a perceived latency of 0ms at the 99th percentile. The system is engineered to ensure that search results are ready for rendering before a user releases a key during typing.
To achieve this, Jan utilized a hybrid search architecture that splits data into a "head" and a "tail." The top 1 million domains from the Tranco list are stored in an in-memory character trie for immediate access. The remaining 240 million domains from the Central Zone Data Service (CZDS) are stored in an SSD-backed memory-mapped block index. To keep the footprint manageable, these 240 million domain names are delta-compressed into fixed-size blocks, requiring approximately 2.5 GB of total disk space.
The Engineering of 'Instant'
The "0ms" claim is a perceived metric rather than a literal measurement of server processing. Jan defines this as results being ready before the `keyUp` event. Based on his measurements of typing speed, this provides a time budget of approximately 121ms for the 99th percentile of users.
On the backend, the API is optimized for extreme efficiency. Most requests are answered within 2ms, and the p99 latency remains around 15ms—including Nginx overhead—even when handling 1.6k requests per second. This high-performance ceiling allows the system to stay well within the 121ms window for users located in Europe.
Why Performance Matters
Wirewiki serves as a tool for inspecting internet infrastructure, including DNS records and delegation. In a competitive landscape, Jan focused on extreme UX quality to differentiate the service, specifically targeting the speed and completeness of autocomplete navigation.
The project serves as a practical demonstration of how high-performance data structures, such as Tries and memory-mapped files, can eliminate the friction of web interfaces. However, the implementation also highlights that network round-trip time (RTT) remains the primary bottleneck. While the API is nearly instantaneous, geographic distance—such as a request traveling from the USA to Europe—can still push latency beyond the threshold of perceived instantaneity.
Future Outlook
While the technical achievement is significant, Jan believes the project is too niche to form the basis of a standalone business. However, he has indicated an openness to providing API access to others if there is paying interest in the technology. For now, the system remains a core component of the Wirewiki user experience, proving that massive datasets can feel weightless through strategic indexing and a deep understanding of human typing patterns.