Hypertable Partitioning
A hypertable is one logical table that the database transparently splits into many physical pieces called chunks. The pattern, popularized by time series extensions on relational engines, lets a normal SQL table scale to huge time series volumes without manual sharding.
Partitioning by time
The main split is by time interval. Each chunk holds rows for a time range, such as one day. Because most queries ask about a recent window, the engine reads only the chunks that overlap that window and skips the rest. This skipping is called chunk exclusion and it is the heart of hypertable speed.
Adding a space dimension
Large deployments also partition by a space key such as device id, so writes spread across chunks instead of all hitting the newest one. This relieves write contention.
- Time chunks keep recent data small and prunable.
- Space chunks spread load across many devices or hosts.
- Old chunks can be compressed or dropped as whole units.
Key idea
A hypertable hides many time and space chunks behind one table, so queries touch only relevant chunks and old data drops as whole units.