ClickHouse / ClickHouse/ClickHouse
Approximate GROUP BY with ORDER BY ... LIMIT distributed query.
Open
comp-distributed
performance
- Dominant language
- C++
- Stars
- 49.9k
- Forks
- 9k
- Avg merge
- 21h 32m
- Merged PRs (30d)
- 515
Description
When doing distributed query with
GROUP BY ... ORDER BY ... LIMIT ...
we can get approximate result if perform ORDER BY and LIMIT on every shard and then merge filtered data on the initiating node.
It will send less data over network making query tremendously faster.
The results can be accurate enough and it will be useful to make this behaviour available under a setting.
The result will be accurate if top keys appeared to be the same on every shard.
We can also allow to tune the accuracy by multiplying limit to some value.
See also:
\- `max_rows_to_group_by` and `group_by_overflow_mode = 'any'`;
Contributor guide
Assessment
This issue has not been assessed yet.