How TimescaleDB compresses time-series data

TimescaleDB achieves compression ratios of up to 98% for time-series data using its hypercore engine, a hybrid row-columnar system that employs specialized algorithms like delta encoding, delta-of-delta, and Gorilla XOR—fundamentally different from PostgreSQL's general-purpose TOAST compression. Unlike TOAST, which handles individual large values, TimescaleDB compression exploits cross-row patterns specific to time-series data, enabling typical compression ratios of 10-100× for sensor data and timestamps. The system automatically converts older row-based chunks into compressed columnar format, improving both storage efficiency and query performance through sparse indexing and vectorized execution.