FAQ - Where are operations on HyperLogLog datatypes processed?
Hyperloglog datatypes can be used to form estimations of set membership across large datasets allowing simalarity, membership and cardinality of intersections or unions of sets to be returned to a client application. Where is the processing of operations on the HLL bin type done, server or client?
The processing of operations across HLL bin types is done completely server side. Only results are returned to the client. The operation (for example the union or intersection computation) happens with an HLL bin on a single record and will take place on the server node holding the partition for that record. The second participant (argument) in the operation is provided by the client api as a list of HLLs. To make that list of HLLs, the application may fetch those from another (one or more) record’s HLL bins.
- For further details on the HLL related APIs, refer to the HyperLogLog documentation page.
HYPERLOGLOG PROCESS SERVER CLIENT