You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@gluten.apache.org by "j7nhai (via GitHub)" <gi...@apache.org> on 2025/03/28 03:06:25 UTC

[I] Shuffle write data size is larger for sort-based columnar shuffle [incubator-gluten]

j7nhai opened a new issue, #9163:
URL: https://github.com/apache/incubator-gluten/issues/9163

   ### Backend
   
   VL (Velox)
   
   ### Bug description
   
   We found that hash columnar shuffle performed poorly and generated too many disk files. 
   
   So we used **sort columnar shuffle**, but it still generated more shuffle write data than vanilla spark in the same stage. For example, for the same stage, the performance is as follows:
   
   gluten:
   ```
   Resource Profile Id: 0
   Total Time Across All Tasks: 123.4 h
   Locality Level Summary: Process local: 10604
   Input Size / Records: 4.3 TiB / 24137351937
   Shuffle Write Size / Records: 4.0 TiB / 24137134849
   Spill (Memory): 39.8 TiB
   Spill (Disk): 3.5 TiB
   Associated Job Ids: 0 
   ```
   
   vanilla spark:
   ```
   Resource Profile Id: 0
   Total Time Across All Tasks: 102.0 h
   Locality Level Summary: Process local: 10492
   Input Size / Records: 4.0 TiB / 24083129089
   Shuffle Write Size / Records: 2.6 TiB / 24083129089
   Spill (Memory): 42.5 TiB
   Spill (Disk): 2.2 TiB
   Associated Job Ids: 0 
   ```
   
   ### Spark version
   
   None
   
   ### Spark configurations
   
   _No response_
   
   ### System information
   
   _No response_
   
   ### Relevant logs
   
   ```bash
   
   ```


-- 
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: commits-unsubscribe@gluten.apache.org.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@gluten.apache.org
For additional commands, e-mail: commits-help@gluten.apache.org


Re: [I] [VL] Shuffle write data size is larger for sort-based columnar shuffle [incubator-gluten]

Posted by "marin-ma (via GitHub)" <gi...@apache.org>.
marin-ma commented on issue #9163:
URL: https://github.com/apache/incubator-gluten/issues/9163#issuecomment-2830864517

   @j7nhai Could you help to check if the issue is fixed?


-- 
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: commits-unsubscribe@gluten.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@gluten.apache.org
For additional commands, e-mail: commits-help@gluten.apache.org


Re: [I] [VL] Shuffle write data size is larger for sort-based columnar shuffle [incubator-gluten]

Posted by "M1zukiqwq (via GitHub)" <gi...@apache.org>.
M1zukiqwq commented on issue #9163:
URL: https://github.com/apache/incubator-gluten/issues/9163#issuecomment-3088362137

   @j7nhai hi, i want to know how do you set to use sort-based shuffle, did you set it by spark.gluten.sql.columnar.shuffle.sort.threshold?


-- 
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: commits-unsubscribe@gluten.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@gluten.apache.org
For additional commands, e-mail: commits-help@gluten.apache.org


Re: [I] [VL] Shuffle write data size is larger for sort-based columnar shuffle [incubator-gluten]

Posted by "j7nhai (via GitHub)" <gi...@apache.org>.
j7nhai closed issue #9163: [VL] Shuffle write data size is larger for sort-based columnar shuffle
URL: https://github.com/apache/incubator-gluten/issues/9163


-- 
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: commits-unsubscribe@gluten.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@gluten.apache.org
For additional commands, e-mail: commits-help@gluten.apache.org


Re: [I] [VL] Shuffle write data size is larger for sort-based columnar shuffle [incubator-gluten]

Posted by "j7nhai (via GitHub)" <gi...@apache.org>.
j7nhai commented on issue #9163:
URL: https://github.com/apache/incubator-gluten/issues/9163#issuecomment-2833322788

   This issue is fixed now 👍 


-- 
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: commits-unsubscribe@gluten.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@gluten.apache.org
For additional commands, e-mail: commits-help@gluten.apache.org


Re: [I] [VL] Shuffle write data size is larger for sort-based columnar shuffle [incubator-gluten]

Posted by "marin-ma (via GitHub)" <gi...@apache.org>.
marin-ma commented on issue #9163:
URL: https://github.com/apache/incubator-gluten/issues/9163#issuecomment-3088396350

   @M1zukiqwq You can set either `spark.gluten.sql.columnar.shuffle.sort.partitions.threshold` or `spark.gluten.sql.columnar.shuffle.sort.columns.threshold` 
   
   Please check https://github.com/apache/incubator-gluten/blob/main/docs/Configuration.md


-- 
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: commits-unsubscribe@gluten.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@gluten.apache.org
For additional commands, e-mail: commits-help@gluten.apache.org