Columnar storage organizes time-series data for analytics over large or historical intervals. When indexed row-store access becomes less efficient, scanning the required partitions can be faster. section 9.7.2
Apply the layout selectively:
- Keep recent measurements in row storage, where indexes support current operational queries. section 9.7.2
- Move partitions older than a chosen retention window into columnar storage for historical analysis. section 9.7.2
- Leave older columnar partitions unindexed when full scans are more efficient than B-tree or BRIN access. section 9.7.2
Recent row data, historical columnar data
A time-series application can retain the most recent three months of measurements in the row store with indexes, then move older partitions to the TimescaleDB columnar store for month- or year-scale analysis. section 9.7.2
