Comparison
Data residencyvsRetention policy
Data residency
the EU customers' rows may not physically leave the EU, so your one global warehouse is no longer allowed.
A requirement that data be stored, and sometimes processed, within a particular jurisdiction. It is the constraint most likely to force an architecture change late, because it invalidates the natural design of one central warehouse. The usual accommodations — regional storage with aggregate-only replication, or per-region pipelines with a federated query layer — both cost significantly more than the single-region version they replace.
Full entry →Retention policy
raw events older than thirteen months are deleted automatically, and nobody has to remember to do it.
A rule for how long each class of data is kept and what happens then. It is simultaneously a cost control, a legal obligation and a risk reduction, since data you no longer hold cannot leak. It is nearly always easier to implement at the partition level than by row, which is another reason partitioning by time is worth doing before you need it.
Full entry →