zenoyang opened a new issue, #10083:
URL: https://github.com/apache/incubator-doris/issues/10083

   ### Search before asking
   
   - [X] I had searched in the 
[issues](https://github.com/apache/incubator-doris/issues?q=is%3Aissue) and 
found no similar issues.
   
   
   ### Description
   
   In our production environment, for 100 million data volumes and cardinality 
of Bitmaps is ten million, for precise deduplication queries, the bottleneck is 
always in the Streaming Aggregate stage, which is single-threaded, so we hope 
to improve the parallelism of aggregation.
   For aggregated tables, when the number of key columns by group by is small, 
it is an easy way to push down aggregates to the Scanner thread because 
Scanners are executed in parallel.
   
   **Performance Testing**
   1. production data:
   
   Test env: 5 BE
   tbl has 80 million rows of data, 100 Bucket
   ```sql
   select k1,
      count(distinct device_id)        
   from tbl1
   group by k1;
   ```
   before: 312.2 s
   after aggregation pushdown: 71.48 s
   
   
   2. data comes from the SSB dataset:
   
   Test environment: 3 BE
   lineorder_bitmap_v1 has 10 Bucket
   Test sql:
   ```sql
   select LO_ORDERPRIORITY, LO_SHIPMODE, count(distinct LO_ORDERKEY)
   from lineorder_bitmap_v1
   group by LO_ORDERPRIORITY, LO_SHIPMODE;
   ```
   
   before: 76.69 s
   after: 23.88s
   
   ### Solution
   
   _No response_
   
   ### Are you willing to submit PR?
   
   - [X] Yes I am willing to submit a PR!
   
   ### Code of Conduct
   
   - [X] I agree to follow this project's [Code of 
Conduct](https://www.apache.org/foundation/policies/conduct)
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to