Loading page...Загружаем страницу...正在加载页面...
AtlasRepo

How we saved 100 terabytes of memory by optimizing 1.1.1.1’s DNS cache

说明与用途

通过优化 1.1.1.1 DNS 缓存,将内存使用量从 100 TB 降至 1 TB,提升了可扩展性并降低了成本。

问题

The 1.1.1.1 DNS cache consumed 100 TB of memory, limiting scalability and increasing operational costs.

结果

Reduced memory consumption from 100 TB to 1 TB, enabling higher query throughput and lower infrastructure costs.

工作流

  • 1. Profileed cache memory usage and identified header/payload overhead.
  • 2. Designed a new record format with 16‑byte header and 4‑byte payload.
  • 3. Implemented the format in the cache codebase and updated serialization logic.
  • 4. Deployed changes to a subset of servers and monitored memory usage.
  • 5. Rolled out to all 1.1.1.1 servers after confirming stability.

证据

We saved 100 terabytes of memory by optimizing 1.1.1.1’s DNS cache

来源

  • 记录日期: 2026-09-07T00:22:03.477Z