Time series databases are the backbone of metrics gathering in large computer systems. The most popular service is a difficult to understand and suboptimal implementation done in Python. Rust's strong typing, safe data reuse, standardized build tool, and easy integration with performance tools have made it possible to build a high performance replacement.
This talk will cover the rearchitecture of the popular graphite metric system redone entirely in Rust and currently used in production. This is not a line-for-line rewrite but it is compatible with the wire protocol and disk format.