Time-Series Databases
Tick data demands columnar storage, and the modern choices trade query latency against retention cost and portability.
Tick data is the stress test of storage: billions of rows, append-heavy, queried by time range and symbol. kdb+ and its q language remain the incumbent, a columnar database paired with a vector programming language that still anchors many bank trading desks. The open-source answer is columnar storage without the license: QuestDB and ClickHouse ingest ticks at millions of rows per second with time-partitioned tables, while Parquet files on object storage have become the default for historical research sets because they are cheap, immutable, and queryable directly from the Polars and DuckDB world. The choice turns on latency, retention, and how research queries production archives.