Afaik, Clickbench requires times to include the transfer of the results back to the clients.
From https://github.com/ClickHouse/ClickBench#output-suppression:
As an end-to-end benchmark, ClickBench submissions should measure back-to-back runtimes, i.e. the combined time
- to send a query from the client (e.g. a native client tool, HTTP, etc.),
- process the query on the server, and
- send the query results back to the client.
Afaict, polars benchmark scripts only includes the query processing in its timings, though.
Is that fine?
Afaik, Clickbench requires times to include the transfer of the results back to the clients.
From https://github.com/ClickHouse/ClickBench#output-suppression:
Afaict, polars benchmark scripts only includes the query processing in its timings, though.
Is that fine?