Basekick-Labs/arc

What it solves

Arc is a SQL-native time-series database designed for high-throughput telemetry data. It solves the problem of "retention cliffs"—where teams are forced to delete or downsample old data due to storage costs—by using an open Parquet storage engine that allows teams to keep full-resolution historical data on their own infrastructure (local disk, S3, Azure, or MinIO) without splitting their hot and cold data paths.

How it works

Arc operates as a single statically-linked binary that integrates an ingestion pipeline, a Parquet storage engine, and a SQL query layer powered by DuckDB. It supports multiple ingestion protocols, including InfluxDB Line Protocol and MessagePack, and automatically merges small Parquet files into larger, optimized ones through a background compaction system to reduce storage size and increase query speed. It provides analytical SQL capabilities (CTEs, window functions, joins) and manages the data lifecycle via retention policies and continuous queries.

Who it’s for

It is primarily built for aerospace telemetry but is suitable for any environment where machines continuously produce data, such as industrial IoT, manufacturing, energy, fleet telemetry, and general observability.

Highlights

• High-performance ingestion: Capable of ingesting over 34 million records per second using MessagePack Columnar.
• Open storage: Stores data as standard Apache Parquet files, preventing vendor lock-in.
• SQL-native: Uses standard analytical SQL instead of proprietary query languages.
• Zero dependencies: Deploys as a single binary with no requirement for JVM, Python, or external clusters.
• Broad compatibility: Compatible with InfluxDB Line Protocol and Telegraf, allowing for low-risk migrations.
• Enterprise-ready: Includes an optional FIPS 140-3 build for regulated environments and supports S3/Azure storage backends.