BTrDB vs FiloDB

BTrDB is a free, MIT-licensed, self-hosted time-series database for dense, high-precision univariate data. It partitions data by time, supports clustered deployments, and records stream changes as new versions that queries can target. Streams have unique UUIDs and can be grouped into collections with tags and annotations. The Python btrdb package supports nanosecond-timestamp queries and insertion through Stream.insert, PyArrow tables, or pandas DataFrames. Statistical queries can return minimum, maximum, mean, count, and standard deviation without reading every underlying point. StreamSets can be filtered and converted to pandas DataFrames or CSV files. Documentation shows Dash and Plotly visualization, Ray-worker serialization, TLS on port 4411, and API-key authentication requiring that TLS port. Arrow-based tabular methods are documented for commercial customers with Arrow-enabled servers. BTrDB reports measured cluster throughput and compression figures, and describes smart-grid aggregation at city or country scale as one target use case.

FiloDB is a free, Apache-2.0 open-source database for distributed, real-time time-series, event, and operational data. It is designed to ingest and query many millions of time series, sharding data across processes with distributed querying. Columnar compression and off-heap memory management support substantial in-memory storage while reducing garbage collection. PromQL is available through an HTTP API and command-line interface; the README describes support as about 60%. Histograms are a first-class column type, with the project reporting possible compression and query-speed improvements of up to two orders of magnitude. Typical ingestion uses Apache Kafka, while the Gateway accepts Influx Line Protocol over TCP. Grafana can connect through the Prometheus-compatible API, and Cassandra is identified as the persistence store. Deployment uses clustered standalone processes and requires Java, Cassandra, Kafka, Rust, a C compiler, and SBT to build.