[jira] [Resolved] (SPARK-37099) Introduce a rank-based filter to optimize top-k computation

["Wenchen Fan (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/SPARK-37099?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Wenchen Fan resolved SPARK-37099.
---------------------------------
    Fix Version/s: 3.5.0
       Resolution: Fixed

Issue resolved by pull request 38799
[https://github.com/apache/spark/pull/38799]

> Introduce a rank-based filter to optimize top-k computation
> -----------------------------------------------------------
>
>                 Key: SPARK-37099
>                 URL: https://issues.apache.org/jira/browse/SPARK-37099
>             Project: Spark
>          Issue Type: Improvement
>          Components: SQL
>    Affects Versions: 3.4.0
>            Reporter: Ruifeng Zheng
>            Priority: Major
>             Fix For: 3.5.0
>
>         Attachments: q67.png, q67_optimized.png, skewed_window.png
>
>
> in JD, we found that more than 90% usage of window function follows this pattern:
> {code:java}
>  select (... (row_number|rank|dense_rank) () over( [partition by ...] order by ... ) as rn)
>     where rn (==|<|<=) k and other conditions{code}
>  
> However, existing physical plan is not optimum:
>  
> 1, we should select local top-k records within each partitions, and then compute the global top-k. this can help reduce the shuffle amount;
>  
> For these three rank functions (row_number|rank|dense_rank), the rank of a key computed on partitial dataset  is always <=  its final rank computed on the whole dataset. so we can safely discard rows with partitial rank > k, anywhere.
>  
>  
> 2, skewed-window: some partition is skewed and take a long time to finish computation.
>  
> A real-world skewed-window case in our system is attached.
>  



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org