// the find
apache/horaedb
Apache HoraeDB (incubating) is a high-performance, distributed, cloud native time-series database.
Apache HoraeDB is a time-series database written in Rust, targeting Prometheus remote-write workloads and IoT ingestion at scale. It's currently mid-rewrite — the main branch is a new engine under active development, the stable engine lives on a separate branch. Not production-ready in the current state.
Prometheus remote-write support is a first-class design goal, not an afterthought — the remote_write crate has its own pooled parser with benchmarks showing real attention to ingestion throughput. Columnar storage with compaction scheduling is properly separated from the query path, which is the right architecture for a TSDB. Apache incubation means governance, IP hygiene, and a real release process — better than most hobby TSDBs at this star count. Rust throughout means memory safety without a GC pause tax, which matters for latency-sensitive time-series workloads.
The main branch is explicitly unstable and the README says it won't be publicly ready until 'end of 2025' — that date has passed and there's no stable release on main yet, so you're choosing between an unstable new engine or a legacy branch that isn't being actively developed. Documentation is thin: the docs/ folder is mostly RFCs and benchmark screenshots, not operational guides for running a cluster. The metric_engine module has almost no implementation visible — data, index, and metric subdirectories each contain a single mod.rs, suggesting the core engine is skeletal. No query language is obvious from the tree; if you need anything beyond Prometheus PromQL-style reads, it's unclear what you get.