There is no dedicated data compression at the moment as our use-case is centered around performance. That being said, we store data very efficiently with minimal overhead.
One interesting consequence is that in many cases, we don't require indexes where other databases do. This synthetically compress the data relative to other DBs by removing the space taken by these indexes while improving query speed. It also means ingestion speed remains fast with O(1) complexity.
We don't store deltas, timestamps are stored as 64-bit int back to back in column. So are other primitives actually.
To avoid using indexes we store timestamps in ascending order. Timestamp interval search uses partitioning to cull chunks of data before lifting data in memory. Once in memory we use binary search on ordered timestamps to find intervals.
That said we do support indexes for key-based lookups
One interesting consequence is that in many cases, we don't require indexes where other databases do. This synthetically compress the data relative to other DBs by removing the space taken by these indexes while improving query speed. It also means ingestion speed remains fast with O(1) complexity.