Welcome to TRustDB (Time series Rust DataBase), a database optimized for handling high-cardinality time series data. The technical report can be found here.
By Desmond Cheong & Nick Young
- Populate a
.envfile with the path of the data folder. We've suggested a default in thesample.envfile.
- To support efficient queries, an inverted index is constructed that maps from label key-value pairs and metric names to series.
- To reduce the storage footprint of this metadata, the inverted index is compressed into a Finite State Transducer (FST).
- To speed up query evaluation, queries are rewritten and evaluated in Disjunctive Normal Form (DNF).
More details can be found in the technical report.
This database is far from complete and we have a lot more ideas to experiment with. Some of the things on our mind can be found on the issues page.
Contributions are welcome!