Advanced
How Arc works underneath: data-time partitioning, the optional ingestion WAL, scheduled file compaction, the query-path caches, and edge-to-hub file sync.
These pages explain mechanisms rather than tasks — useful when tuning a deployment or working out why Arc behaved the way it did.
Data-time partitioning
Why files are partitioned by event time, not ingestion time.
Write-ahead log
The optional WAL that bounds data loss on a crash.
File compaction
Merging small Parquet files on an hourly and daily schedule.
Query caching
The three caches on the query path and what each memoizes.
Edge sync
Shipping Parquet files from an edge spoke to a central hub.